From 97a9376a341dfa5e252b24fb839e5c07aec8a531 Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Wed, 4 Nov 2026 13:07:27 +0800 Subject: [PATCH 1/6] docs(minimax): document the H3 Max editing nodes and templates --- .../partner-nodes/minimax/minimax-h3.mdx | 86 +++++++++++++++++-- .../partner-nodes/minimax/minimax-h3.mdx | 86 +++++++++++++++++-- .../partner-nodes/minimax/minimax-h3.mdx | 86 +++++++++++++++++-- .../partner-nodes/minimax/minimax-h3.mdx | 82 +++++++++++++++++- .../partner-nodes/minimax/minimax-h3.mdx | 86 +++++++++++++++++-- 5 files changed, 403 insertions(+), 23 deletions(-) diff --git a/es/tutorials/partner-nodes/minimax/minimax-h3.mdx b/es/tutorials/partner-nodes/minimax/minimax-h3.mdx index 253870ba17..cc47321d42 100644 --- a/es/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/es/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -2,13 +2,13 @@ title: "Generación de video con la API de MiniMax H3 en ComfyUI" description: "Usa los nodos API de MiniMax H3 en ComfyUI para flujos en la nube de texto a video, primer/último fotograma y referencia a video con audio estéreo nativo." sidebarTitle: "Minimax H3" -translationSourceHash: b6d84062 +translationSourceHash: f247dec4 translationFrom: tutorials/partner-nodes/minimax/minimax-h3.mdx translationBlockHashes: "_intro": a261dbf5 "What MiniMax H3 is good at": 4c84d1e7 "Example outputs": 77256fd5 - "MiniMax H3 API workflows": 9f3a305f + "MiniMax H3 API workflows": 272af121 --- import ReqHint from "/snippets/es/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/es/tutorials/update-reminder.mdx"; @@ -51,19 +51,20 @@ Generación de primer y último fotograma, con el modelo creando el movimiento e ## Flujos de trabajo de la API de MiniMax H3 -La biblioteca de plantillas incluye tres flujos de trabajo de ejemplo de API para el modelo base H3, cuatro para H3 Max y dos para H3 Max Turbo: +La biblioteca de plantillas incluye tres flujos de trabajo de ejemplo de API para el modelo base H3, seis para H3 Max (cuatro de generación y dos de edición) y dos para H3 Max Turbo: - **Text to Video**: Genera videos a partir de prompts de texto - **First-Last-Frame to Video**: Genera un video entre una imagen de primer y último fotograma - **Reference to Video**: Genera videos condicionados por imágenes, videos y audio de referencia - **H3 Max T2V / I2V / FLF2V / R2V**: Los mismos modos a través del modelo MiniMax H3 Max, más rápido, con mejora de prompt opcional y etapas de ampliación a 2K integradas en cada flujo de trabajo - **H3 Max Turbo T2V / I2V**: Texto a video y video de primer/último fotograma a través del modelo MiniMax H3 Max Turbo, el más rápido, con las mismas etapas opcionales de mejora de prompt y ampliación a 2K +- **H3 Max cambio de reparto / reencuadre**: Flujos de trabajo de edición que sustituyen a las personas de un clip terminado o lo ajustan a otra relación de aspecto sin recortar al sujeto ### MiniMax H3 Max -Las cuatro plantillas `api_minimax_h3_max_*` se ejecutan en el modelo MiniMax H3 Max a través de los mismos nodos API: los nodos Text to Video y First-Last-Frame recibieron una opción H3 Max en ComfyUI 0.34.2, y el nodo Reference to Video en 0.34.3. Actualiza ComfyUI a la versión 0.34.3 como mínimo para ejecutarlas todas. +Las seis plantillas `api_minimax_h3_max_*` se ejecutan en el modelo MiniMax H3 Max a través de los mismos nodos API: los nodos Text to Video y First-Last-Frame recibieron una opción H3 Max en ComfyUI 0.34.2, y el nodo Reference to Video en 0.34.3. Las plantillas de cambio de reparto y reencuadre usan los nodos de edición H3 Max añadidos en ComfyUI 0.44.0. Actualiza ComfyUI a la versión 0.44.0 como mínimo para ejecutarlas todas. -H3 Max genera más rápido que en tiempo real y cobra tarifas más bajas que el modelo base: 15.09 créditos/s a 480P y 24.14 créditos/s a 768P (consulta la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing)). Cada plantilla Max añade dos etapas opcionales detrás de un nodo Switch, ambas desactivadas de forma predeterminada: +H3 Max genera más rápido que en tiempo real y cobra tarifas más bajas que el modelo base: 15.09 créditos/s a 480P y 24.14 créditos/s a 768P (consulta la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing)). Cada plantilla Max de generación añade dos etapas opcionales detrás de un nodo Switch, ambas desactivadas de forma predeterminada: 1. **Prompt enhance**: Reescribe el prompt con el nodo MiniMax H3 Context IR antes de la generación. Activa el booleano `Prompt Enhance` para habilitarla 2. **Upscale to 2K**: Ejecuta el nodo Regenerar MiniMax H3 a 2K sobre el video de la primera pasada. Activa el booleano `Upscale to 2K` para habilitarla @@ -278,3 +279,78 @@ Genera videos condicionados por medios de referencia con el modelo H3 Max, más Imagen de referencia para el flujo de trabajo, o usa tus propias imágenes. + +#### MiniMax H3 Max: cambio de reparto + +Sustituye a las personas de un clip terminado sin volver a rodar. El nodo Recast conserva el movimiento, el encuadre, los cortes y el audio original, y genera las nuevas personas a partir de tus fotos de referencia. + + + + + + Abrir en Comfy Cloud + + + Descargar JSON o buscar "Cast Swap: MiniMax H3 Max" en la Biblioteca de plantillas + + + + + + Clip de origen para el flujo de trabajo, o usa el tuyo. + + + Foto de referencia de la nueva persona, o usa la tuya. + + + +Carga tu propio clip y conecta de una a cuatro fotos, una por cada persona nueva. De forma predeterminada las fotos sustituyen a las personas principales del clip de izquierda a derecha, así que conéctalas en el orden que quieras. El prompt es opcional: nombra las fotos como `@Image1` a `@Image4` e indica quién pasa a ser quién, además de lo que debe mantenerse igual. + +- **Clip de origen**: de 5 a 30 segundos, sin ningún plano de más de 15 segundos +- **Fotos de referencia**: al menos 256 x 256 píxeles, hasta cuatro +- **Resolución**: 768P o 1080P +- **Mejorar el realismo**: desactivado de forma predeterminada. Actívalo para obtener una textura de piel más natural y una mejor integración en la escena + +#### MiniMax H3 Max: reencuadre de relación de aspecto + +Ajusta un clip a otra relación de aspecto sin recortar al sujeto. El origen mantiene su posición, su ritmo y su audio en el nuevo lienzo, y el modelo genera la imagen de alrededor. + + + + + + Abrir en Comfy Cloud + + + Descargar JSON o buscar "Aspect Ratio Reframe: MiniMax H3 Max" en la Biblioteca de plantillas + + + + + + Clip de origen para el flujo de trabajo, o usa el tuyo. + + + +La demo incluida convierte un clip vertical en un encuadre 16:9. Carga tu propio clip, elige la relación de aspecto y la resolución de destino, y usa el prompt para describir qué ocupa la zona nueva, por ejemplo continuar el fondo que ya aparece a ambos lados del sujeto. + +- **Clip de origen**: de 0.5 a 300 segundos +- **Relación de aspecto**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 o 9:21 +- **Resolución**: 480P, 768P o 1080P + +#### Nodos de edición de MiniMax H3 Max + +Seis nodos de edición H3 Max parten de un video o de una imagen fija en lugar de un prompt y editan ese material. Dos de ellos vienen con las plantillas anteriores; los otros cuatro están en la biblioteca de nodos bajo **Partner Nodes > video > MiniMax**, listos para añadirlos a tu propio flujo de trabajo. + +| Nodo | Qué hace | Material de partida | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | Sustituye a las personas de un video por las de fotos de referencia, manteniendo el movimiento, la cámara, los cortes y el sonido del original | Video de 5 a 30 segundos, 1 a 4 fotos | +| **MiniMax H3 Max Extend Video** | Continúa un video con material nuevo descrito en el prompt | Video de 1.625 a 60 segundos | +| **MiniMax H3 Max Reframe Video** | Ajusta un video a una nueva relación de aspecto y genera el área de alrededor, manteniendo el ritmo y el sonido | Video de 0.5 a 300 segundos | +| **MiniMax H3 Max Insert Video** | Sustituye el tramo entre dos marcas de tiempo por una escena nueva | Video de hasta 60 segundos, más un tiempo de inicio y de reanudación | +| **MiniMax H3 Max Lip Sync** | Anima una imagen fija para que su boca siga una pista de audio | Una imagen y de 5 a 899 segundos de audio | +| **MiniMax H3 Max Camera Control** | Orbita la cámara alrededor del sujeto de una imagen fija | Una imagen, más un ángulo y una distancia finales | + +En los prompts de Recast, Extend Video e Insert Video, nombra las referencias conectadas como `@Image1`, `@Audio1` o `@Video1`, numeradas en el orden de conexión. El nodo Reference to Video acepta la misma forma abreviada para sus propias referencias. Una etiqueta que apunte más allá de las referencias conectadas detiene la ejecución con un error. + +Las tarifas por nodo y resolución están en la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing). diff --git a/ja/tutorials/partner-nodes/minimax/minimax-h3.mdx b/ja/tutorials/partner-nodes/minimax/minimax-h3.mdx index be25de34c8..4fd55eb881 100644 --- a/ja/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/ja/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -2,13 +2,13 @@ title: "ComfyUIでのMiniMax H3 APIビデオ生成" description: "ComfyUI で MiniMax H3 API ノードを使い、クラウドワークフローでテキストから動画、先頭・末尾フレームから動画、参照から動画を、ネイティブのステレオオーディオ付きで生成します。" sidebarTitle: "Minimax H3" -translationSourceHash: b6d84062 +translationSourceHash: f247dec4 translationFrom: tutorials/partner-nodes/minimax/minimax-h3.mdx translationBlockHashes: "_intro": a261dbf5 "What MiniMax H3 is good at": 4c84d1e7 "Example outputs": 77256fd5 - "MiniMax H3 API workflows": 9f3a305f + "MiniMax H3 API workflows": 272af121 --- import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; @@ -52,19 +52,20 @@ MiniMax H3 は、**ネイティブステレオオーディオ**付きのビデ ## MiniMax H3 API ワークフロー -テンプレートライブラリには、H3 ベースモデル向けに 3 つ、H3 Max 向けに 4 つ、H3 Max Turbo 向けに 2 つの API サンプルワークフローが含まれています: +テンプレートライブラリには、H3 ベースモデル向けに 3 つ、H3 Max 向けに 6 つ(生成 4 つ、編集 2 つ)、H3 Max Turbo 向けに 2 つの API サンプルワークフローが含まれています: - **テキストから動画生成**:テキストプロンプトから動画を生成します。 - **先頭・末尾フレームから動画生成**:先頭フレームと末尾フレームの画像から、その間の動画を生成します。 - **リファレンスから動画生成**:参照画像、動画、オーディオを条件として動画を生成します。 - **H3 Max T2V / I2V / FLF2V / R2V**:より高速な MiniMax H3 Max モデルで同じモードを実行します。各ワークフローには、任意のプロンプト拡張と 2K アップスケール段階が組み込まれています。 - **H3 Max Turbo T2V / I2V**:最速の MiniMax H3 Max Turbo モデルでテキストから動画生成と先頭・末尾フレーム動画を実行します。こちらにも任意のプロンプト拡張と 2K アップスケール段階が組み込まれています。 +- **H3 Max キャスト入れ替え / アスペクト比リフレーム**:完成したクリップの人物を入れ替えたり、被写体を切り取らずに新しいアスペクト比に合わせたりする編集ワークフローです。 ### MiniMax H3 Max -4 つの `api_minimax_h3_max_*` テンプレートは、同じ API ノードを通じて MiniMax H3 Max モデルで実行されます。Text to Video ノードと First-Last-Frame ノードには ComfyUI 0.34.2 で H3 Max オプションが追加され、Reference to Video ノードには 0.34.3 で追加されました。すべて実行するには、ComfyUI を少なくとも 0.34.3 にアップデートしてください。 +6 つの `api_minimax_h3_max_*` テンプレートは、同じ API ノードを通じて MiniMax H3 Max モデルで実行されます。Text to Video ノードと First-Last-Frame ノードには ComfyUI 0.34.2 で H3 Max オプションが追加され、Reference to Video ノードには 0.34.3 で追加されました。キャスト入れ替えとアスペクト比リフレームのテンプレートは、ComfyUI 0.44.0 で追加された H3 Max 編集ノードを使用します。すべて実行するには、ComfyUI を少なくとも 0.44.0 にアップデートしてください。 -H3 Max はリアルタイムよりも速く生成でき、ベースモデルより低いレートで課金されます:480P で 15.09 クレジット/秒、768P で 24.14 クレジット/秒([Minimax 料金表](/ja/tutorials/partner-nodes/pricing) を参照)。各 Max テンプレートには、Switch ノードの背後に 2 つの任意段階があり、どちらもデフォルトでオフです: +H3 Max はリアルタイムよりも速く生成でき、ベースモデルより低いレートで課金されます:480P で 15.09 クレジット/秒、768P で 24.14 クレジット/秒([Minimax 料金表](/ja/tutorials/partner-nodes/pricing) を参照)。各 Max 生成テンプレートには、Switch ノードの背後に 2 つの任意段階があり、どちらもデフォルトでオフです: 1. **プロンプト拡張**:生成前に MiniMax H3 Context IR ノードでプロンプトを書き換えます。`Prompt Enhance` ブール値をオンにすると有効になります 2. **2K へのアップスケール**:最初のパスのビデオに対して MiniMax H3 Regenerate to 2K ノードを実行します。`Upscale to 2K` ブール値をオンにすると有効になります @@ -279,3 +280,78 @@ MiniMaxは、参照駆動型生成(R2V)向けの公式の[フル参照モー ワークフロー用の参照画像です。または、ご自身の画像を使用できます。 + +#### MiniMax H3 Max: キャスト入れ替え + +撮り直しなしで、完成したクリップの人物を入れ替えます。Recast ノードは動き、構図、カット、元のオーディオをそのまま保ち、参照写真から新しい人物を生成します。 + + + + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで「Cast Swap: MiniMax H3 Max」を検索してください + + + + + + ワークフロー用のソースクリップです。または、ご自身の動画を使用できます。 + + + 新しい人物の参照写真です。または、ご自身の写真を使用できます。 + + + +クリップを読み込み、新しい人物ごとに 1 枚、最大 4 枚の写真を接続します。既定では写真がクリップ内の主要な人物を左から右へ置き換えるため、希望する順序で接続してください。プロンプトは任意です。写真を `@Image1` から `@Image4` で指定し、誰が誰になるか、変わってはいけない点を書きます。 + +- **ソースクリップ**: 5〜30 秒、1 ショットは 15 秒以下 +- **参照写真**: 256 x 256 ピクセル以上、最大 4 枚 +- **解像度**: 768P または 1080P +- **リアリズムを強化**: 既定ではオフ。オンにすると肌の質感が自然になり、新しい人物がシーンになじみます + +#### MiniMax H3 Max: アスペクト比のリフレーム + +被写体を切り取らずに、クリップを別のアスペクト比に合わせます。ソースは新しいキャンバス上で位置、タイミング、オーディオを保ち、周囲の映像はモデルが生成します。 + + + + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで「Aspect Ratio Reframe: MiniMax H3 Max」を検索してください + + + + + + ワークフロー用のソースクリップです。または、ご自身の動画を使用できます。 + + + +同梱のデモは縦向きのクリップを 16:9 のフレームに変換します。ご自身のクリップを読み込み、目標のアスペクト比と解像度を選び、新しく加わる領域の内容をプロンプトで説明してください。たとえば、被写体の両側にある背景をそのまま続けるよう指定します。 + +- **ソースクリップ**: 0.5〜300 秒 +- **アスペクト比**: 16:9、9:16、1:1、4:3、3:4、21:9、9:21 +- **解像度**: 480P、768P、1080P + +#### MiniMax H3 Max 編集ノード + +6 つの H3 Max 編集ノードは、プロンプトだけでなく動画や静止画を起点として、その素材を編集します。うち 2 つは上記のテンプレートに含まれており、残りの 4 つはノードライブラリの **Partner Nodes > video > MiniMax** にあり、ご自身のワークフローに追加できます。 + +| ノード | 動作 | 入力素材 | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | 参照写真の人物で動画内の人物を置き換え、元動画の動き、カメラ、カット、音を保ちます | 5〜30 秒の動画、写真 1〜4 枚 | +| **MiniMax H3 Max Extend Video** | プロンプトで説明した新しい映像で動画を延長します | 1.625〜60 秒の動画 | +| **MiniMax H3 Max Reframe Video** | 動画を新しいアスペクト比に収め、周囲の領域を生成します。タイミングと音は保たれます | 0.5〜300 秒の動画 | +| **MiniMax H3 Max Insert Video** | 2 つのタイムスタンプの間を新しく生成したシーンに置き換えます | 最大 60 秒の動画と、開始・再開の時刻 | +| **MiniMax H3 Max Lip Sync** | 静止画の口の動きをオーディオに合わせます | 画像 1 枚、5〜899 秒のオーディオ | +| **MiniMax H3 Max Camera Control** | 静止画の被写体の周りでカメラを回り込ませます | 画像 1 枚と、終了角度・距離 | + +Recast、Extend Video、Insert Video のプロンプトでは、接続した参照素材を `@Image1`、`@Audio1`、`@Video1` のように接続順の番号で指定します。Reference to Video ノードも同じ省略記法を受け付けます。接続数を超えるタグを指定すると、実行はエラーで停止します。 + +ノードと解像度ごとの料金は [Minimax 料金表](/ja/tutorials/partner-nodes/pricing) にあります。 diff --git a/ko/tutorials/partner-nodes/minimax/minimax-h3.mdx b/ko/tutorials/partner-nodes/minimax/minimax-h3.mdx index d16b975973..25db2ae72d 100644 --- a/ko/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/ko/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -2,13 +2,13 @@ title: "ComfyUI에서 MiniMax H3 API 비디오 생성" description: "ComfyUI에서 MiniMax H3 API 노드로 클라우드 워크플로를 실행해 텍스트, 첫/마지막 프레임, 레퍼런스 기반 비디오를 네이티브 스테레오 오디오와 함께 생성하세요." sidebarTitle: "MiniMax H3" -translationSourceHash: b6d84062 +translationSourceHash: f247dec4 translationFrom: tutorials/partner-nodes/minimax/minimax-h3.mdx translationBlockHashes: "_intro": a261dbf5 "What MiniMax H3 is good at": 4c84d1e7 "Example outputs": 77256fd5 - "MiniMax H3 API workflows": 9f3a305f + "MiniMax H3 API workflows": 272af121 --- import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -51,19 +51,20 @@ MiniMax H3는 **네이티브 스테레오 오디오**로 비디오를 생성합 ## MiniMax H3 API 워크플로 -템플릿 라이브러리에는 H3 베이스 모델용 API 예제 워크플로 3개, H3 Max용 4개, H3 Max Turbo용 2개가 포함되어 있습니다: +템플릿 라이브러리에는 H3 베이스 모델용 API 예제 워크플로 3개, H3 Max용 6개(생성 4개, 편집 2개), H3 Max Turbo용 2개가 포함되어 있습니다: - **텍스트 기반 비디오 생성**: 텍스트 프롬프트에서 비디오를 생성합니다. - **첫/마지막 프레임 기반 비디오 생성**: 첫 번째 프레임 이미지와 마지막 프레임 이미지 사이의 비디오를 생성합니다. - **레퍼런스 기반 비디오 생성**: 레퍼런스 이미지, 비디오 및 오디오를 조건으로 비디오를 생성합니다. - **H3 Max T2V / I2V / FLF2V / R2V**: 더 빠른 MiniMax H3 Max 모델로 동일한 모드를 실행합니다. 각 워크플로에는 선택적 프롬프트 개선 및 2K 업스케일 단계가 내장되어 있습니다. - **H3 Max Turbo T2V / I2V**: 가장 빠른 MiniMax H3 Max Turbo 모델로 텍스트 기반 비디오 생성과 첫/마지막 프레임 비디오를 실행합니다. 이 역시 선택적 프롬프트 개선 및 2K 업스케일 단계가 내장되어 있습니다. +- **H3 Max 캐스트 교체 / 화면 비율 리프레임**: 완성된 클립의 인물을 교체하거나 피사체를 자르지 않고 새로운 화면 비율에 맞추는 편집 워크플로입니다. ### MiniMax H3 Max -4개의 `api_minimax_h3_max_*` 템플릿은 동일한 API 노드를 통해 MiniMax H3 Max 모델에서 실행됩니다. Text to Video 및 First-Last-Frame 노드에는 ComfyUI 0.34.2에서, Reference to Video 노드에는 0.34.3에서 H3 Max 옵션이 추가되었습니다. 모두 실행하려면 ComfyUI를 최소 0.34.3으로 업데이트하세요. +6개의 `api_minimax_h3_max_*` 템플릿은 동일한 API 노드를 통해 MiniMax H3 Max 모델에서 실행됩니다. Text to Video 및 First-Last-Frame 노드에는 ComfyUI 0.34.2에서, Reference to Video 노드에는 0.34.3에서 H3 Max 옵션이 추가되었습니다. 캐스트 교체와 화면 비율 리프레임 템플릿은 ComfyUI 0.44.0에서 추가된 H3 Max 편집 노드를 사용합니다. 모두 실행하려면 ComfyUI를 최소 0.44.0으로 업데이트하세요. -H3 Max는 실시간보다 빠르게 생성되며 베이스 모델보다 낮은 요율로 청구됩니다: 480P는 15.09 크레딧/초, 768P는 24.14 크레딧/초입니다([Minimax 요금표](/ko/tutorials/partner-nodes/pricing) 참조). 각 Max 템플릿에는 Switch 노드 뒤에 선택적 단계 2개가 있으며, 둘 다 기본적으로 꺼져 있습니다: +H3 Max는 실시간보다 빠르게 생성되며 베이스 모델보다 낮은 요율로 청구됩니다: 480P는 15.09 크레딧/초, 768P는 24.14 크레딧/초입니다([Minimax 요금표](/ko/tutorials/partner-nodes/pricing) 참조). 각 Max 생성 템플릿에는 Switch 노드 뒤에 선택적 단계 2개가 있으며, 둘 다 기본적으로 꺼져 있습니다: 1. **프롬프트 개선**: 생성 전에 MiniMax H3 Context IR 노드로 프롬프트를 다시 씁니다. `Prompt Enhance` 불리언을 켜면 활성화됩니다 2. **2K로 업스케일**: 첫 번째 패스 비디오에 MiniMax H3 Regenerate to 2K 노드를 실행합니다. `Upscale to 2K` 불리언을 켜면 활성화됩니다 @@ -278,3 +279,78 @@ MiniMax는 레퍼런스 기반 생성(R2V)을 위한 공식 [전체 레퍼런스 워크플로용 레퍼런스 이미지입니다. 또는 직접 만든 이미지를 사용하세요. + +#### MiniMax H3 Max: 캐스트 교체 + +다시 촬영하지 않고 완성된 클립의 인물을 교체합니다. Recast 노드는 움직임, 구도, 컷, 원본 오디오를 그대로 유지하고 참조 사진으로 새 인물을 생성합니다. + + + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Cast Swap: MiniMax H3 Max" 검색 + + + + + + 워크플로용 소스 클립입니다. 또는 직접 만든 영상을 사용하세요. + + + 새 인물의 참조 사진입니다. 또는 직접 만든 사진을 사용하세요. + + + +클립을 불러오고 새 인물마다 한 장씩, 최대 네 장의 사진을 연결합니다. 기본 설정에서는 사진이 클립의 주요 인물을 왼쪽에서 오른쪽으로 교체하므로 원하는 순서대로 연결하세요. 프롬프트는 선택 사항입니다. 사진을 `@Image1`부터 `@Image4`로 지칭하고 누가 누구로 바뀌는지, 무엇이 그대로여야 하는지 적습니다. + +- **소스 클립**: 5~30초, 한 샷은 15초 이하 +- **참조 사진**: 최소 256 x 256 픽셀, 최대 네 장 +- **해상도**: 768P 또는 1080P +- **사실감 강화**: 기본적으로 꺼져 있습니다. 켜면 피부 질감이 더 자연스럽고 새 인물이 장면에 더 잘 어울립니다 + +#### MiniMax H3 Max: 화면 비율 리프레임 + +피사체를 자르지 않고 클립을 다른 화면 비율에 맞춥니다. 소스는 새 캔버스에서 위치, 타이밍, 오디오를 유지하고 주변 화면은 모델이 생성합니다. + + + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Aspect Ratio Reframe: MiniMax H3 Max" 검색 + + + + + + 워크플로용 소스 클립입니다. 또는 직접 만든 영상을 사용하세요. + + + +포함된 데모는 세로 클립을 16:9 프레임으로 바꿉니다. 직접 만든 클립을 불러오고 목표 화면 비율과 해상도를 선택한 뒤, 새로 채워질 영역을 프롬프트로 설명하세요. 예를 들어 피사체 양옆에 이미 있는 배경을 이어서 그리도록 지정합니다. + +- **소스 클립**: 0.5~300초 +- **화면 비율**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, 9:21 +- **해상도**: 480P, 768P, 1080P + +#### MiniMax H3 Max 편집 노드 + +여섯 H3 Max 편집 노드는 프롬프트만이 아니라 비디오나 정지 이미지에서 시작해 그 소재를 편집합니다. 이 중 두 개는 위의 템플릿에 포함되어 있고, 나머지 네 개는 노드 라이브러리의 **Partner Nodes > video > MiniMax** 아래에 있어 직접 워크플로에 추가할 수 있습니다. + +| 노드 | 기능 | 시작 소재 | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | 참조 사진의 인물로 비디오 속 인물을 교체하며, 원본의 움직임, 카메라, 컷, 소리를 유지합니다 | 5~30초 비디오, 사진 1~4장 | +| **MiniMax H3 Max Extend Video** | 프롬프트로 설명한 새 장면으로 비디오를 연장합니다 | 1.625~60초 비디오 | +| **MiniMax H3 Max Reframe Video** | 비디오를 새 화면 비율에 맞추고 주변 영역을 생성하며, 타이밍과 소리는 유지합니다 | 0.5~300초 비디오 | +| **MiniMax H3 Max Insert Video** | 두 타임스탬프 사이 구간을 새로 생성한 장면으로 교체합니다 | 최대 60초 비디오와 시작/재개 시각 | +| **MiniMax H3 Max Lip Sync** | 정지 이미지의 입 모양을 오디오에 맞춰 움직입니다 | 이미지 1장, 5~899초 오디오 | +| **MiniMax H3 Max Camera Control** | 정지 이미지의 피사체 주위로 카메라를 회전시킵니다 | 이미지 1장과 종료 각도·거리 | + +Recast, Extend Video, Insert Video의 프롬프트에서는 연결한 참조 소재를 `@Image1`, `@Audio1`, `@Video1`처럼 연결 순서대로 지정합니다. Reference to Video 노드도 같은 축약 표기를 지원합니다. 연결한 개수를 넘는 태그를 쓰면 실행이 오류로 중단됩니다. + +노드와 해상도별 요율은 [Minimax 요금표](/ko/tutorials/partner-nodes/pricing)에 있습니다. diff --git a/tutorials/partner-nodes/minimax/minimax-h3.mdx b/tutorials/partner-nodes/minimax/minimax-h3.mdx index 8c54aecf0b..f8424cd9f2 100644 --- a/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -44,19 +44,20 @@ First-last-frame generation, with the model creating the motion between two fram ## MiniMax H3 API workflows -The template library ships with three API example workflows for the base H3 model, four for H3 Max, and two for H3 Max Turbo: +The template library ships with three API example workflows for the base H3 model, six for H3 Max (four generation, two editing), and two for H3 Max Turbo: - **Text to Video**: Generate videos from text prompts - **First-Last-Frame to Video**: Generate a video between a first and last frame image - **Reference to Video**: Generate videos conditioned on reference images, videos, and audio - **H3 Max T2V / I2V / FLF2V / R2V**: The same modes through the faster MiniMax H3 Max model, with optional prompt enhancement and 2K upscaling stages built into each workflow - **H3 Max Turbo T2V / I2V**: Text-to-video and first/last-frame video through the fastest MiniMax H3 Max Turbo model, with the same optional prompt enhancement and 2K upscaling stages +- **H3 Max Cast Swap / Aspect Ratio Reframe**: Editing workflows that swap the people in a finished clip, or fit it to a new aspect ratio without cropping the subject ### MiniMax H3 Max -The four `api_minimax_h3_max_*` templates run on the MiniMax H3 Max model through the same API nodes: the Text to Video and First-Last-Frame nodes gained an H3 Max option in ComfyUI 0.34.2, and the Reference to Video node in 0.34.3. Update ComfyUI to at least 0.34.3 to run all of them. +The six `api_minimax_h3_max_*` templates run on the MiniMax H3 Max model through the same API nodes: the Text to Video and First-Last-Frame nodes gained an H3 Max option in ComfyUI 0.34.2, and the Reference to Video node in 0.34.3. The Cast Swap and Aspect Ratio Reframe templates use the H3 Max editing nodes added in ComfyUI 0.44.0. Update ComfyUI to at least 0.44.0 to run all of them. -H3 Max generates faster than real time and bills at lower rates than the base model: 15.09 credits/sec at 480P and 24.14 credits/sec at 768P (see the [Minimax pricing table](/tutorials/partner-nodes/pricing)). Each Max template adds two optional stages behind a Switch node, both off by default: +H3 Max generates faster than real time and bills at lower rates than the base model: 15.09 credits/sec at 480P and 24.14 credits/sec at 768P (see the [Minimax pricing table](/tutorials/partner-nodes/pricing)). Each Max generation template adds two optional stages behind a Switch node, both off by default: 1. **Prompt enhance**: Rewrites the prompt with the MiniMax H3 Context IR node before generation. Turn on the `Prompt Enhance` boolean to enable it 2. **Upscale to 2K**: Runs the MiniMax H3 Regenerate to 2K node on the first-pass video. Turn on the `Upscale to 2K` boolean to enable it @@ -271,3 +272,78 @@ Generate videos conditioned on reference media on the faster H3 Max model. The R Reference image for the workflow, or use your own images. + +#### MiniMax H3 Max: Cast Swap + +Replace the people in a finished clip without a reshoot. The Recast node keeps the motion, the framing, the cuts and the original audio, and generates the new people from your reference photos. + + + + + + Open in Comfy Cloud + + + Download JSON or search "Cast Swap: MiniMax H3 Max" in Template Library + + + + + + Source clip for the workflow, or use your own. + + + Reference photo of the new person, or use your own. + + + +Load your own clip and connect one to four photos, one per new person. By default the photos replace the main people in the clip from left to right, so connect them in the order you want. The prompt is optional: name the photos as `@Image1` to `@Image4` and say who becomes whom, plus anything that has to stay the same. + +- **Source clip**: 5 to 30 seconds, with no single shot longer than 15 seconds +- **Reference photos**: at least 256 x 256 pixels, up to four +- **Resolution**: 768P or 1080P +- **Enhance realism**: off by default. Turn it on for more natural skin texture and a better blend into the scene + +#### MiniMax H3 Max: Aspect Ratio Reframe + +Fit a clip to a different aspect ratio without cropping the subject. The source keeps its position, timing and audio on the new canvas, and the model generates the surrounding picture. + + + + + + Open in Comfy Cloud + + + Download JSON or search "Aspect Ratio Reframe: MiniMax H3 Max" in Template Library + + + + + + Source clip for the workflow, or use your own. + + + +The bundled demo turns a vertical portrait clip into a 16:9 frame. Load your own clip, pick the target aspect ratio and resolution, and use the prompt to describe what fills the new area, for example continuing the background that is already behind the subject on both sides. + +- **Source clip**: 0.5 to 300 seconds +- **Aspect ratio**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 or 9:21 +- **Resolution**: 480P, 768P or 1080P + +#### MiniMax H3 Max editing nodes + +Six H3 Max editing nodes start from a video or a still image instead of a prompt alone and edit that material. Two of them ship with the templates above; the other four are in the node library under **Partner Nodes > video > MiniMax**, ready to drop into your own workflow. + +| Node | What it does | Starting media | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | Replaces the people in a video with people from reference photos, keeping the motion, camera, cuts and sound of the source | 5-30 second video, 1-4 photos | +| **MiniMax H3 Max Extend Video** | Continues a video with new footage described in the prompt | 1.625-60 second video | +| **MiniMax H3 Max Reframe Video** | Fits a video into a new aspect ratio and generates the area around it, keeping timing and sound | 0.5-300 second video | +| **MiniMax H3 Max Insert Video** | Replaces the section between two timestamps with a newly generated scene | Video up to 60 seconds, plus a start and resume time | +| **MiniMax H3 Max Lip Sync** | Animates a still image so the mouth follows an audio track | One image, 5-899 seconds of audio | +| **MiniMax H3 Max Camera Control** | Orbits the camera around the subject in a still image | One image, plus an end angle and distance | + +In the Recast, Extend Video and Insert Video prompts, refer to the connected references as `@Image1`, `@Audio1` or `@Video1`, numbered in connection order. The Reference to Video node accepts the same shorthand for its own references. A tag that points past the references you connected stops the run with an error. + +Rates per node and resolution are in the [Minimax pricing table](/tutorials/partner-nodes/pricing). diff --git a/zh/tutorials/partner-nodes/minimax/minimax-h3.mdx b/zh/tutorials/partner-nodes/minimax/minimax-h3.mdx index d376905d38..f6176bd5b1 100644 --- a/zh/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/zh/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -2,13 +2,13 @@ title: "在 ComfyUI 中使用 MiniMax H3 API 进行视频生成" description: "在 ComfyUI 中使用 MiniMax H3 API 节点,通过云端工作流进行文生视频、首尾帧视频和参考生视频,支持原生立体声音频。" sidebarTitle: "Minimax H3" -translationSourceHash: b6d84062 +translationSourceHash: f247dec4 translationFrom: tutorials/partner-nodes/minimax/minimax-h3.mdx translationBlockHashes: "_intro": a261dbf5 "What MiniMax H3 is good at": 4c84d1e7 "Example outputs": 77256fd5 - "MiniMax H3 API workflows": 9f3a305f + "MiniMax H3 API workflows": 272af121 --- import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -51,19 +51,20 @@ MiniMax H3 可生成带有**原生立体声音频**的视频:人声、音效 ## MiniMax H3 API 工作流 -模板库为 H3 基础模型随附三个 API 示例工作流,为 H3 Max 随附四个,为 H3 Max Turbo 随附两个: +模板库为 H3 基础模型随附三个 API 示例工作流,为 H3 Max 随附六个(四个生成、两个编辑),为 H3 Max Turbo 随附两个: - **文生视频**:根据文本提示词生成视频 - **首尾帧视频**:在首帧和末帧图像之间生成视频 - **参考生成视频**:根据参考图像、视频和音频生成视频 - **H3 Max T2V / I2V / FLF2V / R2V**:通过更快的 MiniMax H3 Max 模型运行相同的模式,每个工作流都内置了可选的提示词增强和 2K 上采样阶段 - **H3 Max Turbo T2V / I2V**:通过最快的 MiniMax H3 Max Turbo 模型运行文生视频与首尾帧视频,同样内置可选的提示词增强和 2K 上采样阶段 +- **H3 Max 角色替换 / 宽高比重构**:编辑工作流,用于替换已完成片段中的人物,或在不裁切主体的情况下把片段适配到新的宽高比 ### MiniMax H3 Max -这四个 `api_minimax_h3_max_*` 模板通过相同的 API 节点在 MiniMax H3 Max 模型上运行:Text to Video 和 First-Last-Frame 节点在 ComfyUI 0.34.2 中加入了 H3 Max 选项,Reference to Video 节点则在 0.34.3 中加入。请将 ComfyUI 更新到至少 0.34.3 以运行全部模板。 +这六个 `api_minimax_h3_max_*` 模板通过相同的 API 节点在 MiniMax H3 Max 模型上运行:Text to Video 和 First-Last-Frame 节点在 ComfyUI 0.34.2 中加入了 H3 Max 选项,Reference to Video 节点则在 0.34.3 中加入。角色替换和宽高比重构模板使用 ComfyUI 0.44.0 新增的 H3 Max 编辑节点。请将 ComfyUI 更新到至少 0.44.0 以运行全部模板。 -H3 Max 以快于实时的速度生成,费率低于基础模型:480P 为 15.09 积分/秒,768P 为 24.14 积分/秒(见 [Minimax 定价表](/zh/tutorials/partner-nodes/pricing))。每个 Max 模板在 Switch 节点后提供两个可选阶段,默认均为关闭: +H3 Max 以快于实时的速度生成,费率低于基础模型:480P 为 15.09 积分/秒,768P 为 24.14 积分/秒(见 [Minimax 定价表](/zh/tutorials/partner-nodes/pricing))。每个 Max 生成模板在 Switch 节点后提供两个可选阶段,默认均为关闭: 1. **提示词增强**:生成前使用 MiniMax H3 Context IR 节点改写提示词。打开 `Prompt Enhance` 布尔值即可启用 2. **上采样至 2K**:对首次生成的视频运行 MiniMax H3 Regenerate to 2K 节点。打开 `Upscale to 2K` 布尔值即可启用 @@ -278,3 +279,78 @@ MiniMax 为参考驱动生成(R2V)发布了一份官方的 [全参考模式 该工作流的参考图像,也可以使用你自己的图像。 + +#### MiniMax H3 Max:角色替换 + +无需重拍就能替换已完成片段中的人物。Recast 节点保留动作、构图、剪辑点和原始音频,并根据参考照片生成新的人物。 + + + + + + 在 Comfy Cloud 中打开 + + + 下载 JSON,或在模板库中搜索“Cast Swap: MiniMax H3 Max” + + + + + + 该工作流的源片段,也可以使用你自己的视频。 + + + 新人物的参考照片,也可以使用你自己的照片。 + + + +载入你自己的片段,并连接一到四张照片,每个新人物一张。默认情况下,照片会从左到右替换片段中的主要人物,所以请按你想要的顺序连接。提示词是可选的:用 `@Image1` 到 `@Image4` 指代照片,说明谁替换谁,以及哪些内容必须保持不变。 + +- **源片段**:5 到 30 秒,单个镜头不超过 15 秒 +- **参考照片**:至少 256 x 256 像素,最多四张 +- **分辨率**:768P 或 1080P +- **增强真实感**:默认关闭。开启后皮肤纹理更自然,新人物与场景融合得更好 + +#### MiniMax H3 Max:宽高比重构 + +在不裁切主体的情况下把片段适配到新的宽高比。源内容在新画布上保持位置、时长和音频,模型负责生成周围的画面。 + + + + + + 在 Comfy Cloud 中打开 + + + 下载 JSON,或在模板库中搜索“Aspect Ratio Reframe: MiniMax H3 Max” + + + + + + 该工作流的源片段,也可以使用你自己的视频。 + + + +自带示例把竖屏片段转换成 16:9 画面。载入你自己的片段,选择目标宽高比和分辨率,再用提示词描述新区域的内容,例如延续主体两侧已有的背景。 + +- **源片段**:0.5 到 300 秒 +- **宽高比**:16:9、9:16、1:1、4:3、3:4、21:9 或 9:21 +- **分辨率**:480P、768P 或 1080P + +#### MiniMax H3 Max 编辑节点 + +六个 H3 Max 编辑节点从视频或静态图像出发,而不是只靠提示词,并对这些素材进行编辑。其中两个已随上面的模板提供,另外四个位于节点库的 **Partner Nodes > video > MiniMax** 下,可以直接拖进你自己的工作流。 + +| 节点 | 作用 | 起始素材 | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | 用参考照片中的人物替换视频中的人物,保留原视频的动作、镜头、剪辑和声音 | 5-30 秒视频,1-4 张照片 | +| **MiniMax H3 Max Extend Video** | 按提示词描述的新内容延长视频 | 1.625-60 秒视频 | +| **MiniMax H3 Max Reframe Video** | 把视频适配到新的宽高比并生成周围区域,保留时长和声音 | 0.5-300 秒视频 | +| **MiniMax H3 Max Insert Video** | 用新生成的场景替换两个时间点之间的段落 | 最长 60 秒视频,另需开始和恢复时间 | +| **MiniMax H3 Max Lip Sync** | 让静态图像的口型跟随音频 | 一张图像,5-899 秒音频 | +| **MiniMax H3 Max Camera Control** | 围绕静态图像中的主体做镜头环绕 | 一张图像,加上结束角度和距离 | + +在 Recast、Extend Video 和 Insert Video 的提示词中,用 `@Image1`、`@Audio1` 或 `@Video1` 指代已连接的参考素材,编号按连接顺序。Reference to Video 节点也接受同样的简写形式。指向超出已连接数量的标签会让运行报错停止。 + +各节点和各分辨率的费率见 [Minimax 定价表](/zh/tutorials/partner-nodes/pricing)。 From e2dd0f10f456318ec2cf296a19706781e17e606d Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Wed, 4 Nov 2026 19:15:00 +0800 Subject: [PATCH 2/6] docs(minimax): move the H3 Max editing nodes to their own page The H3 page covers generation; the six editing nodes, the Cast Swap and Aspect Ratio Reframe templates, and their limits now live on tutorials/partner-nodes/minimax/minimax-h3-max-editing with all five locales and nav entries. --- docs.json | 15 ++- .../minimax/minimax-h3-max-editing.mdx | 96 +++++++++++++++++++ .../partner-nodes/minimax/minimax-h3.mdx | 88 ++--------------- .../minimax/minimax-h3-max-editing.mdx | 96 +++++++++++++++++++ .../partner-nodes/minimax/minimax-h3.mdx | 88 ++--------------- .../minimax/minimax-h3-max-editing.mdx | 96 +++++++++++++++++++ .../partner-nodes/minimax/minimax-h3.mdx | 88 ++--------------- .../minimax/minimax-h3-max-editing.mdx | 89 +++++++++++++++++ .../partner-nodes/minimax/minimax-h3.mdx | 84 +--------------- .../minimax/minimax-h3-max-editing.mdx | 96 +++++++++++++++++++ .../partner-nodes/minimax/minimax-h3.mdx | 88 ++--------------- 11 files changed, 516 insertions(+), 408 deletions(-) create mode 100644 es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx create mode 100644 ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx create mode 100644 ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx create mode 100644 tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx create mode 100644 zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx diff --git a/docs.json b/docs.json index 97cacfd89a..f8197f7110 100644 --- a/docs.json +++ b/docs.json @@ -805,7 +805,8 @@ "tutorials/partner-nodes/minimax/overview", "tutorials/partner-nodes/minimax/minimax-01", "tutorials/partner-nodes/minimax/minimax-s2v-01", - "tutorials/partner-nodes/minimax/minimax-h3" + "tutorials/partner-nodes/minimax/minimax-h3", + "tutorials/partner-nodes/minimax/minimax-h3-max-editing" ] }, { @@ -4916,7 +4917,8 @@ { "group": "MiniMax", "pages": [ - "zh/tutorials/partner-nodes/minimax/minimax-h3" + "zh/tutorials/partner-nodes/minimax/minimax-h3", + "zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing" ] }, { @@ -8995,7 +8997,8 @@ { "group": "MiniMax", "pages": [ - "ja/tutorials/partner-nodes/minimax/minimax-h3" + "ja/tutorials/partner-nodes/minimax/minimax-h3", + "ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing" ] }, { @@ -13091,7 +13094,8 @@ { "group": "MiniMax", "pages": [ - "ko/tutorials/partner-nodes/minimax/minimax-h3" + "ko/tutorials/partner-nodes/minimax/minimax-h3", + "ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing" ] }, { @@ -17124,7 +17128,8 @@ "es/tutorials/partner-nodes/minimax/overview", "es/tutorials/partner-nodes/minimax/minimax-01", "es/tutorials/partner-nodes/minimax/minimax-s2v-01", - "es/tutorials/partner-nodes/minimax/minimax-h3" + "es/tutorials/partner-nodes/minimax/minimax-h3", + "es/tutorials/partner-nodes/minimax/minimax-h3-max-editing" ] }, { diff --git a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx new file mode 100644 index 0000000000..6741912c6e --- /dev/null +++ b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -0,0 +1,96 @@ +--- +title: "Edición con MiniMax H3 Max en ComfyUI" +description: "Cambia el reparto, reencuadra, extiende o inserta tomas, sincroniza labios y orbita la cámara en clips finalizados con los nodos API de edición MiniMax H3 Max." +sidebarTitle: "Edición de Minimax H3 Max" +translationSourceHash: e1fd584f +translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +translationBlockHashes: + "_intro": f127db6f + "MiniMax H3 Max: Cast Swap": cd212e2a + "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 + "MiniMax H3 Max editing nodes": 43f08f03 +--- +import ReqHint from "/snippets/es/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/es/tutorials/update-reminder.mdx"; + +Seis nodos de edición de MiniMax H3 Max parten de un video o de una imagen fija en lugar de solo un prompt y editan ese material. Se ejecutan en los servidores de MiniMax a través de tu cuenta de Comfy API, igual que los demás nodos API de H3, y los nodos de edición se agregaron en ComfyUI 0.44.0. + +Dos de ellos incluyen plantillas de flujo de trabajo: **Cast Swap** (`api_minimax_h3_max_recast`) y **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`). Los otros cuatro están en la biblioteca de nodos, dentro de **Nodos asociados > video > MiniMax**, listos para agregarlos a tu propio flujo de trabajo. + +Para generar con MiniMax H3, H3 Max y H3 Max Turbo, consulta [Generación de video con la API de MiniMax H3](/es/tutorials/partner-nodes/minimax/minimax-h3). + + + + +## MiniMax H3 Max: Cast Swap + +Reemplaza a las personas de un clip finalizado sin necesidad de volver a rodar. El nodo Recast conserva el movimiento, el encuadre, los cortes y el audio original, y genera a las nuevas personas a partir de tus fotos de referencia. + + + + + + Abrir en Comfy Cloud + + + Descarga el JSON o busca "Cast Swap: MiniMax H3 Max" en la Biblioteca de plantillas + + + + + + Clip de origen del flujo de trabajo, o usa el tuyo. + + + Foto de referencia de la nueva persona, o usa la tuya. + + + +Carga tu propio clip y conecta de una a cuatro fotos, una por cada nueva persona. De forma predeterminada, las fotos reemplazan a las personas principales del clip de izquierda a derecha, así que conéctalas en el orden que quieras. El prompt es opcional: nombra las fotos como `@Image1` a `@Image4` y di quién se convierte en quién, además de todo lo que deba permanecer igual. + +- **Clip de origen**: de 5 a 30 segundos, sin ningún plano individual de más de 15 segundos +- **Fotos de referencia**: al menos 256 x 256 píxeles, hasta cuatro +- **Resolución**: 768P o 1080P +- **Enhance realism**: desactivado de forma predeterminada. Actívalo para obtener una textura de piel más natural y una mejor integración en la escena + +## MiniMax H3 Max: Reencuadre de relación de aspecto + +Ajusta un clip a una relación de aspecto distinta sin recortar al sujeto. La fuente conserva su posición, su sincronización y su audio sobre el nuevo lienzo, y el modelo genera la imagen de alrededor. + + + + + + Abrir en Comfy Cloud + + + Descarga el JSON o busca "Aspect Ratio Reframe: MiniMax H3 Max" en Template Library + + + + + + Clip de origen para el flujo de trabajo, o usa el tuyo. + + + +La demo incluida convierte un clip vertical de retrato en un encuadre de 16:9. Carga tu propio clip, elige la relación de aspecto y la resolución objetivo, y usa el prompt para describir qué rellena la nueva zona; por ejemplo, continuar el fondo que ya está detrás del sujeto a ambos lados. + +- **Clip de origen**: de 0.5 a 300 segundos +- **Relación de aspecto**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 o 9:21 +- **Resolución**: 480P, 768P o 1080P + +## Nodos de edición de MiniMax H3 Max + +| Nodo | Qué hace | Medio inicial | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | Reemplaza a las personas de un video con personas de fotos de referencia, manteniendo el movimiento, la cámara, los cortes y el sonido de la fuente | Video de 5-30 segundos, 1-4 fotos | +| **MiniMax H3 Max Extend Video** | Continúa un video con metraje nuevo descrito en el prompt | Video de 1.625-60 segundos | +| **MiniMax H3 Max Reframe Video** | Ajusta un video a una nueva relación de aspecto y genera el área que lo rodea, manteniendo la sincronización y el sonido | Video de 0.5-300 segundos | +| **MiniMax H3 Max Insert Video** | Reemplaza la sección entre dos marcas de tiempo con una escena recién generada | Video de hasta 60 segundos, más un tiempo de inicio y de reanudación | +| **MiniMax H3 Max Lip Sync** | Anima una imagen fija para que la boca siga una pista de audio | Una imagen, 5-899 segundos de audio | +| **MiniMax H3 Max Camera Control** | Orbita la cámara alrededor del sujeto en una imagen fija | Una imagen, más un ángulo y una distancia finales | + +En los prompts de Recast, Extend Video e Insert Video, haz referencia a las referencias conectadas como `@Image1`, `@Audio1` o `@Video1`, numeradas en el orden de conexión. El nodo Reference to Video acepta la misma forma abreviada para sus propias referencias. Una etiqueta que apunta más allá de las referencias que conectaste detiene la ejecución con un error. + +Las tarifas por nodo y resolución están en la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing). diff --git a/es/tutorials/partner-nodes/minimax/minimax-h3.mdx b/es/tutorials/partner-nodes/minimax/minimax-h3.mdx index cc47321d42..e7b363e2c3 100644 --- a/es/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/es/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -2,13 +2,13 @@ title: "Generación de video con la API de MiniMax H3 en ComfyUI" description: "Usa los nodos API de MiniMax H3 en ComfyUI para flujos en la nube de texto a video, primer/último fotograma y referencia a video con audio estéreo nativo." sidebarTitle: "Minimax H3" -translationSourceHash: f247dec4 +translationSourceHash: fc98a97a translationFrom: tutorials/partner-nodes/minimax/minimax-h3.mdx translationBlockHashes: "_intro": a261dbf5 "What MiniMax H3 is good at": 4c84d1e7 "Example outputs": 77256fd5 - "MiniMax H3 API workflows": 272af121 + "MiniMax H3 API workflows": 9f3a305f --- import ReqHint from "/snippets/es/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/es/tutorials/update-reminder.mdx"; @@ -51,20 +51,19 @@ Generación de primer y último fotograma, con el modelo creando el movimiento e ## Flujos de trabajo de la API de MiniMax H3 -La biblioteca de plantillas incluye tres flujos de trabajo de ejemplo de API para el modelo base H3, seis para H3 Max (cuatro de generación y dos de edición) y dos para H3 Max Turbo: +La biblioteca de plantillas incluye tres flujos de trabajo de ejemplo de API para el modelo base H3, cuatro para H3 Max y dos para H3 Max Turbo: - **Text to Video**: Genera videos a partir de prompts de texto - **First-Last-Frame to Video**: Genera un video entre una imagen de primer y último fotograma - **Reference to Video**: Genera videos condicionados por imágenes, videos y audio de referencia - **H3 Max T2V / I2V / FLF2V / R2V**: Los mismos modos a través del modelo MiniMax H3 Max, más rápido, con mejora de prompt opcional y etapas de ampliación a 2K integradas en cada flujo de trabajo - **H3 Max Turbo T2V / I2V**: Texto a video y video de primer/último fotograma a través del modelo MiniMax H3 Max Turbo, el más rápido, con las mismas etapas opcionales de mejora de prompt y ampliación a 2K -- **H3 Max cambio de reparto / reencuadre**: Flujos de trabajo de edición que sustituyen a las personas de un clip terminado o lo ajustan a otra relación de aspecto sin recortar al sujeto ### MiniMax H3 Max -Las seis plantillas `api_minimax_h3_max_*` se ejecutan en el modelo MiniMax H3 Max a través de los mismos nodos API: los nodos Text to Video y First-Last-Frame recibieron una opción H3 Max en ComfyUI 0.34.2, y el nodo Reference to Video en 0.34.3. Las plantillas de cambio de reparto y reencuadre usan los nodos de edición H3 Max añadidos en ComfyUI 0.44.0. Actualiza ComfyUI a la versión 0.44.0 como mínimo para ejecutarlas todas. +Las cuatro plantillas `api_minimax_h3_max_*` se ejecutan en el modelo MiniMax H3 Max a través de los mismos nodos API: los nodos Text to Video y First-Last-Frame recibieron una opción H3 Max en ComfyUI 0.34.2, y el nodo Reference to Video en 0.34.3. Actualiza ComfyUI a la versión 0.34.3 como mínimo para ejecutarlas todas. -H3 Max genera más rápido que en tiempo real y cobra tarifas más bajas que el modelo base: 15.09 créditos/s a 480P y 24.14 créditos/s a 768P (consulta la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing)). Cada plantilla Max de generación añade dos etapas opcionales detrás de un nodo Switch, ambas desactivadas de forma predeterminada: +H3 Max genera más rápido que en tiempo real y cobra tarifas más bajas que el modelo base: 15.09 créditos/s a 480P y 24.14 créditos/s a 768P (consulta la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing)). Cada plantilla Max añade dos etapas opcionales detrás de un nodo Switch, ambas desactivadas de forma predeterminada: 1. **Prompt enhance**: Reescribe el prompt con el nodo MiniMax H3 Context IR antes de la generación. Activa el booleano `Prompt Enhance` para habilitarla 2. **Upscale to 2K**: Ejecuta el nodo Regenerar MiniMax H3 a 2K sobre el video de la primera pasada. Activa el booleano `Upscale to 2K` para habilitarla @@ -75,6 +74,8 @@ Ten en cuenta lo siguiente al conectar una plantilla Max: - **Iguala las duraciones entre nodos**: la generación y la mejora de prompt deben usar la misma duración, o las etapas se desincronizan. El nodo Regenerar MiniMax H3 a 2K no tiene entrada de duración; vuelve a renderizar el video de 768P enviado con su duración original - **Duplica las entradas adicionales**: los tres nodos no comparten el contexto adicional automáticamente. Si agregas un video de referencia, audio u otra imagen, conecta los mismos cables tanto en el nodo Context IR como en el nodo Regenerar MiniMax H3 a 2K +Para cambiar a las personas de un clip terminado, reencuadrarlo, extenderlo o insertar una toma nueva, usa los nodos de edición H3 Max: consulta [MiniMax H3 Max Edición](/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing). + ### MiniMax H3 Texto a Video Genera videos a partir de prompts de texto mediante la API de MiniMax H3, con audio estéreo nativo. @@ -279,78 +280,3 @@ Genera videos condicionados por medios de referencia con el modelo H3 Max, más Imagen de referencia para el flujo de trabajo, o usa tus propias imágenes. - -#### MiniMax H3 Max: cambio de reparto - -Sustituye a las personas de un clip terminado sin volver a rodar. El nodo Recast conserva el movimiento, el encuadre, los cortes y el audio original, y genera las nuevas personas a partir de tus fotos de referencia. - - - - - - Abrir en Comfy Cloud - - - Descargar JSON o buscar "Cast Swap: MiniMax H3 Max" en la Biblioteca de plantillas - - - - - - Clip de origen para el flujo de trabajo, o usa el tuyo. - - - Foto de referencia de la nueva persona, o usa la tuya. - - - -Carga tu propio clip y conecta de una a cuatro fotos, una por cada persona nueva. De forma predeterminada las fotos sustituyen a las personas principales del clip de izquierda a derecha, así que conéctalas en el orden que quieras. El prompt es opcional: nombra las fotos como `@Image1` a `@Image4` e indica quién pasa a ser quién, además de lo que debe mantenerse igual. - -- **Clip de origen**: de 5 a 30 segundos, sin ningún plano de más de 15 segundos -- **Fotos de referencia**: al menos 256 x 256 píxeles, hasta cuatro -- **Resolución**: 768P o 1080P -- **Mejorar el realismo**: desactivado de forma predeterminada. Actívalo para obtener una textura de piel más natural y una mejor integración en la escena - -#### MiniMax H3 Max: reencuadre de relación de aspecto - -Ajusta un clip a otra relación de aspecto sin recortar al sujeto. El origen mantiene su posición, su ritmo y su audio en el nuevo lienzo, y el modelo genera la imagen de alrededor. - - - - - - Abrir en Comfy Cloud - - - Descargar JSON o buscar "Aspect Ratio Reframe: MiniMax H3 Max" en la Biblioteca de plantillas - - - - - - Clip de origen para el flujo de trabajo, o usa el tuyo. - - - -La demo incluida convierte un clip vertical en un encuadre 16:9. Carga tu propio clip, elige la relación de aspecto y la resolución de destino, y usa el prompt para describir qué ocupa la zona nueva, por ejemplo continuar el fondo que ya aparece a ambos lados del sujeto. - -- **Clip de origen**: de 0.5 a 300 segundos -- **Relación de aspecto**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 o 9:21 -- **Resolución**: 480P, 768P o 1080P - -#### Nodos de edición de MiniMax H3 Max - -Seis nodos de edición H3 Max parten de un video o de una imagen fija en lugar de un prompt y editan ese material. Dos de ellos vienen con las plantillas anteriores; los otros cuatro están en la biblioteca de nodos bajo **Partner Nodes > video > MiniMax**, listos para añadirlos a tu propio flujo de trabajo. - -| Nodo | Qué hace | Material de partida | -| :--- | :----------- | :------------- | -| **MiniMax H3 Max Recast** | Sustituye a las personas de un video por las de fotos de referencia, manteniendo el movimiento, la cámara, los cortes y el sonido del original | Video de 5 a 30 segundos, 1 a 4 fotos | -| **MiniMax H3 Max Extend Video** | Continúa un video con material nuevo descrito en el prompt | Video de 1.625 a 60 segundos | -| **MiniMax H3 Max Reframe Video** | Ajusta un video a una nueva relación de aspecto y genera el área de alrededor, manteniendo el ritmo y el sonido | Video de 0.5 a 300 segundos | -| **MiniMax H3 Max Insert Video** | Sustituye el tramo entre dos marcas de tiempo por una escena nueva | Video de hasta 60 segundos, más un tiempo de inicio y de reanudación | -| **MiniMax H3 Max Lip Sync** | Anima una imagen fija para que su boca siga una pista de audio | Una imagen y de 5 a 899 segundos de audio | -| **MiniMax H3 Max Camera Control** | Orbita la cámara alrededor del sujeto de una imagen fija | Una imagen, más un ángulo y una distancia finales | - -En los prompts de Recast, Extend Video e Insert Video, nombra las referencias conectadas como `@Image1`, `@Audio1` o `@Video1`, numeradas en el orden de conexión. El nodo Reference to Video acepta la misma forma abreviada para sus propias referencias. Una etiqueta que apunte más allá de las referencias conectadas detiene la ejecución con un error. - -Las tarifas por nodo y resolución están en la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing). diff --git a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx new file mode 100644 index 0000000000..f918449123 --- /dev/null +++ b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -0,0 +1,96 @@ +--- +title: "ComfyUI での MiniMax H3 Max 編集" +description: "ComfyUI の MiniMax H3 Max 編集 APIノードを使用すると、完成したクリップのキャスト交換、フレーミング変更、ショットの延長や挿入、リップシンク、カメラの周回が行えます。" +sidebarTitle: "Minimax H3 Max 編集" +translationSourceHash: e1fd584f +translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +translationBlockHashes: + "_intro": f127db6f + "MiniMax H3 Max: Cast Swap": cd212e2a + "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 + "MiniMax H3 Max editing nodes": 43f08f03 +--- +import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; + +MiniMax H3 Maxの編集系ノード6種は、プロンプトだけではなくビデオや静止画を出発点として、そのマテリアルを編集します。他のH3 APIノードと同様に、お使いのComfy APIアカウントを通じてMiniMaxのサーバー上で実行され、これらの編集系ノードはComfyUI 0.44.0で追加されました。 + +このうち2つにはワークフローテンプレートが付属しています。**Cast Swap**(`api_minimax_h3_max_recast`)と**Aspect Ratio Reframe**(`api_minimax_h3_max_reframe`)です。残りの4つは**Partner Nodes > video > MiniMax**のノードライブラリにあり、自身のワークフローにそのまま追加できます。 + +MiniMax H3、H3 Max、H3 Max Turboでの生成については、[MiniMax H3 API ビデオ生成](/ja/tutorials/partner-nodes/minimax/minimax-h3)を参照してください。 + + + + +## MiniMax H3 Max: キャストスワップ + +撮り直しなしで、完成したクリップに登場する人物を入れ替えます。Recastノードはモーション、フレーミング、カット、オリジナルのオーディオを維持したまま、参照写真から新しい人物を生成します。 + + + + + + Comfy Cloud で開く + + + JSONをダウンロードするか、テンプレートライブラリで「Cast Swap: MiniMax H3 Max」を検索してください + + + + + + このワークフロー用のソースクリップです。ご自身のクリップを使用してもかまいません。 + + + 新しい人物の参照写真です。ご自身の写真を使用してもかまいません。 + + + +ご自身のクリップを読み込み、新しい人物1人につき1枚、最大4枚までの写真を接続します。デフォルトでは、写真はクリップ内の主要な人物を左から右へ置き換えるため、希望する順番に接続してください。プロンプトは任意です。写真を `@Image1` から `@Image4` として指定し、誰が誰になるのか、また変わらずに保つ必要があるものを記述してください。 + +- **ソースクリップ**: 5~30秒、1カットが15秒を超えないこと +- **参照写真**: 最小 256 x 256 ピクセル、最大4枚 +- **解像度**: 768P または 1080P +- **リアリズムを強化**: デフォルトはオフ。より自然な肌の質感とシーンへのより良いなじみを得るにはオンにします + +## MiniMax H3 Max: アスペクト比リフレーム + +被写体をクロップせずにクリップを別のアスペクト比にフィットさせます。ソースは新しいキャンバス上で位置、タイミング、オーディオを維持し、モデルが周囲の映像を生成します。 + + + + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで「Aspect Ratio Reframe: MiniMax H3 Max」を検索してください + + + + + + ワークフロー用のソースクリップです。ご自身のクリップを使用することもできます。 + + + +同梱のデモでは、垂直のポートレートクリップを 16:9 のフレームに変換します。ご自身のクリップを読み込み、目標のアスペクト比と解像度を選択し、プロンプトで新しい領域を何で埋めるかを記述してください。たとえば、被写体の両側にすでに存在する背景を続けるよう指定します。 + +- **ソースクリップ**: 0.5〜300 秒 +- **アスペクト比**: 16:9、9:16、1:1、4:3、3:4、21:9、9:21 +- **解像度**: 480P、768P、1080P + +## MiniMax H3 Max 編集ノード + +| ノード | 機能 | 入力メディア | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | ビデオ内の人物を参照写真の人物に置き換えます。動き、カメラワーク、カット、ソースのサウンドは維持されます | 5-30秒のビデオ、1-4枚の写真 | +| **MiniMax H3 Max Extend Video** | プロンプトで記述した新しい映像でビデオを続けます | 1.625-60秒のビデオ | +| **MiniMax H3 Max Reframe Video** | ビデオを新しいアスペクト比に収め、その周囲の領域を生成します。タイミングとサウンドは維持されます | 0.5-300秒のビデオ | +| **MiniMax H3 Max Insert Video** | 2つのタイムスタンプの間の区間を、新しく生成したシーンに置き換えます | 最大60秒のビデオ、および開始時刻と再開時刻 | +| **MiniMax H3 Max Lip Sync** | 静止画像をアニメーション化し、口元がオーディオトラックに合わせて動くようにします | 画像1枚、5-899秒のオーディオ | +| **MiniMax H3 Max Camera Control** | 静止画像内の被写体の周りをカメラが周回します | 画像1枚、および終了角度と距離 | + +Recast、Extend Video、Insert Videoのプロンプトでは、接続した参照を `@Image1`、`@Audio1`、`@Video1` として参照します。番号は接続順に付けられます。Reference to Videoノードも、自身の参照に同じ省略記法を使用できます。接続した参照を超えるタグを指定すると、エラーで実行が停止します。 + +ノードおよび解像度ごとの料金は [Minimax料金表](/ja/tutorials/partner-nodes/pricing) に記載されています。 diff --git a/ja/tutorials/partner-nodes/minimax/minimax-h3.mdx b/ja/tutorials/partner-nodes/minimax/minimax-h3.mdx index 4fd55eb881..53c43e0f57 100644 --- a/ja/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/ja/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -2,13 +2,13 @@ title: "ComfyUIでのMiniMax H3 APIビデオ生成" description: "ComfyUI で MiniMax H3 API ノードを使い、クラウドワークフローでテキストから動画、先頭・末尾フレームから動画、参照から動画を、ネイティブのステレオオーディオ付きで生成します。" sidebarTitle: "Minimax H3" -translationSourceHash: f247dec4 +translationSourceHash: fc98a97a translationFrom: tutorials/partner-nodes/minimax/minimax-h3.mdx translationBlockHashes: "_intro": a261dbf5 "What MiniMax H3 is good at": 4c84d1e7 "Example outputs": 77256fd5 - "MiniMax H3 API workflows": 272af121 + "MiniMax H3 API workflows": 9f3a305f --- import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; @@ -52,20 +52,19 @@ MiniMax H3 は、**ネイティブステレオオーディオ**付きのビデ ## MiniMax H3 API ワークフロー -テンプレートライブラリには、H3 ベースモデル向けに 3 つ、H3 Max 向けに 6 つ(生成 4 つ、編集 2 つ)、H3 Max Turbo 向けに 2 つの API サンプルワークフローが含まれています: +テンプレートライブラリには、H3 ベースモデル向けに 3 つ、H3 Max 向けに 4 つ、H3 Max Turbo 向けに 2 つの API サンプルワークフローが含まれています: - **テキストから動画生成**:テキストプロンプトから動画を生成します。 - **先頭・末尾フレームから動画生成**:先頭フレームと末尾フレームの画像から、その間の動画を生成します。 - **リファレンスから動画生成**:参照画像、動画、オーディオを条件として動画を生成します。 - **H3 Max T2V / I2V / FLF2V / R2V**:より高速な MiniMax H3 Max モデルで同じモードを実行します。各ワークフローには、任意のプロンプト拡張と 2K アップスケール段階が組み込まれています。 - **H3 Max Turbo T2V / I2V**:最速の MiniMax H3 Max Turbo モデルでテキストから動画生成と先頭・末尾フレーム動画を実行します。こちらにも任意のプロンプト拡張と 2K アップスケール段階が組み込まれています。 -- **H3 Max キャスト入れ替え / アスペクト比リフレーム**:完成したクリップの人物を入れ替えたり、被写体を切り取らずに新しいアスペクト比に合わせたりする編集ワークフローです。 ### MiniMax H3 Max -6 つの `api_minimax_h3_max_*` テンプレートは、同じ API ノードを通じて MiniMax H3 Max モデルで実行されます。Text to Video ノードと First-Last-Frame ノードには ComfyUI 0.34.2 で H3 Max オプションが追加され、Reference to Video ノードには 0.34.3 で追加されました。キャスト入れ替えとアスペクト比リフレームのテンプレートは、ComfyUI 0.44.0 で追加された H3 Max 編集ノードを使用します。すべて実行するには、ComfyUI を少なくとも 0.44.0 にアップデートしてください。 +4 つの `api_minimax_h3_max_*` テンプレートは、同じ API ノードを通じて MiniMax H3 Max モデルで実行されます。Text to Video ノードと First-Last-Frame ノードには ComfyUI 0.34.2 で H3 Max オプションが追加され、Reference to Video ノードには 0.34.3 で追加されました。すべて実行するには、ComfyUI を少なくとも 0.34.3 にアップデートしてください。 -H3 Max はリアルタイムよりも速く生成でき、ベースモデルより低いレートで課金されます:480P で 15.09 クレジット/秒、768P で 24.14 クレジット/秒([Minimax 料金表](/ja/tutorials/partner-nodes/pricing) を参照)。各 Max 生成テンプレートには、Switch ノードの背後に 2 つの任意段階があり、どちらもデフォルトでオフです: +H3 Max はリアルタイムよりも速く生成でき、ベースモデルより低いレートで課金されます:480P で 15.09 クレジット/秒、768P で 24.14 クレジット/秒([Minimax 料金表](/ja/tutorials/partner-nodes/pricing) を参照)。各 Max テンプレートには、Switch ノードの背後に 2 つの任意段階があり、どちらもデフォルトでオフです: 1. **プロンプト拡張**:生成前に MiniMax H3 Context IR ノードでプロンプトを書き換えます。`Prompt Enhance` ブール値をオンにすると有効になります 2. **2K へのアップスケール**:最初のパスのビデオに対して MiniMax H3 Regenerate to 2K ノードを実行します。`Upscale to 2K` ブール値をオンにすると有効になります @@ -76,6 +75,8 @@ Max テンプレートを配線する際は、以下の点に注意してくだ - **ノード間で再生時間を一致させる**:生成とプロンプト拡張は同じ再生時間を使用する必要があります。一致しないと各段階がずれます。Regenerate to 2K ノードには再生時間の入力がなく、送信された 768P ビデオを元の再生時間で再レンダリングします - **追加入力は複製する**:3 つのノードは追加コンテキストを自動的に共有しません。参照ビデオ、オーディオ、別の画像を追加した場合は、同じ配線を Context IR ノードと Regenerate ノードの両方に接続してください +完成したクリップの人物を差し替えたり、アスペクト比を変更したり、延長や新しいショットの挿入を行う場合は、H3 Max 編集ノードを使います。詳しくは [MiniMax H3 Max 編集](/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing) を参照してください。 + ### MiniMax H3 テキストから動画生成 MiniMax H3 API 経由で、テキストプロンプトから動画を生成します。ステレオオーディオにもネイティブ対応しています。 @@ -280,78 +281,3 @@ MiniMaxは、参照駆動型生成(R2V)向けの公式の[フル参照モー ワークフロー用の参照画像です。または、ご自身の画像を使用できます。 - -#### MiniMax H3 Max: キャスト入れ替え - -撮り直しなしで、完成したクリップの人物を入れ替えます。Recast ノードは動き、構図、カット、元のオーディオをそのまま保ち、参照写真から新しい人物を生成します。 - - - - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで「Cast Swap: MiniMax H3 Max」を検索してください - - - - - - ワークフロー用のソースクリップです。または、ご自身の動画を使用できます。 - - - 新しい人物の参照写真です。または、ご自身の写真を使用できます。 - - - -クリップを読み込み、新しい人物ごとに 1 枚、最大 4 枚の写真を接続します。既定では写真がクリップ内の主要な人物を左から右へ置き換えるため、希望する順序で接続してください。プロンプトは任意です。写真を `@Image1` から `@Image4` で指定し、誰が誰になるか、変わってはいけない点を書きます。 - -- **ソースクリップ**: 5〜30 秒、1 ショットは 15 秒以下 -- **参照写真**: 256 x 256 ピクセル以上、最大 4 枚 -- **解像度**: 768P または 1080P -- **リアリズムを強化**: 既定ではオフ。オンにすると肌の質感が自然になり、新しい人物がシーンになじみます - -#### MiniMax H3 Max: アスペクト比のリフレーム - -被写体を切り取らずに、クリップを別のアスペクト比に合わせます。ソースは新しいキャンバス上で位置、タイミング、オーディオを保ち、周囲の映像はモデルが生成します。 - - - - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで「Aspect Ratio Reframe: MiniMax H3 Max」を検索してください - - - - - - ワークフロー用のソースクリップです。または、ご自身の動画を使用できます。 - - - -同梱のデモは縦向きのクリップを 16:9 のフレームに変換します。ご自身のクリップを読み込み、目標のアスペクト比と解像度を選び、新しく加わる領域の内容をプロンプトで説明してください。たとえば、被写体の両側にある背景をそのまま続けるよう指定します。 - -- **ソースクリップ**: 0.5〜300 秒 -- **アスペクト比**: 16:9、9:16、1:1、4:3、3:4、21:9、9:21 -- **解像度**: 480P、768P、1080P - -#### MiniMax H3 Max 編集ノード - -6 つの H3 Max 編集ノードは、プロンプトだけでなく動画や静止画を起点として、その素材を編集します。うち 2 つは上記のテンプレートに含まれており、残りの 4 つはノードライブラリの **Partner Nodes > video > MiniMax** にあり、ご自身のワークフローに追加できます。 - -| ノード | 動作 | 入力素材 | -| :--- | :----------- | :------------- | -| **MiniMax H3 Max Recast** | 参照写真の人物で動画内の人物を置き換え、元動画の動き、カメラ、カット、音を保ちます | 5〜30 秒の動画、写真 1〜4 枚 | -| **MiniMax H3 Max Extend Video** | プロンプトで説明した新しい映像で動画を延長します | 1.625〜60 秒の動画 | -| **MiniMax H3 Max Reframe Video** | 動画を新しいアスペクト比に収め、周囲の領域を生成します。タイミングと音は保たれます | 0.5〜300 秒の動画 | -| **MiniMax H3 Max Insert Video** | 2 つのタイムスタンプの間を新しく生成したシーンに置き換えます | 最大 60 秒の動画と、開始・再開の時刻 | -| **MiniMax H3 Max Lip Sync** | 静止画の口の動きをオーディオに合わせます | 画像 1 枚、5〜899 秒のオーディオ | -| **MiniMax H3 Max Camera Control** | 静止画の被写体の周りでカメラを回り込ませます | 画像 1 枚と、終了角度・距離 | - -Recast、Extend Video、Insert Video のプロンプトでは、接続した参照素材を `@Image1`、`@Audio1`、`@Video1` のように接続順の番号で指定します。Reference to Video ノードも同じ省略記法を受け付けます。接続数を超えるタグを指定すると、実行はエラーで停止します。 - -ノードと解像度ごとの料金は [Minimax 料金表](/ja/tutorials/partner-nodes/pricing) にあります。 diff --git a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx new file mode 100644 index 0000000000..3e1997e596 --- /dev/null +++ b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -0,0 +1,96 @@ +--- +title: "ComfyUI의 MiniMax H3 Max 편집" +description: "ComfyUI의 MiniMax H3 Max 편집 API 노드로 완성된 클립에서 출연진을 교체하고, 구도를 재설정하고, 장면을 연장하거나 삽입하고, 립 싱크를 맞추고, 카메라를 회전시킬 수 있습니다." +sidebarTitle: "Minimax H3 Max 편집" +translationSourceHash: e1fd584f +translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +translationBlockHashes: + "_intro": f127db6f + "MiniMax H3 Max: Cast Swap": cd212e2a + "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 + "MiniMax H3 Max editing nodes": 43f08f03 +--- +import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; + +여섯 개의 MiniMax H3 Max 편집 노드는 프롬프트만 사용하는 대신 비디오나 정지 이미지에서 시작하여 해당 소재를 편집합니다. 이 노드들은 다른 H3 API 노드와 마찬가지로 Comfy API 계정을 통해 MiniMax의 서버에서 실행되며, 편집 노드는 ComfyUI 0.44.0에서 추가되었습니다. + +이 중 두 개는 워크플로 템플릿과 함께 제공됩니다: **Cast Swap** (`api_minimax_h3_max_recast`)과 **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`)입니다. 나머지 네 개는 **Partner Nodes > video > MiniMax** 아래 노드 라이브러리에 있으며, 자신의 워크플로에 바로 추가할 수 있습니다. + +MiniMax H3, H3 Max, H3 Max Turbo로 생성하려면 [MiniMax H3 API 비디오 생성](/ko/tutorials/partner-nodes/minimax/minimax-h3)을 참조하세요. + + + + +## MiniMax H3 Max: Cast Swap + +재촬영 없이 완성된 클립 속 인물을 교체합니다. Recast 노드는 모션, 프레이밍, 컷, 원본 오디오를 그대로 유지하고, 참조 사진으로 새로운 인물을 생성합니다. + + + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Cast Swap: MiniMax H3 Max" 검색 + + + + + + 워크플로용 소스 클립이며, 직접 준비한 클립을 사용해도 됩니다. + + + 새 인물의 참조 사진이며, 직접 준비한 사진을 사용해도 됩니다. + + + +직접 준비한 클립을 불러오고, 새 인물마다 한 장씩 1~4장의 사진을 연결합니다. 기본적으로 사진은 클립의 주요 인물을 왼쪽에서 오른쪽 순서로 교체하므로, 원하는 순서대로 연결하세요. 프롬프트는 선택 사항입니다. 사진의 이름을 `@Image1`부터 `@Image4`까지로 지정하고, 누가 누구로 바뀌는지와 함께 그대로 유지되어야 하는 요소를 설명하세요. + +- **소스 클립**: 5~30초, 단일 샷은 15초를 넘지 않아야 합니다 +- **참조 사진**: 최소 256 x 256 픽셀, 최대 4장 +- **해상도**: 768P 또는 1080P +- **사실감 향상**: 기본적으로 꺼짐입니다. 더 자연스러운 피부 질감과 장면에 더 잘 어우러지는 결과를 원하면 켜세요 + +## MiniMax H3 Max: 화면 비율 리프레임 + +피사체를 잘라내지 않고 클립을 다른 화면 비율에 맞춥니다. 소스는 새 캔버스에서 위치, 타이밍, 오디오를 그대로 유지하며, 모델이 주변 화면을 생성합니다. + + + + + + Comfy Cloud에서 열기 + + + JSON을 다운로드하거나 템플릿 라이브러리에서 "Aspect Ratio Reframe: MiniMax H3 Max"를 검색하세요 + + + + + + 워크플로에 사용할 소스 클립이며, 직접 준비한 클립을 사용해도 됩니다. + + + +포함된 데모는 수직 인물 클립을 16:9 프레임으로 변환합니다. 직접 준비한 클립을 불러오고, 대상 화면 비율과 해상도를 선택한 뒤, 프롬프트로 새 영역을 무엇으로 채울지 설명하세요. 예를 들어 피사체 양옆 뒤에 이미 있는 배경을 이어서 그리도록 지정할 수 있습니다. + +- **소스 클립**: 0.5 ~ 300초 +- **화면 비율**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 또는 9:21 +- **해상도**: 480P, 768P 또는 1080P + +## MiniMax H3 Max 편집 노드 + +| 노드 | 기능 | 시작 미디어 | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | 비디오 속 인물을 레퍼런스 사진의 인물로 교체하며, 소스의 움직임, 카메라, 컷, 사운드를 유지합니다 | 5-30초 비디오, 사진 1-4장 | +| **MiniMax H3 Max Extend Video** | 프롬프트에 설명된 새로운 영상으로 비디오를 이어서 생성합니다 | 1.625-60초 비디오 | +| **MiniMax H3 Max Reframe Video** | 비디오를 새로운 화면 비율에 맞추고 주변 영역을 생성하며, 타이밍과 사운드를 유지합니다 | 0.5-300초 비디오 | +| **MiniMax H3 Max Insert Video** | 두 타임스탬프 사이의 구간을 새로 생성된 장면으로 교체합니다 | 최대 60초 비디오와 시작 및 재개 시간 | +| **MiniMax H3 Max Lip Sync** | 정지 이미지를 애니메이션으로 만들어 입 모양이 오디오 트랙을 따라가게 합니다 | 이미지 1장, 5-899초 오디오 | +| **MiniMax H3 Max Camera Control** | 정지 이미지 속 피사체 주위로 카메라를 회전시킵니다 | 이미지 1장과 종료 각도 및 거리 | + +Recast, Extend Video, Insert Video 프롬프트에서는 연결된 레퍼런스를 연결 순서대로 번호가 매겨진 `@Image1`, `@Audio1`, `@Video1`로 참조하세요. 레퍼런스 기반 비디오 생성 노드도 자체 레퍼런스에 동일한 축약 표기를 사용할 수 있습니다. 연결한 레퍼런스 범위를 벗어난 태그를 지정하면 실행이 오류와 함께 중단됩니다. + +노드 및 해상도별 요금은 [Minimax 가격표](/ko/tutorials/partner-nodes/pricing)에 나와 있습니다. diff --git a/ko/tutorials/partner-nodes/minimax/minimax-h3.mdx b/ko/tutorials/partner-nodes/minimax/minimax-h3.mdx index 25db2ae72d..c6ff060029 100644 --- a/ko/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/ko/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -2,13 +2,13 @@ title: "ComfyUI에서 MiniMax H3 API 비디오 생성" description: "ComfyUI에서 MiniMax H3 API 노드로 클라우드 워크플로를 실행해 텍스트, 첫/마지막 프레임, 레퍼런스 기반 비디오를 네이티브 스테레오 오디오와 함께 생성하세요." sidebarTitle: "MiniMax H3" -translationSourceHash: f247dec4 +translationSourceHash: fc98a97a translationFrom: tutorials/partner-nodes/minimax/minimax-h3.mdx translationBlockHashes: "_intro": a261dbf5 "What MiniMax H3 is good at": 4c84d1e7 "Example outputs": 77256fd5 - "MiniMax H3 API workflows": 272af121 + "MiniMax H3 API workflows": 9f3a305f --- import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -51,20 +51,19 @@ MiniMax H3는 **네이티브 스테레오 오디오**로 비디오를 생성합 ## MiniMax H3 API 워크플로 -템플릿 라이브러리에는 H3 베이스 모델용 API 예제 워크플로 3개, H3 Max용 6개(생성 4개, 편집 2개), H3 Max Turbo용 2개가 포함되어 있습니다: +템플릿 라이브러리에는 H3 베이스 모델용 API 예제 워크플로 3개, H3 Max용 4개, H3 Max Turbo용 2개가 포함되어 있습니다: - **텍스트 기반 비디오 생성**: 텍스트 프롬프트에서 비디오를 생성합니다. - **첫/마지막 프레임 기반 비디오 생성**: 첫 번째 프레임 이미지와 마지막 프레임 이미지 사이의 비디오를 생성합니다. - **레퍼런스 기반 비디오 생성**: 레퍼런스 이미지, 비디오 및 오디오를 조건으로 비디오를 생성합니다. - **H3 Max T2V / I2V / FLF2V / R2V**: 더 빠른 MiniMax H3 Max 모델로 동일한 모드를 실행합니다. 각 워크플로에는 선택적 프롬프트 개선 및 2K 업스케일 단계가 내장되어 있습니다. - **H3 Max Turbo T2V / I2V**: 가장 빠른 MiniMax H3 Max Turbo 모델로 텍스트 기반 비디오 생성과 첫/마지막 프레임 비디오를 실행합니다. 이 역시 선택적 프롬프트 개선 및 2K 업스케일 단계가 내장되어 있습니다. -- **H3 Max 캐스트 교체 / 화면 비율 리프레임**: 완성된 클립의 인물을 교체하거나 피사체를 자르지 않고 새로운 화면 비율에 맞추는 편집 워크플로입니다. ### MiniMax H3 Max -6개의 `api_minimax_h3_max_*` 템플릿은 동일한 API 노드를 통해 MiniMax H3 Max 모델에서 실행됩니다. Text to Video 및 First-Last-Frame 노드에는 ComfyUI 0.34.2에서, Reference to Video 노드에는 0.34.3에서 H3 Max 옵션이 추가되었습니다. 캐스트 교체와 화면 비율 리프레임 템플릿은 ComfyUI 0.44.0에서 추가된 H3 Max 편집 노드를 사용합니다. 모두 실행하려면 ComfyUI를 최소 0.44.0으로 업데이트하세요. +4개의 `api_minimax_h3_max_*` 템플릿은 동일한 API 노드를 통해 MiniMax H3 Max 모델에서 실행됩니다. Text to Video 및 First-Last-Frame 노드에는 ComfyUI 0.34.2에서, Reference to Video 노드에는 0.34.3에서 H3 Max 옵션이 추가되었습니다. 모두 실행하려면 ComfyUI를 최소 0.34.3으로 업데이트하세요. -H3 Max는 실시간보다 빠르게 생성되며 베이스 모델보다 낮은 요율로 청구됩니다: 480P는 15.09 크레딧/초, 768P는 24.14 크레딧/초입니다([Minimax 요금표](/ko/tutorials/partner-nodes/pricing) 참조). 각 Max 생성 템플릿에는 Switch 노드 뒤에 선택적 단계 2개가 있으며, 둘 다 기본적으로 꺼져 있습니다: +H3 Max는 실시간보다 빠르게 생성되며 베이스 모델보다 낮은 요율로 청구됩니다: 480P는 15.09 크레딧/초, 768P는 24.14 크레딧/초입니다([Minimax 요금표](/ko/tutorials/partner-nodes/pricing) 참조). 각 Max 템플릿에는 Switch 노드 뒤에 선택적 단계 2개가 있으며, 둘 다 기본적으로 꺼져 있습니다: 1. **프롬프트 개선**: 생성 전에 MiniMax H3 Context IR 노드로 프롬프트를 다시 씁니다. `Prompt Enhance` 불리언을 켜면 활성화됩니다 2. **2K로 업스케일**: 첫 번째 패스 비디오에 MiniMax H3 Regenerate to 2K 노드를 실행합니다. `Upscale to 2K` 불리언을 켜면 활성화됩니다 @@ -75,6 +74,8 @@ Max 템플릿을 연결할 때 다음 사항에 유의하세요: - **노드 간 재생 시간 일치**: 생성과 프롬프트 개선은 동일한 재생 시간을 사용해야 합니다. 그렇지 않으면 단계가 어긋납니다. Regenerate to 2K 노드에는 재생 시간 입력이 없으며, 제출된 768P 비디오를 원래 재생 시간으로 다시 렌더링합니다 - **추가 입력은 복제**: 3개의 노드는 추가 컨텍스트를 자동으로 공유하지 않습니다. 참조 비디오, 오디오 또는 다른 이미지를 추가했다면 동일한 연결을 Context IR 노드와 Regenerate 노드 모두에 연결하세요 +완성된 클립의 인물을 교체하거나 화면 비율을 바꾸거나 클립을 연장하고 새 장면을 삽입하려면 H3 Max 편집 노드를 사용하세요. 자세한 내용은 [MiniMax H3 Max 편집](/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing)을 참고하세요. + ### MiniMax H3 텍스트 기반 비디오 생성 MiniMax H3 API를 통해 텍스트 프롬프트에서 비디오를 생성할 수 있으며, 스테레오 오디오를 기본 지원합니다. @@ -279,78 +280,3 @@ MiniMax는 레퍼런스 기반 생성(R2V)을 위한 공식 [전체 레퍼런스 워크플로용 레퍼런스 이미지입니다. 또는 직접 만든 이미지를 사용하세요. - -#### MiniMax H3 Max: 캐스트 교체 - -다시 촬영하지 않고 완성된 클립의 인물을 교체합니다. Recast 노드는 움직임, 구도, 컷, 원본 오디오를 그대로 유지하고 참조 사진으로 새 인물을 생성합니다. - - - - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Cast Swap: MiniMax H3 Max" 검색 - - - - - - 워크플로용 소스 클립입니다. 또는 직접 만든 영상을 사용하세요. - - - 새 인물의 참조 사진입니다. 또는 직접 만든 사진을 사용하세요. - - - -클립을 불러오고 새 인물마다 한 장씩, 최대 네 장의 사진을 연결합니다. 기본 설정에서는 사진이 클립의 주요 인물을 왼쪽에서 오른쪽으로 교체하므로 원하는 순서대로 연결하세요. 프롬프트는 선택 사항입니다. 사진을 `@Image1`부터 `@Image4`로 지칭하고 누가 누구로 바뀌는지, 무엇이 그대로여야 하는지 적습니다. - -- **소스 클립**: 5~30초, 한 샷은 15초 이하 -- **참조 사진**: 최소 256 x 256 픽셀, 최대 네 장 -- **해상도**: 768P 또는 1080P -- **사실감 강화**: 기본적으로 꺼져 있습니다. 켜면 피부 질감이 더 자연스럽고 새 인물이 장면에 더 잘 어울립니다 - -#### MiniMax H3 Max: 화면 비율 리프레임 - -피사체를 자르지 않고 클립을 다른 화면 비율에 맞춥니다. 소스는 새 캔버스에서 위치, 타이밍, 오디오를 유지하고 주변 화면은 모델이 생성합니다. - - - - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Aspect Ratio Reframe: MiniMax H3 Max" 검색 - - - - - - 워크플로용 소스 클립입니다. 또는 직접 만든 영상을 사용하세요. - - - -포함된 데모는 세로 클립을 16:9 프레임으로 바꿉니다. 직접 만든 클립을 불러오고 목표 화면 비율과 해상도를 선택한 뒤, 새로 채워질 영역을 프롬프트로 설명하세요. 예를 들어 피사체 양옆에 이미 있는 배경을 이어서 그리도록 지정합니다. - -- **소스 클립**: 0.5~300초 -- **화면 비율**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, 9:21 -- **해상도**: 480P, 768P, 1080P - -#### MiniMax H3 Max 편집 노드 - -여섯 H3 Max 편집 노드는 프롬프트만이 아니라 비디오나 정지 이미지에서 시작해 그 소재를 편집합니다. 이 중 두 개는 위의 템플릿에 포함되어 있고, 나머지 네 개는 노드 라이브러리의 **Partner Nodes > video > MiniMax** 아래에 있어 직접 워크플로에 추가할 수 있습니다. - -| 노드 | 기능 | 시작 소재 | -| :--- | :----------- | :------------- | -| **MiniMax H3 Max Recast** | 참조 사진의 인물로 비디오 속 인물을 교체하며, 원본의 움직임, 카메라, 컷, 소리를 유지합니다 | 5~30초 비디오, 사진 1~4장 | -| **MiniMax H3 Max Extend Video** | 프롬프트로 설명한 새 장면으로 비디오를 연장합니다 | 1.625~60초 비디오 | -| **MiniMax H3 Max Reframe Video** | 비디오를 새 화면 비율에 맞추고 주변 영역을 생성하며, 타이밍과 소리는 유지합니다 | 0.5~300초 비디오 | -| **MiniMax H3 Max Insert Video** | 두 타임스탬프 사이 구간을 새로 생성한 장면으로 교체합니다 | 최대 60초 비디오와 시작/재개 시각 | -| **MiniMax H3 Max Lip Sync** | 정지 이미지의 입 모양을 오디오에 맞춰 움직입니다 | 이미지 1장, 5~899초 오디오 | -| **MiniMax H3 Max Camera Control** | 정지 이미지의 피사체 주위로 카메라를 회전시킵니다 | 이미지 1장과 종료 각도·거리 | - -Recast, Extend Video, Insert Video의 프롬프트에서는 연결한 참조 소재를 `@Image1`, `@Audio1`, `@Video1`처럼 연결 순서대로 지정합니다. Reference to Video 노드도 같은 축약 표기를 지원합니다. 연결한 개수를 넘는 태그를 쓰면 실행이 오류로 중단됩니다. - -노드와 해상도별 요율은 [Minimax 요금표](/ko/tutorials/partner-nodes/pricing)에 있습니다. diff --git a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx new file mode 100644 index 0000000000..d143575209 --- /dev/null +++ b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -0,0 +1,89 @@ +--- +title: "MiniMax H3 Max Editing in ComfyUI" +description: "Swap the cast, reframe, extend or insert shots, lip sync, and orbit the camera in finished clips with the MiniMax H3 Max editing API nodes in ComfyUI." +sidebarTitle: "Minimax H3 Max Editing" +--- +import ReqHint from "/snippets/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; + +Six MiniMax H3 Max editing nodes start from a video or a still image instead of a prompt alone and edit that material. They run on MiniMax's servers through your Comfy API account like the other H3 API nodes, and the editing nodes were added in ComfyUI 0.44.0. + +Two of them ship with workflow templates: **Cast Swap** (`api_minimax_h3_max_recast`) and **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`). The other four are in the node library under **Partner Nodes > video > MiniMax**, ready to drop into your own workflow. + +For generation with MiniMax H3, H3 Max, and H3 Max Turbo, see [MiniMax H3 API Video Generation](/tutorials/partner-nodes/minimax/minimax-h3). + + + + +## MiniMax H3 Max: Cast Swap + +Replace the people in a finished clip without a reshoot. The Recast node keeps the motion, the framing, the cuts and the original audio, and generates the new people from your reference photos. + + + + + + Open in Comfy Cloud + + + Download JSON or search "Cast Swap: MiniMax H3 Max" in Template Library + + + + + + Source clip for the workflow, or use your own. + + + Reference photo of the new person, or use your own. + + + +Load your own clip and connect one to four photos, one per new person. By default the photos replace the main people in the clip from left to right, so connect them in the order you want. The prompt is optional: name the photos as `@Image1` to `@Image4` and say who becomes whom, plus anything that has to stay the same. + +- **Source clip**: 5 to 30 seconds, with no single shot longer than 15 seconds +- **Reference photos**: at least 256 x 256 pixels, up to four +- **Resolution**: 768P or 1080P +- **Enhance realism**: off by default. Turn it on for more natural skin texture and a better blend into the scene + +## MiniMax H3 Max: Aspect Ratio Reframe + +Fit a clip to a different aspect ratio without cropping the subject. The source keeps its position, timing and audio on the new canvas, and the model generates the surrounding picture. + + + + + + Open in Comfy Cloud + + + Download JSON or search "Aspect Ratio Reframe: MiniMax H3 Max" in Template Library + + + + + + Source clip for the workflow, or use your own. + + + +The bundled demo turns a vertical portrait clip into a 16:9 frame. Load your own clip, pick the target aspect ratio and resolution, and use the prompt to describe what fills the new area, for example continuing the background that is already behind the subject on both sides. + +- **Source clip**: 0.5 to 300 seconds +- **Aspect ratio**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 or 9:21 +- **Resolution**: 480P, 768P or 1080P + +## MiniMax H3 Max editing nodes + +| Node | What it does | Starting media | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | Replaces the people in a video with people from reference photos, keeping the motion, camera, cuts and sound of the source | 5-30 second video, 1-4 photos | +| **MiniMax H3 Max Extend Video** | Continues a video with new footage described in the prompt | 1.625-60 second video | +| **MiniMax H3 Max Reframe Video** | Fits a video into a new aspect ratio and generates the area around it, keeping timing and sound | 0.5-300 second video | +| **MiniMax H3 Max Insert Video** | Replaces the section between two timestamps with a newly generated scene | Video up to 60 seconds, plus a start and resume time | +| **MiniMax H3 Max Lip Sync** | Animates a still image so the mouth follows an audio track | One image, 5-899 seconds of audio | +| **MiniMax H3 Max Camera Control** | Orbits the camera around the subject in a still image | One image, plus an end angle and distance | + +In the Recast, Extend Video and Insert Video prompts, refer to the connected references as `@Image1`, `@Audio1` or `@Video1`, numbered in connection order. The Reference to Video node accepts the same shorthand for its own references. A tag that points past the references you connected stops the run with an error. + +Rates per node and resolution are in the [Minimax pricing table](/tutorials/partner-nodes/pricing). diff --git a/tutorials/partner-nodes/minimax/minimax-h3.mdx b/tutorials/partner-nodes/minimax/minimax-h3.mdx index f8424cd9f2..1a321ca45e 100644 --- a/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -44,20 +44,19 @@ First-last-frame generation, with the model creating the motion between two fram ## MiniMax H3 API workflows -The template library ships with three API example workflows for the base H3 model, six for H3 Max (four generation, two editing), and two for H3 Max Turbo: +The template library ships with three API example workflows for the base H3 model, four for H3 Max, and two for H3 Max Turbo: - **Text to Video**: Generate videos from text prompts - **First-Last-Frame to Video**: Generate a video between a first and last frame image - **Reference to Video**: Generate videos conditioned on reference images, videos, and audio - **H3 Max T2V / I2V / FLF2V / R2V**: The same modes through the faster MiniMax H3 Max model, with optional prompt enhancement and 2K upscaling stages built into each workflow - **H3 Max Turbo T2V / I2V**: Text-to-video and first/last-frame video through the fastest MiniMax H3 Max Turbo model, with the same optional prompt enhancement and 2K upscaling stages -- **H3 Max Cast Swap / Aspect Ratio Reframe**: Editing workflows that swap the people in a finished clip, or fit it to a new aspect ratio without cropping the subject ### MiniMax H3 Max -The six `api_minimax_h3_max_*` templates run on the MiniMax H3 Max model through the same API nodes: the Text to Video and First-Last-Frame nodes gained an H3 Max option in ComfyUI 0.34.2, and the Reference to Video node in 0.34.3. The Cast Swap and Aspect Ratio Reframe templates use the H3 Max editing nodes added in ComfyUI 0.44.0. Update ComfyUI to at least 0.44.0 to run all of them. +The four `api_minimax_h3_max_*` templates run on the MiniMax H3 Max model through the same API nodes: the Text to Video and First-Last-Frame nodes gained an H3 Max option in ComfyUI 0.34.2, and the Reference to Video node in 0.34.3. Update ComfyUI to at least 0.34.3 to run all of them. -H3 Max generates faster than real time and bills at lower rates than the base model: 15.09 credits/sec at 480P and 24.14 credits/sec at 768P (see the [Minimax pricing table](/tutorials/partner-nodes/pricing)). Each Max generation template adds two optional stages behind a Switch node, both off by default: +H3 Max generates faster than real time and bills at lower rates than the base model: 15.09 credits/sec at 480P and 24.14 credits/sec at 768P (see the [Minimax pricing table](/tutorials/partner-nodes/pricing)). Each Max template adds two optional stages behind a Switch node, both off by default: 1. **Prompt enhance**: Rewrites the prompt with the MiniMax H3 Context IR node before generation. Turn on the `Prompt Enhance` boolean to enable it 2. **Upscale to 2K**: Runs the MiniMax H3 Regenerate to 2K node on the first-pass video. Turn on the `Upscale to 2K` boolean to enable it @@ -68,6 +67,8 @@ Keep the following in mind when wiring a Max template: - **Match durations across nodes**: Generation and prompt enhance must use the same duration, or the stages drift out of sync. The Regenerate to 2K node has no duration input; it re-renders the submitted 768P video at its original duration - **Duplicate extra inputs**: The three nodes do not share extra context automatically. If you add a reference video, audio, or another image, plug the same wires into both the Context IR and Regenerate nodes +To swap the cast in a finished clip, reframe it, extend it or insert a new shot, use the H3 Max editing nodes: see [MiniMax H3 Max Editing](/tutorials/partner-nodes/minimax/minimax-h3-max-editing). + ### MiniMax H3 Text to Video Generate videos from text prompts through the MiniMax H3 API, with native stereo audio. @@ -272,78 +273,3 @@ Generate videos conditioned on reference media on the faster H3 Max model. The R Reference image for the workflow, or use your own images. - -#### MiniMax H3 Max: Cast Swap - -Replace the people in a finished clip without a reshoot. The Recast node keeps the motion, the framing, the cuts and the original audio, and generates the new people from your reference photos. - - - - - - Open in Comfy Cloud - - - Download JSON or search "Cast Swap: MiniMax H3 Max" in Template Library - - - - - - Source clip for the workflow, or use your own. - - - Reference photo of the new person, or use your own. - - - -Load your own clip and connect one to four photos, one per new person. By default the photos replace the main people in the clip from left to right, so connect them in the order you want. The prompt is optional: name the photos as `@Image1` to `@Image4` and say who becomes whom, plus anything that has to stay the same. - -- **Source clip**: 5 to 30 seconds, with no single shot longer than 15 seconds -- **Reference photos**: at least 256 x 256 pixels, up to four -- **Resolution**: 768P or 1080P -- **Enhance realism**: off by default. Turn it on for more natural skin texture and a better blend into the scene - -#### MiniMax H3 Max: Aspect Ratio Reframe - -Fit a clip to a different aspect ratio without cropping the subject. The source keeps its position, timing and audio on the new canvas, and the model generates the surrounding picture. - - - - - - Open in Comfy Cloud - - - Download JSON or search "Aspect Ratio Reframe: MiniMax H3 Max" in Template Library - - - - - - Source clip for the workflow, or use your own. - - - -The bundled demo turns a vertical portrait clip into a 16:9 frame. Load your own clip, pick the target aspect ratio and resolution, and use the prompt to describe what fills the new area, for example continuing the background that is already behind the subject on both sides. - -- **Source clip**: 0.5 to 300 seconds -- **Aspect ratio**: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 or 9:21 -- **Resolution**: 480P, 768P or 1080P - -#### MiniMax H3 Max editing nodes - -Six H3 Max editing nodes start from a video or a still image instead of a prompt alone and edit that material. Two of them ship with the templates above; the other four are in the node library under **Partner Nodes > video > MiniMax**, ready to drop into your own workflow. - -| Node | What it does | Starting media | -| :--- | :----------- | :------------- | -| **MiniMax H3 Max Recast** | Replaces the people in a video with people from reference photos, keeping the motion, camera, cuts and sound of the source | 5-30 second video, 1-4 photos | -| **MiniMax H3 Max Extend Video** | Continues a video with new footage described in the prompt | 1.625-60 second video | -| **MiniMax H3 Max Reframe Video** | Fits a video into a new aspect ratio and generates the area around it, keeping timing and sound | 0.5-300 second video | -| **MiniMax H3 Max Insert Video** | Replaces the section between two timestamps with a newly generated scene | Video up to 60 seconds, plus a start and resume time | -| **MiniMax H3 Max Lip Sync** | Animates a still image so the mouth follows an audio track | One image, 5-899 seconds of audio | -| **MiniMax H3 Max Camera Control** | Orbits the camera around the subject in a still image | One image, plus an end angle and distance | - -In the Recast, Extend Video and Insert Video prompts, refer to the connected references as `@Image1`, `@Audio1` or `@Video1`, numbered in connection order. The Reference to Video node accepts the same shorthand for its own references. A tag that points past the references you connected stops the run with an error. - -Rates per node and resolution are in the [Minimax pricing table](/tutorials/partner-nodes/pricing). diff --git a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx new file mode 100644 index 0000000000..8391dd1836 --- /dev/null +++ b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -0,0 +1,96 @@ +--- +title: "ComfyUI 中的 MiniMax H3 Max 编辑" +description: "使用 ComfyUI 中的 MiniMax H3 Max 编辑 API 节点,在已完成片段中替换角色、重新构图、延长或插入镜头、对口型,并环绕相机。" +sidebarTitle: "MiniMax H3 Max 编辑" +translationSourceHash: e1fd584f +translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +translationBlockHashes: + "_intro": f127db6f + "MiniMax H3 Max: Cast Swap": cd212e2a + "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 + "MiniMax H3 Max editing nodes": 43f08f03 +--- +import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; + +六个 MiniMax H3 Max 编辑节点以视频或静态图像为起点,而不只依赖提示词,并对这些素材进行编辑。与其他 H3 API 节点一样,它们通过你的 Comfy API 账户在 MiniMax 的服务器上运行;编辑节点于 ComfyUI 0.44.0 中加入。 + +其中两个自带工作流模板:**Cast Swap**(`api_minimax_h3_max_recast`)和 **Aspect Ratio Reframe**(`api_minimax_h3_max_reframe`)。其余四个位于节点库中的 **合作节点 > 视频 > MiniMax** 分类下,可直接拖入你自己的的工作流。 + +如需使用 MiniMax H3、H3 Max 和 H3 Max Turbo 生成,请参阅 [MiniMax H3 API 视频生成](/zh/tutorials/partner-nodes/minimax/minimax-h3)。 + + + + +## MiniMax H3 Max:角色替换 + +无需重新拍摄,就能替换已完成视频片段中的人物。Recast 节点会保留动作、构图、剪辑和原始音频,并根据你的参考照片生成新的人物。 + + + + + + 在 Comfy Cloud 中打开 + + + 下载 JSON,或在模板库中搜索 "Cast Swap: MiniMax H3 Max" + + + + + + 该工作流使用的来源片段,也可以使用你自己的片段。 + + + 新人物的人像参考照片,也可以使用你自己的照片。 + + + +加载你自己的片段,并连接一到四张照片,每个新人物一张。默认情况下,照片会从左到右依次替换片段中的主要人物,因此请按你想要的顺序连接它们。提示词是可选的:用 `@Image1` 到 `@Image4` 指代照片,说明谁替换谁,以及任何必须保持不变的内容。 + +- **来源片段**:时长 5 至 30 秒,单个镜头不超过 15 秒 +- **参考照片**:至少 256 x 256 像素,最多四张 +- **分辨率**:768P 或 1080P +- **增强真实感**:默认关闭。开启后可获得更自然的皮肤质感,并更好地融入场景 + +## MiniMax H3 Max:宽高比重构 + +将片段适配到不同的宽高比,同时不裁剪主体。来源片段在新的画布上保留其位置、时间与音频,由模型生成周围的画面。 + + + + + + 在 Comfy Cloud 中打开 + + + 下载 JSON,或在模板库中搜索“Aspect Ratio Reframe: MiniMax H3 Max” + + + + + + 该工作流使用的来源片段,也可以使用你自己的片段。 + + + +内置演示将一段竖直人像片段转换为 16:9 画面。加载你自己的片段,选择目标宽高比与分辨率,并用提示词描述新区域中填充的内容,例如延续主体两侧已有的背景。 + +- **来源片段**:0.5 至 300 秒 +- **宽高比**:16:9、9:16、1:1、4:3、3:4、21:9 或 9:21 +- **分辨率**:480P、768P 或 1080P + +## MiniMax H3 Max 编辑节点 + +| 节点 | 功能 | 起始媒体 | +| :--- | :----------- | :------------- | +| **MiniMax H3 Max Recast** | 用参考照片中的人物替换视频中的人物,保留来源视频的运动、相机、剪辑和声音 | 5-30 秒视频,1-4 张照片 | +| **MiniMax H3 Max Extend Video** | 用提示词中描述的新素材延续视频 | 1.625-60 秒视频 | +| **MiniMax H3 Max Reframe Video** | 将视频适配到新的比例,并生成其周围区域,保留时间节奏和声音 | 0.5-300 秒视频 | +| **MiniMax H3 Max Insert Video** | 用新生成的场景替换两个时间戳之间的片段 | 最长 60 秒的视频,外加起始时间和恢复时间 | +| **MiniMax H3 Max Lip Sync** | 让静态图像动起来,使嘴型跟随音频轨道 | 一张图像,5-899 秒音频 | +| **MiniMax H3 Max Camera Control** | 让相机围绕静态图像中的主体环绕运动 | 一张图像,外加结束角度和距离 | + +在 Recast、Extend Video 和 Insert Video 的提示词中,请以 `@Image1`、`@Audio1` 或 `@Video1` 指代已连接的参考,编号遵循连接顺序。Reference to Video 节点对其自身的参考也接受同样的简写形式。如果标签指向的参考超出你所连接的参考范围,运行将因错误而停止。 + +各节点和各分辨率的费率见 [Minimax 价格表](/zh/tutorials/partner-nodes/pricing)。 diff --git a/zh/tutorials/partner-nodes/minimax/minimax-h3.mdx b/zh/tutorials/partner-nodes/minimax/minimax-h3.mdx index f6176bd5b1..61aee5ca17 100644 --- a/zh/tutorials/partner-nodes/minimax/minimax-h3.mdx +++ b/zh/tutorials/partner-nodes/minimax/minimax-h3.mdx @@ -2,13 +2,13 @@ title: "在 ComfyUI 中使用 MiniMax H3 API 进行视频生成" description: "在 ComfyUI 中使用 MiniMax H3 API 节点,通过云端工作流进行文生视频、首尾帧视频和参考生视频,支持原生立体声音频。" sidebarTitle: "Minimax H3" -translationSourceHash: f247dec4 +translationSourceHash: fc98a97a translationFrom: tutorials/partner-nodes/minimax/minimax-h3.mdx translationBlockHashes: "_intro": a261dbf5 "What MiniMax H3 is good at": 4c84d1e7 "Example outputs": 77256fd5 - "MiniMax H3 API workflows": 272af121 + "MiniMax H3 API workflows": 9f3a305f --- import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -51,20 +51,19 @@ MiniMax H3 可生成带有**原生立体声音频**的视频:人声、音效 ## MiniMax H3 API 工作流 -模板库为 H3 基础模型随附三个 API 示例工作流,为 H3 Max 随附六个(四个生成、两个编辑),为 H3 Max Turbo 随附两个: +模板库为 H3 基础模型随附三个 API 示例工作流,为 H3 Max 随附四个,为 H3 Max Turbo 随附两个: - **文生视频**:根据文本提示词生成视频 - **首尾帧视频**:在首帧和末帧图像之间生成视频 - **参考生成视频**:根据参考图像、视频和音频生成视频 - **H3 Max T2V / I2V / FLF2V / R2V**:通过更快的 MiniMax H3 Max 模型运行相同的模式,每个工作流都内置了可选的提示词增强和 2K 上采样阶段 - **H3 Max Turbo T2V / I2V**:通过最快的 MiniMax H3 Max Turbo 模型运行文生视频与首尾帧视频,同样内置可选的提示词增强和 2K 上采样阶段 -- **H3 Max 角色替换 / 宽高比重构**:编辑工作流,用于替换已完成片段中的人物,或在不裁切主体的情况下把片段适配到新的宽高比 ### MiniMax H3 Max -这六个 `api_minimax_h3_max_*` 模板通过相同的 API 节点在 MiniMax H3 Max 模型上运行:Text to Video 和 First-Last-Frame 节点在 ComfyUI 0.34.2 中加入了 H3 Max 选项,Reference to Video 节点则在 0.34.3 中加入。角色替换和宽高比重构模板使用 ComfyUI 0.44.0 新增的 H3 Max 编辑节点。请将 ComfyUI 更新到至少 0.44.0 以运行全部模板。 +这四个 `api_minimax_h3_max_*` 模板通过相同的 API 节点在 MiniMax H3 Max 模型上运行:Text to Video 和 First-Last-Frame 节点在 ComfyUI 0.34.2 中加入了 H3 Max 选项,Reference to Video 节点则在 0.34.3 中加入。请将 ComfyUI 更新到至少 0.34.3 以运行全部模板。 -H3 Max 以快于实时的速度生成,费率低于基础模型:480P 为 15.09 积分/秒,768P 为 24.14 积分/秒(见 [Minimax 定价表](/zh/tutorials/partner-nodes/pricing))。每个 Max 生成模板在 Switch 节点后提供两个可选阶段,默认均为关闭: +H3 Max 以快于实时的速度生成,费率低于基础模型:480P 为 15.09 积分/秒,768P 为 24.14 积分/秒(见 [Minimax 定价表](/zh/tutorials/partner-nodes/pricing))。每个 Max 模板在 Switch 节点后提供两个可选阶段,默认均为关闭: 1. **提示词增强**:生成前使用 MiniMax H3 Context IR 节点改写提示词。打开 `Prompt Enhance` 布尔值即可启用 2. **上采样至 2K**:对首次生成的视频运行 MiniMax H3 Regenerate to 2K 节点。打开 `Upscale to 2K` 布尔值即可启用 @@ -75,6 +74,8 @@ H3 Max 以快于实时的速度生成,费率低于基础模型:480P 为 15.0 - **各节点时长保持一致**:生成和提示词增强必须使用相同时长,否则各阶段会不同步。Regenerate to 2K 节点没有时长输入,它会以原始时长重新渲染提交的 768P 视频 - **额外输入需重复接线**:这三个节点不会自动共享额外上下文。如果添加参考视频、音频或另一张图像,请将相同的线同时接入 Context IR 和 Regenerate 节点 +要把成片中的角色替换成其他人、改变画幅、延长片段或插入新镜头,请使用 H3 Max 编辑节点,参见 [MiniMax H3 Max 编辑](/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing)。 + ### MiniMax H3 文生视频 通过 MiniMax H3 API 从文本提示词生成视频,并输出原生立体声音频。 @@ -279,78 +280,3 @@ MiniMax 为参考驱动生成(R2V)发布了一份官方的 [全参考模式 该工作流的参考图像,也可以使用你自己的图像。 - -#### MiniMax H3 Max:角色替换 - -无需重拍就能替换已完成片段中的人物。Recast 节点保留动作、构图、剪辑点和原始音频,并根据参考照片生成新的人物。 - - - - - - 在 Comfy Cloud 中打开 - - - 下载 JSON,或在模板库中搜索“Cast Swap: MiniMax H3 Max” - - - - - - 该工作流的源片段,也可以使用你自己的视频。 - - - 新人物的参考照片,也可以使用你自己的照片。 - - - -载入你自己的片段,并连接一到四张照片,每个新人物一张。默认情况下,照片会从左到右替换片段中的主要人物,所以请按你想要的顺序连接。提示词是可选的:用 `@Image1` 到 `@Image4` 指代照片,说明谁替换谁,以及哪些内容必须保持不变。 - -- **源片段**:5 到 30 秒,单个镜头不超过 15 秒 -- **参考照片**:至少 256 x 256 像素,最多四张 -- **分辨率**:768P 或 1080P -- **增强真实感**:默认关闭。开启后皮肤纹理更自然,新人物与场景融合得更好 - -#### MiniMax H3 Max:宽高比重构 - -在不裁切主体的情况下把片段适配到新的宽高比。源内容在新画布上保持位置、时长和音频,模型负责生成周围的画面。 - - - - - - 在 Comfy Cloud 中打开 - - - 下载 JSON,或在模板库中搜索“Aspect Ratio Reframe: MiniMax H3 Max” - - - - - - 该工作流的源片段,也可以使用你自己的视频。 - - - -自带示例把竖屏片段转换成 16:9 画面。载入你自己的片段,选择目标宽高比和分辨率,再用提示词描述新区域的内容,例如延续主体两侧已有的背景。 - -- **源片段**:0.5 到 300 秒 -- **宽高比**:16:9、9:16、1:1、4:3、3:4、21:9 或 9:21 -- **分辨率**:480P、768P 或 1080P - -#### MiniMax H3 Max 编辑节点 - -六个 H3 Max 编辑节点从视频或静态图像出发,而不是只靠提示词,并对这些素材进行编辑。其中两个已随上面的模板提供,另外四个位于节点库的 **Partner Nodes > video > MiniMax** 下,可以直接拖进你自己的工作流。 - -| 节点 | 作用 | 起始素材 | -| :--- | :----------- | :------------- | -| **MiniMax H3 Max Recast** | 用参考照片中的人物替换视频中的人物,保留原视频的动作、镜头、剪辑和声音 | 5-30 秒视频,1-4 张照片 | -| **MiniMax H3 Max Extend Video** | 按提示词描述的新内容延长视频 | 1.625-60 秒视频 | -| **MiniMax H3 Max Reframe Video** | 把视频适配到新的宽高比并生成周围区域,保留时长和声音 | 0.5-300 秒视频 | -| **MiniMax H3 Max Insert Video** | 用新生成的场景替换两个时间点之间的段落 | 最长 60 秒视频,另需开始和恢复时间 | -| **MiniMax H3 Max Lip Sync** | 让静态图像的口型跟随音频 | 一张图像,5-899 秒音频 | -| **MiniMax H3 Max Camera Control** | 围绕静态图像中的主体做镜头环绕 | 一张图像,加上结束角度和距离 | - -在 Recast、Extend Video 和 Insert Video 的提示词中,用 `@Image1`、`@Audio1` 或 `@Video1` 指代已连接的参考素材,编号按连接顺序。Reference to Video 节点也接受同样的简写形式。指向超出已连接数量的标签会让运行报错停止。 - -各节点和各分辨率的费率见 [Minimax 定价表](/zh/tutorials/partner-nodes/pricing)。 From 1c8d251993ee874a7cdf80b0de683740db4e2288 Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Wed, 4 Nov 2026 19:25:24 +0800 Subject: [PATCH 3/6] docs(minimax): add the 0.4-2.5 source aspect ratio limit to Cast Swap The Recast node rejects clips outside that range and the template passes the clip straight through, so the source-clip limits have to say it. --- es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx | 4 ++-- ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx | 4 ++-- ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx | 4 ++-- tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx | 2 +- zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx | 4 ++-- 5 files changed, 9 insertions(+), 9 deletions(-) diff --git a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 6741912c6e..fa4bbb0f17 100644 --- a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,7 +2,7 @@ title: "Edición con MiniMax H3 Max en ComfyUI" description: "Cambia el reparto, reencuadra, extiende o inserta tomas, sincroniza labios y orbita la cámara en clips finalizados con los nodos API de edición MiniMax H3 Max." sidebarTitle: "Edición de Minimax H3 Max" -translationSourceHash: e1fd584f +translationSourceHash: bf58ccdc translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": f127db6f @@ -48,7 +48,7 @@ Reemplaza a las personas de un clip finalizado sin necesidad de volver a rodar. Carga tu propio clip y conecta de una a cuatro fotos, una por cada nueva persona. De forma predeterminada, las fotos reemplazan a las personas principales del clip de izquierda a derecha, así que conéctalas en el orden que quieras. El prompt es opcional: nombra las fotos como `@Image1` a `@Image4` y di quién se convierte en quién, además de todo lo que deba permanecer igual. -- **Clip de origen**: de 5 a 30 segundos, sin ningún plano individual de más de 15 segundos +- **Clip de origen**: de 5 a 30 segundos, sin ningún plano individual de más de 15 segundos, y con una relación de aspecto entre 0.4 y 2.5 - **Fotos de referencia**: al menos 256 x 256 píxeles, hasta cuatro - **Resolución**: 768P o 1080P - **Enhance realism**: desactivado de forma predeterminada. Actívalo para obtener una textura de piel más natural y una mejor integración en la escena diff --git a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index f918449123..f5c51ff636 100644 --- a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,7 +2,7 @@ title: "ComfyUI での MiniMax H3 Max 編集" description: "ComfyUI の MiniMax H3 Max 編集 APIノードを使用すると、完成したクリップのキャスト交換、フレーミング変更、ショットの延長や挿入、リップシンク、カメラの周回が行えます。" sidebarTitle: "Minimax H3 Max 編集" -translationSourceHash: e1fd584f +translationSourceHash: bf58ccdc translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": f127db6f @@ -48,7 +48,7 @@ MiniMax H3、H3 Max、H3 Max Turboでの生成については、[MiniMax H3 API ご自身のクリップを読み込み、新しい人物1人につき1枚、最大4枚までの写真を接続します。デフォルトでは、写真はクリップ内の主要な人物を左から右へ置き換えるため、希望する順番に接続してください。プロンプトは任意です。写真を `@Image1` から `@Image4` として指定し、誰が誰になるのか、また変わらずに保つ必要があるものを記述してください。 -- **ソースクリップ**: 5~30秒、1カットが15秒を超えないこと +- **ソースクリップ**: 5~30秒、1カットが15秒を超えないこと、アスペクト比は0.4~2.5の範囲であること - **参照写真**: 最小 256 x 256 ピクセル、最大4枚 - **解像度**: 768P または 1080P - **リアリズムを強化**: デフォルトはオフ。より自然な肌の質感とシーンへのより良いなじみを得るにはオンにします diff --git a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 3e1997e596..48f192a68e 100644 --- a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,7 +2,7 @@ title: "ComfyUI의 MiniMax H3 Max 편집" description: "ComfyUI의 MiniMax H3 Max 편집 API 노드로 완성된 클립에서 출연진을 교체하고, 구도를 재설정하고, 장면을 연장하거나 삽입하고, 립 싱크를 맞추고, 카메라를 회전시킬 수 있습니다." sidebarTitle: "Minimax H3 Max 편집" -translationSourceHash: e1fd584f +translationSourceHash: bf58ccdc translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": f127db6f @@ -48,7 +48,7 @@ MiniMax H3, H3 Max, H3 Max Turbo로 생성하려면 [MiniMax H3 API 비디오 직접 준비한 클립을 불러오고, 새 인물마다 한 장씩 1~4장의 사진을 연결합니다. 기본적으로 사진은 클립의 주요 인물을 왼쪽에서 오른쪽 순서로 교체하므로, 원하는 순서대로 연결하세요. 프롬프트는 선택 사항입니다. 사진의 이름을 `@Image1`부터 `@Image4`까지로 지정하고, 누가 누구로 바뀌는지와 함께 그대로 유지되어야 하는 요소를 설명하세요. -- **소스 클립**: 5~30초, 단일 샷은 15초를 넘지 않아야 합니다 +- **소스 클립**: 5~30초, 단일 샷은 15초를 넘지 않아야 하며 가로세로 비율이 0.4~2.5 사이여야 합니다 - **참조 사진**: 최소 256 x 256 픽셀, 최대 4장 - **해상도**: 768P 또는 1080P - **사실감 향상**: 기본적으로 꺼짐입니다. 더 자연스러운 피부 질감과 장면에 더 잘 어우러지는 결과를 원하면 켜세요 diff --git a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index d143575209..6097784949 100644 --- a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -41,7 +41,7 @@ Replace the people in a finished clip without a reshoot. The Recast node keeps t Load your own clip and connect one to four photos, one per new person. By default the photos replace the main people in the clip from left to right, so connect them in the order you want. The prompt is optional: name the photos as `@Image1` to `@Image4` and say who becomes whom, plus anything that has to stay the same. -- **Source clip**: 5 to 30 seconds, with no single shot longer than 15 seconds +- **Source clip**: 5 to 30 seconds, with no single shot longer than 15 seconds, and an aspect ratio between 0.4 and 2.5 - **Reference photos**: at least 256 x 256 pixels, up to four - **Resolution**: 768P or 1080P - **Enhance realism**: off by default. Turn it on for more natural skin texture and a better blend into the scene diff --git a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 8391dd1836..0b5bfff762 100644 --- a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,7 +2,7 @@ title: "ComfyUI 中的 MiniMax H3 Max 编辑" description: "使用 ComfyUI 中的 MiniMax H3 Max 编辑 API 节点,在已完成片段中替换角色、重新构图、延长或插入镜头、对口型,并环绕相机。" sidebarTitle: "MiniMax H3 Max 编辑" -translationSourceHash: e1fd584f +translationSourceHash: bf58ccdc translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": f127db6f @@ -48,7 +48,7 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; 加载你自己的片段,并连接一到四张照片,每个新人物一张。默认情况下,照片会从左到右依次替换片段中的主要人物,因此请按你想要的顺序连接它们。提示词是可选的:用 `@Image1` 到 `@Image4` 指代照片,说明谁替换谁,以及任何必须保持不变的内容。 -- **来源片段**:时长 5 至 30 秒,单个镜头不超过 15 秒 +- **来源片段**:时长 5 至 30 秒,单个镜头不超过 15 秒,宽高比在 0.4 到 2.5 之间 - **参考照片**:至少 256 x 256 像素,最多四张 - **分辨率**:768P 或 1080P - **增强真实感**:默认关闭。开启后可获得更自然的皮肤质感,并更好地融入场景 From de4198bea351f05d12b4aa21a13cdf865967269a Mon Sep 17 00:00:00 2001 From: "Daxiong (Lin)" Date: Mon, 9 Nov 2026 15:58:14 +0800 Subject: [PATCH 4/6] docs: note the third H3 Max editing template and the Extend Turbo option --- .../minimax/minimax-h3-max-editing.mdx | 16 ++++++++-------- .../minimax/minimax-h3-max-editing.mdx | 14 +++++++------- .../minimax/minimax-h3-max-editing.mdx | 14 +++++++------- .../minimax/minimax-h3-max-editing.mdx | 8 ++++---- .../minimax/minimax-h3-max-editing.mdx | 16 ++++++++-------- 5 files changed, 34 insertions(+), 34 deletions(-) diff --git a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index fa4bbb0f17..442e33d6f9 100644 --- a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,20 +2,20 @@ title: "Edición con MiniMax H3 Max en ComfyUI" description: "Cambia el reparto, reencuadra, extiende o inserta tomas, sincroniza labios y orbita la cámara en clips finalizados con los nodos API de edición MiniMax H3 Max." sidebarTitle: "Edición de Minimax H3 Max" -translationSourceHash: bf58ccdc +translationSourceHash: 810d7886 translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: - "_intro": f127db6f - "MiniMax H3 Max: Cast Swap": cd212e2a + "_intro": bf838aaa + "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 43f08f03 + "MiniMax H3 Max editing nodes": 64b7f083 --- import ReqHint from "/snippets/es/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/es/tutorials/update-reminder.mdx"; -Seis nodos de edición de MiniMax H3 Max parten de un video o de una imagen fija en lugar de solo un prompt y editan ese material. Se ejecutan en los servidores de MiniMax a través de tu cuenta de Comfy API, igual que los demás nodos API de H3, y los nodos de edición se agregaron en ComfyUI 0.44.0. +Seis nodos de edición de MiniMax H3 Max parten de un video o de una imagen fija y editan ese material. Se ejecutan a través de tu cuenta de Comfy API, igual que los demás nodos API de H3, y se agregaron en ComfyUI 0.44.0. -Dos de ellos incluyen plantillas de flujo de trabajo: **Cast Swap** (`api_minimax_h3_max_recast`) y **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`). Los otros cuatro están en la biblioteca de nodos, dentro de **Nodos asociados > video > MiniMax**, listos para agregarlos a tu propio flujo de trabajo. +Tres de ellos incluyen plantillas de flujo de trabajo: **Cast Swap** (`api_minimax_h3_max_recast`), **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`) y **Camera Control** (`api_minimax_h3_max_camera_control`). Los otros tres están en la biblioteca de nodos, dentro de **Nodos asociados > video > MiniMax**, listos para agregarlos a tu propio flujo de trabajo. Las dos primeras se explican a continuación; la [plantilla Camera Control](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/api_minimax_h3_max_camera_control.json) orbita alrededor de una imagen fija y añade una base musical al clip. Para generar con MiniMax H3, H3 Max y H3 Max Turbo, consulta [Generación de video con la API de MiniMax H3](/es/tutorials/partner-nodes/minimax/minimax-h3). @@ -24,7 +24,7 @@ Para generar con MiniMax H3, H3 Max y H3 Max Turbo, consulta [Generación de vid ## MiniMax H3 Max: Cast Swap -Reemplaza a las personas de un clip finalizado sin necesidad de volver a rodar. El nodo Recast conserva el movimiento, el encuadre, los cortes y el audio original, y genera a las nuevas personas a partir de tus fotos de referencia. +Reemplaza a las personas de un clip finalizado sin necesidad de volver a rodar: el nodo Recast conserva el movimiento, el encuadre, los cortes y el audio original, y genera a las nuevas personas a partir de tus fotos de referencia. @@ -85,7 +85,7 @@ La demo incluida convierte un clip vertical de retrato en un encuadre de 16:9. C | Nodo | Qué hace | Medio inicial | | :--- | :----------- | :------------- | | **MiniMax H3 Max Recast** | Reemplaza a las personas de un video con personas de fotos de referencia, manteniendo el movimiento, la cámara, los cortes y el sonido de la fuente | Video de 5-30 segundos, 1-4 fotos | -| **MiniMax H3 Max Extend Video** | Continúa un video con metraje nuevo descrito en el prompt | Video de 1.625-60 segundos | +| **MiniMax H3 Max Extend Video** | Continúa un video con metraje nuevo descrito en el prompt | Video de 1.625-60 segundos (H3 Max o H3 Max Turbo) | | **MiniMax H3 Max Reframe Video** | Ajusta un video a una nueva relación de aspecto y genera el área que lo rodea, manteniendo la sincronización y el sonido | Video de 0.5-300 segundos | | **MiniMax H3 Max Insert Video** | Reemplaza la sección entre dos marcas de tiempo con una escena recién generada | Video de hasta 60 segundos, más un tiempo de inicio y de reanudación | | **MiniMax H3 Max Lip Sync** | Anima una imagen fija para que la boca siga una pista de audio | Una imagen, 5-899 segundos de audio | diff --git a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index f5c51ff636..18bae88afe 100644 --- a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,20 +2,20 @@ title: "ComfyUI での MiniMax H3 Max 編集" description: "ComfyUI の MiniMax H3 Max 編集 APIノードを使用すると、完成したクリップのキャスト交換、フレーミング変更、ショットの延長や挿入、リップシンク、カメラの周回が行えます。" sidebarTitle: "Minimax H3 Max 編集" -translationSourceHash: bf58ccdc +translationSourceHash: 810d7886 translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: - "_intro": f127db6f - "MiniMax H3 Max: Cast Swap": cd212e2a + "_intro": bf838aaa + "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 43f08f03 + "MiniMax H3 Max editing nodes": 64b7f083 --- import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; -MiniMax H3 Maxの編集系ノード6種は、プロンプトだけではなくビデオや静止画を出発点として、そのマテリアルを編集します。他のH3 APIノードと同様に、お使いのComfy APIアカウントを通じてMiniMaxのサーバー上で実行され、これらの編集系ノードはComfyUI 0.44.0で追加されました。 +MiniMax H3 Maxの編集系ノード6種は、ビデオや静止画を出発点としてそのマテリアルを編集します。他のH3 APIノードと同様にComfy APIアカウントを通じて実行され、ComfyUI 0.44.0で追加されました。 -このうち2つにはワークフローテンプレートが付属しています。**Cast Swap**(`api_minimax_h3_max_recast`)と**Aspect Ratio Reframe**(`api_minimax_h3_max_reframe`)です。残りの4つは**Partner Nodes > video > MiniMax**のノードライブラリにあり、自身のワークフローにそのまま追加できます。 +このうち3つにはワークフローテンプレートが付属しています。**Cast Swap**(`api_minimax_h3_max_recast`)、**Aspect Ratio Reframe**(`api_minimax_h3_max_reframe`)、**Camera Control**(`api_minimax_h3_max_camera_control`)です。残りの3つは**Partner Nodes > video > MiniMax**のノードライブラリにあり、自身のワークフローにそのまま追加できます。最初の2つは以下で解説します。[Camera Controlのテンプレート](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/api_minimax_h3_max_camera_control.json)は静止画の周りをカメラが周回し、クリップに音楽を重ねます。 MiniMax H3、H3 Max、H3 Max Turboでの生成については、[MiniMax H3 API ビデオ生成](/ja/tutorials/partner-nodes/minimax/minimax-h3)を参照してください。 @@ -85,7 +85,7 @@ MiniMax H3、H3 Max、H3 Max Turboでの生成については、[MiniMax H3 API | ノード | 機能 | 入力メディア | | :--- | :----------- | :------------- | | **MiniMax H3 Max Recast** | ビデオ内の人物を参照写真の人物に置き換えます。動き、カメラワーク、カット、ソースのサウンドは維持されます | 5-30秒のビデオ、1-4枚の写真 | -| **MiniMax H3 Max Extend Video** | プロンプトで記述した新しい映像でビデオを続けます | 1.625-60秒のビデオ | +| **MiniMax H3 Max Extend Video** | プロンプトで記述した新しい映像でビデオを続けます | 1.625-60秒のビデオ(H3 Max または H3 Max Turbo) | | **MiniMax H3 Max Reframe Video** | ビデオを新しいアスペクト比に収め、その周囲の領域を生成します。タイミングとサウンドは維持されます | 0.5-300秒のビデオ | | **MiniMax H3 Max Insert Video** | 2つのタイムスタンプの間の区間を、新しく生成したシーンに置き換えます | 最大60秒のビデオ、および開始時刻と再開時刻 | | **MiniMax H3 Max Lip Sync** | 静止画像をアニメーション化し、口元がオーディオトラックに合わせて動くようにします | 画像1枚、5-899秒のオーディオ | diff --git a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 48f192a68e..93337898ff 100644 --- a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,20 +2,20 @@ title: "ComfyUI의 MiniMax H3 Max 편집" description: "ComfyUI의 MiniMax H3 Max 편집 API 노드로 완성된 클립에서 출연진을 교체하고, 구도를 재설정하고, 장면을 연장하거나 삽입하고, 립 싱크를 맞추고, 카메라를 회전시킬 수 있습니다." sidebarTitle: "Minimax H3 Max 편집" -translationSourceHash: bf58ccdc +translationSourceHash: 810d7886 translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: - "_intro": f127db6f - "MiniMax H3 Max: Cast Swap": cd212e2a + "_intro": bf838aaa + "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 43f08f03 + "MiniMax H3 Max editing nodes": 64b7f083 --- import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; -여섯 개의 MiniMax H3 Max 편집 노드는 프롬프트만 사용하는 대신 비디오나 정지 이미지에서 시작하여 해당 소재를 편집합니다. 이 노드들은 다른 H3 API 노드와 마찬가지로 Comfy API 계정을 통해 MiniMax의 서버에서 실행되며, 편집 노드는 ComfyUI 0.44.0에서 추가되었습니다. +여섯 개의 MiniMax H3 Max 편집 노드는 비디오나 정지 이미지를 출발점으로 삼아 해당 소재를 편집합니다. 다른 H3 API 노드와 마찬가지로 Comfy API 계정을 통해 실행되며, ComfyUI 0.44.0에서 추가되었습니다. -이 중 두 개는 워크플로 템플릿과 함께 제공됩니다: **Cast Swap** (`api_minimax_h3_max_recast`)과 **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`)입니다. 나머지 네 개는 **Partner Nodes > video > MiniMax** 아래 노드 라이브러리에 있으며, 자신의 워크플로에 바로 추가할 수 있습니다. +이 중 세 개는 워크플로 템플릿과 함께 제공됩니다: **Cast Swap** (`api_minimax_h3_max_recast`), **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`), **Camera Control** (`api_minimax_h3_max_camera_control`)입니다. 나머지 세 개는 **Partner Nodes > video > MiniMax** 아래 노드 라이브러리에 있으며, 자신의 워크플로에 바로 추가할 수 있습니다. 처음 두 개는 아래에서 설명합니다. [Camera Control 템플릿](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/api_minimax_h3_max_camera_control.json)은 정지 이미지 주위로 카메라를 회전시키고 클립에 음악을 깔아 줍니다. MiniMax H3, H3 Max, H3 Max Turbo로 생성하려면 [MiniMax H3 API 비디오 생성](/ko/tutorials/partner-nodes/minimax/minimax-h3)을 참조하세요. @@ -85,7 +85,7 @@ MiniMax H3, H3 Max, H3 Max Turbo로 생성하려면 [MiniMax H3 API 비디오 | 노드 | 기능 | 시작 미디어 | | :--- | :----------- | :------------- | | **MiniMax H3 Max Recast** | 비디오 속 인물을 레퍼런스 사진의 인물로 교체하며, 소스의 움직임, 카메라, 컷, 사운드를 유지합니다 | 5-30초 비디오, 사진 1-4장 | -| **MiniMax H3 Max Extend Video** | 프롬프트에 설명된 새로운 영상으로 비디오를 이어서 생성합니다 | 1.625-60초 비디오 | +| **MiniMax H3 Max Extend Video** | 프롬프트에 설명된 새로운 영상으로 비디오를 이어서 생성합니다 | 1.625-60초 비디오(H3 Max 또는 H3 Max Turbo) | | **MiniMax H3 Max Reframe Video** | 비디오를 새로운 화면 비율에 맞추고 주변 영역을 생성하며, 타이밍과 사운드를 유지합니다 | 0.5-300초 비디오 | | **MiniMax H3 Max Insert Video** | 두 타임스탬프 사이의 구간을 새로 생성된 장면으로 교체합니다 | 최대 60초 비디오와 시작 및 재개 시간 | | **MiniMax H3 Max Lip Sync** | 정지 이미지를 애니메이션으로 만들어 입 모양이 오디오 트랙을 따라가게 합니다 | 이미지 1장, 5-899초 오디오 | diff --git a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 6097784949..37e2294183 100644 --- a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -6,9 +6,9 @@ sidebarTitle: "Minimax H3 Max Editing" import ReqHint from "/snippets/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; -Six MiniMax H3 Max editing nodes start from a video or a still image instead of a prompt alone and edit that material. They run on MiniMax's servers through your Comfy API account like the other H3 API nodes, and the editing nodes were added in ComfyUI 0.44.0. +Six MiniMax H3 Max editing nodes take a video or a still image as their starting point and edit that material. They run through your Comfy API account like the other H3 API nodes, and were added in ComfyUI 0.44.0. -Two of them ship with workflow templates: **Cast Swap** (`api_minimax_h3_max_recast`) and **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`). The other four are in the node library under **Partner Nodes > video > MiniMax**, ready to drop into your own workflow. +Three of them ship with workflow templates: **Cast Swap** (`api_minimax_h3_max_recast`), **Aspect Ratio Reframe** (`api_minimax_h3_max_reframe`) and **Camera Control** (`api_minimax_h3_max_camera_control`). The other three are in the node library under **Partner Nodes > video > MiniMax**, ready to drop into your own workflow. The first two are walked through below; the [Camera Control template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/api_minimax_h3_max_camera_control.json) orbits a still image and lays a music bed under the clip. For generation with MiniMax H3, H3 Max, and H3 Max Turbo, see [MiniMax H3 API Video Generation](/tutorials/partner-nodes/minimax/minimax-h3). @@ -17,7 +17,7 @@ For generation with MiniMax H3, H3 Max, and H3 Max Turbo, see [MiniMax H3 API Vi ## MiniMax H3 Max: Cast Swap -Replace the people in a finished clip without a reshoot. The Recast node keeps the motion, the framing, the cuts and the original audio, and generates the new people from your reference photos. +Replace the people in a finished clip without a reshoot: the node keeps the motion, the framing, the cuts and the original audio, and generates the new people from your reference photos. @@ -78,7 +78,7 @@ The bundled demo turns a vertical portrait clip into a 16:9 frame. Load your own | Node | What it does | Starting media | | :--- | :----------- | :------------- | | **MiniMax H3 Max Recast** | Replaces the people in a video with people from reference photos, keeping the motion, camera, cuts and sound of the source | 5-30 second video, 1-4 photos | -| **MiniMax H3 Max Extend Video** | Continues a video with new footage described in the prompt | 1.625-60 second video | +| **MiniMax H3 Max Extend Video** | Continues a video with new footage described in the prompt | 1.625-60 second video, on H3 Max or H3 Max Turbo | | **MiniMax H3 Max Reframe Video** | Fits a video into a new aspect ratio and generates the area around it, keeping timing and sound | 0.5-300 second video | | **MiniMax H3 Max Insert Video** | Replaces the section between two timestamps with a newly generated scene | Video up to 60 seconds, plus a start and resume time | | **MiniMax H3 Max Lip Sync** | Animates a still image so the mouth follows an audio track | One image, 5-899 seconds of audio | diff --git a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 0b5bfff762..f5091a5996 100644 --- a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,20 +2,20 @@ title: "ComfyUI 中的 MiniMax H3 Max 编辑" description: "使用 ComfyUI 中的 MiniMax H3 Max 编辑 API 节点,在已完成片段中替换角色、重新构图、延长或插入镜头、对口型,并环绕相机。" sidebarTitle: "MiniMax H3 Max 编辑" -translationSourceHash: bf58ccdc +translationSourceHash: 810d7886 translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: - "_intro": f127db6f - "MiniMax H3 Max: Cast Swap": cd212e2a + "_intro": bf838aaa + "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 43f08f03 + "MiniMax H3 Max editing nodes": 64b7f083 --- import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; -六个 MiniMax H3 Max 编辑节点以视频或静态图像为起点,而不只依赖提示词,并对这些素材进行编辑。与其他 H3 API 节点一样,它们通过你的 Comfy API 账户在 MiniMax 的服务器上运行;编辑节点于 ComfyUI 0.44.0 中加入。 +六个 MiniMax H3 Max 编辑节点以视频或静态图像为起点,并对该素材进行编辑。与其他 H3 API 节点一样,它们通过你的 Comfy API 账户运行,于 ComfyUI 0.44.0 中加入。 -其中两个自带工作流模板:**Cast Swap**(`api_minimax_h3_max_recast`)和 **Aspect Ratio Reframe**(`api_minimax_h3_max_reframe`)。其余四个位于节点库中的 **合作节点 > 视频 > MiniMax** 分类下,可直接拖入你自己的的工作流。 +其中三个自带工作流模板:**Cast Swap**(`api_minimax_h3_max_recast`)、**Aspect Ratio Reframe**(`api_minimax_h3_max_reframe`)和 **Camera Control**(`api_minimax_h3_max_camera_control`)。其余三个位于节点库中的 **合作节点 > 视频 > MiniMax** 分类下,可直接拖入你的工作流。前两个模板见下文;[Camera Control 模板](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/api_minimax_h3_max_camera_control.json)会让相机围绕静态图像环绕,并为片段配上音乐。 如需使用 MiniMax H3、H3 Max 和 H3 Max Turbo 生成,请参阅 [MiniMax H3 API 视频生成](/zh/tutorials/partner-nodes/minimax/minimax-h3)。 @@ -24,7 +24,7 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; ## MiniMax H3 Max:角色替换 -无需重新拍摄,就能替换已完成视频片段中的人物。Recast 节点会保留动作、构图、剪辑和原始音频,并根据你的参考照片生成新的人物。 +无需重新拍摄,就能替换已完成视频片段中的人物:Recast 节点会保留动作、构图、剪辑和原始音频,并根据你的参考照片生成新的人物。 @@ -85,7 +85,7 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; | 节点 | 功能 | 起始媒体 | | :--- | :----------- | :------------- | | **MiniMax H3 Max Recast** | 用参考照片中的人物替换视频中的人物,保留来源视频的运动、相机、剪辑和声音 | 5-30 秒视频,1-4 张照片 | -| **MiniMax H3 Max Extend Video** | 用提示词中描述的新素材延续视频 | 1.625-60 秒视频 | +| **MiniMax H3 Max Extend Video** | 用提示词中描述的新素材延续视频 | 1.625-60 秒视频,可选 H3 Max 或 H3 Max Turbo | | **MiniMax H3 Max Reframe Video** | 将视频适配到新的比例,并生成其周围区域,保留时间节奏和声音 | 0.5-300 秒视频 | | **MiniMax H3 Max Insert Video** | 用新生成的场景替换两个时间戳之间的片段 | 最长 60 秒的视频,外加起始时间和恢复时间 | | **MiniMax H3 Max Lip Sync** | 让静态图像动起来,使嘴型跟随音频轨道 | 一张图像,5-899 秒音频 | From 59003781e1c7a16d8560ae45589ef5465b9d9287 Mon Sep 17 00:00:00 2001 From: "Daxiong (Lin)" Date: Mon, 9 Nov 2026 17:00:30 +0800 Subject: [PATCH 5/6] docs: scope MiniMax H3 Max reference tags per node and fix locale titles --- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 8 ++++---- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 8 ++++---- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 8 ++++---- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 2 +- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 8 ++++---- 5 files changed, 17 insertions(+), 17 deletions(-) diff --git a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 442e33d6f9..62b9623ef4 100644 --- a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -1,14 +1,14 @@ --- -title: "Edición con MiniMax H3 Max en ComfyUI" +title: "MiniMax H3 Max en ComfyUI: edición" description: "Cambia el reparto, reencuadra, extiende o inserta tomas, sincroniza labios y orbita la cámara en clips finalizados con los nodos API de edición MiniMax H3 Max." sidebarTitle: "Edición de Minimax H3 Max" -translationSourceHash: 810d7886 +translationSourceHash: ba2b10f5 translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": bf838aaa "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 64b7f083 + "MiniMax H3 Max editing nodes": 2757b0aa --- import ReqHint from "/snippets/es/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/es/tutorials/update-reminder.mdx"; @@ -91,6 +91,6 @@ La demo incluida convierte un clip vertical de retrato en un encuadre de 16:9. C | **MiniMax H3 Max Lip Sync** | Anima una imagen fija para que la boca siga una pista de audio | Una imagen, 5-899 segundos de audio | | **MiniMax H3 Max Camera Control** | Orbita la cámara alrededor del sujeto en una imagen fija | Una imagen, más un ángulo y una distancia finales | -En los prompts de Recast, Extend Video e Insert Video, haz referencia a las referencias conectadas como `@Image1`, `@Audio1` o `@Video1`, numeradas en el orden de conexión. El nodo Reference to Video acepta la misma forma abreviada para sus propias referencias. Una etiqueta que apunta más allá de las referencias que conectaste detiene la ejecución con un error. +En los prompts, haz referencia a las referencias conectadas como `@Image1`, `@Video1` o `@Audio1`, numeradas en el orden de conexión. Recast comprueba las etiquetas `@Image`, Extend Video las `@Audio`, Insert Video las `@Image` y `@Video`, y Reference to Video las tres. Una etiqueta de un tipo comprobado que apunta más allá de las referencias conectadas de ese tipo detiene la ejecución con un error; las etiquetas de otro tipo se envían tal cual. Las tarifas por nodo y resolución están en la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing). diff --git a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 18bae88afe..cfcd5861a3 100644 --- a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -1,14 +1,14 @@ --- -title: "ComfyUI での MiniMax H3 Max 編集" +title: "MiniMax H3 Max: ComfyUI 編集" description: "ComfyUI の MiniMax H3 Max 編集 APIノードを使用すると、完成したクリップのキャスト交換、フレーミング変更、ショットの延長や挿入、リップシンク、カメラの周回が行えます。" sidebarTitle: "Minimax H3 Max 編集" -translationSourceHash: 810d7886 +translationSourceHash: ba2b10f5 translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": bf838aaa "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 64b7f083 + "MiniMax H3 Max editing nodes": 2757b0aa --- import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -91,6 +91,6 @@ MiniMax H3、H3 Max、H3 Max Turboでの生成については、[MiniMax H3 API | **MiniMax H3 Max Lip Sync** | 静止画像をアニメーション化し、口元がオーディオトラックに合わせて動くようにします | 画像1枚、5-899秒のオーディオ | | **MiniMax H3 Max Camera Control** | 静止画像内の被写体の周りをカメラが周回します | 画像1枚、および終了角度と距離 | -Recast、Extend Video、Insert Videoのプロンプトでは、接続した参照を `@Image1`、`@Audio1`、`@Video1` として参照します。番号は接続順に付けられます。Reference to Videoノードも、自身の参照に同じ省略記法を使用できます。接続した参照を超えるタグを指定すると、エラーで実行が停止します。 +プロンプトでは、接続した参照を `@Image1`、`@Video1`、`@Audio1` として参照します。番号は接続順に付けられます。Recast は `@Image` タグを、Extend Video は `@Audio` タグを、Insert Video は `@Image` と `@Video` のタグを、Reference to Video は3種類すべてのタグを検査します。検査対象のタグが、その種類で接続した参照の範囲を超えると、エラーで実行が停止します。それ以外の種類のタグはそのまま渡されます。 ノードおよび解像度ごとの料金は [Minimax料金表](/ja/tutorials/partner-nodes/pricing) に記載されています。 diff --git a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 93337898ff..e1748de8cf 100644 --- a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -1,14 +1,14 @@ --- -title: "ComfyUI의 MiniMax H3 Max 편집" +title: "MiniMax H3 Max: ComfyUI 영상 편집" description: "ComfyUI의 MiniMax H3 Max 편집 API 노드로 완성된 클립에서 출연진을 교체하고, 구도를 재설정하고, 장면을 연장하거나 삽입하고, 립 싱크를 맞추고, 카메라를 회전시킬 수 있습니다." sidebarTitle: "Minimax H3 Max 편집" -translationSourceHash: 810d7886 +translationSourceHash: ba2b10f5 translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": bf838aaa "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 64b7f083 + "MiniMax H3 Max editing nodes": 2757b0aa --- import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -91,6 +91,6 @@ MiniMax H3, H3 Max, H3 Max Turbo로 생성하려면 [MiniMax H3 API 비디오 | **MiniMax H3 Max Lip Sync** | 정지 이미지를 애니메이션으로 만들어 입 모양이 오디오 트랙을 따라가게 합니다 | 이미지 1장, 5-899초 오디오 | | **MiniMax H3 Max Camera Control** | 정지 이미지 속 피사체 주위로 카메라를 회전시킵니다 | 이미지 1장과 종료 각도 및 거리 | -Recast, Extend Video, Insert Video 프롬프트에서는 연결된 레퍼런스를 연결 순서대로 번호가 매겨진 `@Image1`, `@Audio1`, `@Video1`로 참조하세요. 레퍼런스 기반 비디오 생성 노드도 자체 레퍼런스에 동일한 축약 표기를 사용할 수 있습니다. 연결한 레퍼런스 범위를 벗어난 태그를 지정하면 실행이 오류와 함께 중단됩니다. +프롬프트에서는 연결된 레퍼런스를 연결 순서대로 번호가 매겨진 `@Image1`, `@Video1`, `@Audio1`로 참조하세요. Recast는 `@Image` 태그를, Extend Video는 `@Audio` 태그를, Insert Video는 `@Image`와 `@Video` 태그를, Reference to Video는 세 가지 태그를 모두 검사합니다. 검사 대상 태그가 해당 종류로 연결한 레퍼런스 범위를 벗어나면 실행이 오류와 함께 중단됩니다. 다른 종류의 태그는 그대로 전달됩니다. 노드 및 해상도별 요금은 [Minimax 가격표](/ko/tutorials/partner-nodes/pricing)에 나와 있습니다. diff --git a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 37e2294183..ed339a16c2 100644 --- a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -84,6 +84,6 @@ The bundled demo turns a vertical portrait clip into a 16:9 frame. Load your own | **MiniMax H3 Max Lip Sync** | Animates a still image so the mouth follows an audio track | One image, 5-899 seconds of audio | | **MiniMax H3 Max Camera Control** | Orbits the camera around the subject in a still image | One image, plus an end angle and distance | -In the Recast, Extend Video and Insert Video prompts, refer to the connected references as `@Image1`, `@Audio1` or `@Video1`, numbered in connection order. The Reference to Video node accepts the same shorthand for its own references. A tag that points past the references you connected stops the run with an error. +In prompts, refer to connected references as `@Image1`, `@Video1` or `@Audio1`, numbered in connection order. Recast checks `@Image` tags, Extend Video checks `@Audio` tags, Insert Video checks `@Image` and `@Video` tags, and Reference to Video checks all three. A tag of a checked kind that points past the connected references of that kind stops the run with an error; a tag of another kind is passed through unchanged. Rates per node and resolution are in the [Minimax pricing table](/tutorials/partner-nodes/pricing). diff --git a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index f5091a5996..152a005f8d 100644 --- a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -1,14 +1,14 @@ --- -title: "ComfyUI 中的 MiniMax H3 Max 编辑" +title: "MiniMax H3 Max:ComfyUI 编辑" description: "使用 ComfyUI 中的 MiniMax H3 Max 编辑 API 节点,在已完成片段中替换角色、重新构图、延长或插入镜头、对口型,并环绕相机。" sidebarTitle: "MiniMax H3 Max 编辑" -translationSourceHash: 810d7886 +translationSourceHash: ba2b10f5 translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": bf838aaa "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 64b7f083 + "MiniMax H3 Max editing nodes": 2757b0aa --- import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -91,6 +91,6 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; | **MiniMax H3 Max Lip Sync** | 让静态图像动起来,使嘴型跟随音频轨道 | 一张图像,5-899 秒音频 | | **MiniMax H3 Max Camera Control** | 让相机围绕静态图像中的主体环绕运动 | 一张图像,外加结束角度和距离 | -在 Recast、Extend Video 和 Insert Video 的提示词中,请以 `@Image1`、`@Audio1` 或 `@Video1` 指代已连接的参考,编号遵循连接顺序。Reference to Video 节点对其自身的参考也接受同样的简写形式。如果标签指向的参考超出你所连接的参考范围,运行将因错误而停止。 +在提示词中,请以 `@Image1`、`@Video1` 或 `@Audio1` 指代已连接的参考,编号遵循连接顺序。Recast 检查 `@Image` 标签,Extend Video 检查 `@Audio` 标签,Insert Video 检查 `@Image` 和 `@Video` 标签,Reference to Video 则检查这三种标签。如果某个受检查类型的标签指向的参考超出该类型已连接的参考数量,运行将因错误而停止;其他类型的标签会原样传递。 各节点和各分辨率的费率见 [Minimax 价格表](/zh/tutorials/partner-nodes/pricing)。 From 14020e464a7c84fb5dc40de7fbead6eca6527355 Mon Sep 17 00:00:00 2001 From: "Daxiong (Lin)" Date: Mon, 9 Nov 2026 17:08:42 +0800 Subject: [PATCH 6/6] docs: limit MiniMax H3 Max reference-tag guidance to the nodes that read tags --- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 6 +++--- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 6 +++--- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 6 +++--- tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx | 2 +- .../partner-nodes/minimax/minimax-h3-max-editing.mdx | 6 +++--- 5 files changed, 13 insertions(+), 13 deletions(-) diff --git a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 62b9623ef4..04b4136925 100644 --- a/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/es/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,13 +2,13 @@ title: "MiniMax H3 Max en ComfyUI: edición" description: "Cambia el reparto, reencuadra, extiende o inserta tomas, sincroniza labios y orbita la cámara en clips finalizados con los nodos API de edición MiniMax H3 Max." sidebarTitle: "Edición de Minimax H3 Max" -translationSourceHash: ba2b10f5 +translationSourceHash: fda273ce translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": bf838aaa "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 2757b0aa + "MiniMax H3 Max editing nodes": 4c3396f2 --- import ReqHint from "/snippets/es/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/es/tutorials/update-reminder.mdx"; @@ -91,6 +91,6 @@ La demo incluida convierte un clip vertical de retrato en un encuadre de 16:9. C | **MiniMax H3 Max Lip Sync** | Anima una imagen fija para que la boca siga una pista de audio | Una imagen, 5-899 segundos de audio | | **MiniMax H3 Max Camera Control** | Orbita la cámara alrededor del sujeto en una imagen fija | Una imagen, más un ángulo y una distancia finales | -En los prompts, haz referencia a las referencias conectadas como `@Image1`, `@Video1` o `@Audio1`, numeradas en el orden de conexión. Recast comprueba las etiquetas `@Image`, Extend Video las `@Audio`, Insert Video las `@Image` y `@Video`, y Reference to Video las tres. Una etiqueta de un tipo comprobado que apunta más allá de las referencias conectadas de ese tipo detiene la ejecución con un error; las etiquetas de otro tipo se envían tal cual. +En los prompts de Recast, Extend Video e Insert Video, haz referencia a las referencias conectadas como `@Image1`, `@Video1` o `@Audio1`, numeradas en el orden de conexión. Recast comprueba las etiquetas `@Image`, Extend Video las `@Audio` e Insert Video las `@Image` y `@Video`; una etiqueta de un tipo comprobado que apunta más allá de las referencias conectadas de ese tipo detiene la ejecución con un error, y las de otro tipo se envían tal cual. El nodo de generación [Reference to Video](/es/tutorials/partner-nodes/minimax/minimax-h3#minimax-h3-referencia-a-video) resuelve las tres. Las tarifas por nodo y resolución están en la [tabla de precios de Minimax](/es/tutorials/partner-nodes/pricing). diff --git a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index cfcd5861a3..bc506d0853 100644 --- a/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/ja/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,13 +2,13 @@ title: "MiniMax H3 Max: ComfyUI 編集" description: "ComfyUI の MiniMax H3 Max 編集 APIノードを使用すると、完成したクリップのキャスト交換、フレーミング変更、ショットの延長や挿入、リップシンク、カメラの周回が行えます。" sidebarTitle: "Minimax H3 Max 編集" -translationSourceHash: ba2b10f5 +translationSourceHash: fda273ce translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": bf838aaa "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 2757b0aa + "MiniMax H3 Max editing nodes": 4c3396f2 --- import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -91,6 +91,6 @@ MiniMax H3、H3 Max、H3 Max Turboでの生成については、[MiniMax H3 API | **MiniMax H3 Max Lip Sync** | 静止画像をアニメーション化し、口元がオーディオトラックに合わせて動くようにします | 画像1枚、5-899秒のオーディオ | | **MiniMax H3 Max Camera Control** | 静止画像内の被写体の周りをカメラが周回します | 画像1枚、および終了角度と距離 | -プロンプトでは、接続した参照を `@Image1`、`@Video1`、`@Audio1` として参照します。番号は接続順に付けられます。Recast は `@Image` タグを、Extend Video は `@Audio` タグを、Insert Video は `@Image` と `@Video` のタグを、Reference to Video は3種類すべてのタグを検査します。検査対象のタグが、その種類で接続した参照の範囲を超えると、エラーで実行が停止します。それ以外の種類のタグはそのまま渡されます。 +Recast、Extend Video、Insert Videoのプロンプトでは、接続した参照を `@Image1`、`@Video1`、`@Audio1` として参照します。番号は接続順に付けられます。Recast は `@Image` タグを、Extend Video は `@Audio` タグを、Insert Video は `@Image` と `@Video` のタグを検査します。検査対象のタグが、その種類で接続した参照の範囲を超えると、エラーで実行が停止します。それ以外の種類のタグはそのまま渡されます。[Reference to Video](/ja/tutorials/partner-nodes/minimax/minimax-h3#minimax-h3-参照からビデオへ) 生成ノードは3種類すべてのタグを解決します。 ノードおよび解像度ごとの料金は [Minimax料金表](/ja/tutorials/partner-nodes/pricing) に記載されています。 diff --git a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index e1748de8cf..c103ff434a 100644 --- a/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/ko/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,13 +2,13 @@ title: "MiniMax H3 Max: ComfyUI 영상 편집" description: "ComfyUI의 MiniMax H3 Max 편집 API 노드로 완성된 클립에서 출연진을 교체하고, 구도를 재설정하고, 장면을 연장하거나 삽입하고, 립 싱크를 맞추고, 카메라를 회전시킬 수 있습니다." sidebarTitle: "Minimax H3 Max 편집" -translationSourceHash: ba2b10f5 +translationSourceHash: fda273ce translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": bf838aaa "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 2757b0aa + "MiniMax H3 Max editing nodes": 4c3396f2 --- import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -91,6 +91,6 @@ MiniMax H3, H3 Max, H3 Max Turbo로 생성하려면 [MiniMax H3 API 비디오 | **MiniMax H3 Max Lip Sync** | 정지 이미지를 애니메이션으로 만들어 입 모양이 오디오 트랙을 따라가게 합니다 | 이미지 1장, 5-899초 오디오 | | **MiniMax H3 Max Camera Control** | 정지 이미지 속 피사체 주위로 카메라를 회전시킵니다 | 이미지 1장과 종료 각도 및 거리 | -프롬프트에서는 연결된 레퍼런스를 연결 순서대로 번호가 매겨진 `@Image1`, `@Video1`, `@Audio1`로 참조하세요. Recast는 `@Image` 태그를, Extend Video는 `@Audio` 태그를, Insert Video는 `@Image`와 `@Video` 태그를, Reference to Video는 세 가지 태그를 모두 검사합니다. 검사 대상 태그가 해당 종류로 연결한 레퍼런스 범위를 벗어나면 실행이 오류와 함께 중단됩니다. 다른 종류의 태그는 그대로 전달됩니다. +Recast, Extend Video, Insert Video 프롬프트에서는 연결된 레퍼런스를 연결 순서대로 번호가 매겨진 `@Image1`, `@Video1`, `@Audio1`로 참조하세요. Recast는 `@Image` 태그를, Extend Video는 `@Audio` 태그를, Insert Video는 `@Image`와 `@Video` 태그를 검사합니다. 검사 대상 태그가 해당 종류로 연결한 레퍼런스 범위를 벗어나면 실행이 오류와 함께 중단되고, 다른 종류의 태그는 그대로 전달됩니다. [Reference to Video](/ko/tutorials/partner-nodes/minimax/minimax-h3#minimax-h3-레퍼런스-기반-비디오-생성) 생성 노드는 세 가지 태그를 모두 해석합니다. 노드 및 해상도별 요금은 [Minimax 가격표](/ko/tutorials/partner-nodes/pricing)에 나와 있습니다. diff --git a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index ed339a16c2..5bd8385cb5 100644 --- a/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -84,6 +84,6 @@ The bundled demo turns a vertical portrait clip into a 16:9 frame. Load your own | **MiniMax H3 Max Lip Sync** | Animates a still image so the mouth follows an audio track | One image, 5-899 seconds of audio | | **MiniMax H3 Max Camera Control** | Orbits the camera around the subject in a still image | One image, plus an end angle and distance | -In prompts, refer to connected references as `@Image1`, `@Video1` or `@Audio1`, numbered in connection order. Recast checks `@Image` tags, Extend Video checks `@Audio` tags, Insert Video checks `@Image` and `@Video` tags, and Reference to Video checks all three. A tag of a checked kind that points past the connected references of that kind stops the run with an error; a tag of another kind is passed through unchanged. +In the Recast, Extend Video and Insert Video prompts, refer to connected references as `@Image1`, `@Video1` or `@Audio1`, numbered in connection order. Recast checks `@Image` tags, Extend Video `@Audio` tags, and Insert Video `@Image` and `@Video` tags; a tag of a checked kind that points past the connected references of that kind stops the run with an error, and other kinds pass through unchanged. The [Reference to Video](/tutorials/partner-nodes/minimax/minimax-h3#minimax-h3-reference-to-video) generation node resolves all three. Rates per node and resolution are in the [Minimax pricing table](/tutorials/partner-nodes/pricing). diff --git a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx index 152a005f8d..7c7dcb4ed3 100644 --- a/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx +++ b/zh/tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx @@ -2,13 +2,13 @@ title: "MiniMax H3 Max:ComfyUI 编辑" description: "使用 ComfyUI 中的 MiniMax H3 Max 编辑 API 节点,在已完成片段中替换角色、重新构图、延长或插入镜头、对口型,并环绕相机。" sidebarTitle: "MiniMax H3 Max 编辑" -translationSourceHash: ba2b10f5 +translationSourceHash: fda273ce translationFrom: tutorials/partner-nodes/minimax/minimax-h3-max-editing.mdx translationBlockHashes: "_intro": bf838aaa "MiniMax H3 Max: Cast Swap": 226f3eef "MiniMax H3 Max: Aspect Ratio Reframe": 4b23a419 - "MiniMax H3 Max editing nodes": 2757b0aa + "MiniMax H3 Max editing nodes": 4c3396f2 --- import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -91,6 +91,6 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; | **MiniMax H3 Max Lip Sync** | 让静态图像动起来,使嘴型跟随音频轨道 | 一张图像,5-899 秒音频 | | **MiniMax H3 Max Camera Control** | 让相机围绕静态图像中的主体环绕运动 | 一张图像,外加结束角度和距离 | -在提示词中,请以 `@Image1`、`@Video1` 或 `@Audio1` 指代已连接的参考,编号遵循连接顺序。Recast 检查 `@Image` 标签,Extend Video 检查 `@Audio` 标签,Insert Video 检查 `@Image` 和 `@Video` 标签,Reference to Video 则检查这三种标签。如果某个受检查类型的标签指向的参考超出该类型已连接的参考数量,运行将因错误而停止;其他类型的标签会原样传递。 +在 Recast、Extend Video 和 Insert Video 的提示词中,请以 `@Image1`、`@Video1` 或 `@Audio1` 指代已连接的参考,编号遵循连接顺序。Recast 检查 `@Image` 标签,Extend Video 检查 `@Audio` 标签,Insert Video 检查 `@Image` 和 `@Video` 标签;受检查类型的标签若指向超出该类型已连接的参考数量,运行将因错误而停止,其他类型则原样传递。[Reference to Video](/zh/tutorials/partner-nodes/minimax/minimax-h3#minimax-h3-参考转视频) 生成节点会解析这三种标签。 各节点和各分辨率的费率见 [Minimax 价格表](/zh/tutorials/partner-nodes/pricing)。