Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 31 additions & 0 deletions changelog/index.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,37 @@ description: "ComfyUI changelog: the latest features, improvements, and bug fixe
icon: "clock-rotate-left"
---

<Update label="v0.45.0" description="November 10, 2026">

**Partner Node Updates**
* [**Pika H3 Flex**](https://github.com/Comfy-Org/ComfyUI/pull/17509): Pika's tuned MiniMax H3 as a video node and a reference to video node
* [**Pika audio nodes**](https://github.com/Comfy-Org/ComfyUI/pull/17328): Pika Soundtrack writes a synchronized track for a clip, Pika SFX makes a single sound effect, and Pika Music writes a song from a description with optional lyrics
* [**Kling 4.0 Flash**](https://github.com/Comfy-Org/ComfyUI/pull/17536): A faster Kling 4.0 model on the Kling 4.0 nodes, plus an option to generate the original audio
* [**Breeze TTS 2**](https://github.com/Comfy-Org/ComfyUI/pull/17611): Text to Speech, Voice Design, Voice Clone and Voice Selector nodes
* [**Seedream 5.5**](https://github.com/Comfy-Org/ComfyUI/pull/17631): A new model option on the ByteDance Seedream node and on Layer Separation
* [**Seedance 2.5 4K**](https://github.com/Comfy-Org/ComfyUI/pull/17579): A 4k output resolution, available to enterprise accounts
* [**Gemini Omni Flash 1.2**](https://github.com/Comfy-Org/ComfyUI/pull/17560): A new model option on the Gemini Omni video node, with the clip length set in the prompt up to 15 seconds
* [**GPT Image 2.6 GIF and SVG**](https://github.com/Comfy-Org/ComfyUI/pull/17591): Generate an animated GIF, returned as frames plus a frame rate, or an SVG from a prompt
* [**Tencent HY**](https://github.com/Comfy-Org/ComfyUI/pull/17534): Text generation with Tencent's Hy4 model, from a prompt and optional images
* [**Qwen Image on Comfy Router**](https://github.com/Comfy-Org/ComfyUI/pull/17527): The Qwen Image nodes now run through Comfy Router

**Open-Source Model Support**
* [**MiniMax-H3-Image**](https://github.com/Comfy-Org/ComfyUI/pull/17655): MiniMax's open image model, with text to image and editing in one checkpoint at about 16 MP
* [**RelayVSR**](https://github.com/Comfy-Org/ComfyUI/pull/17511): Video upscaling that runs the model on every 15th frame and restores the remaining frames in the VSR decoder
* [**WeVisOne-VL**](https://github.com/Comfy-Org/ComfyUI/pull/17632): Grounding detection, segmentation, points and counting from a WeVisOne-VL text encoder
* [**UltraTex**](https://github.com/Comfy-Org/ComfyUI/pull/17636): Paint a reference image's look onto a mesh as a texture, rendered as one image per view
* [**Qwen 3.5 video input**](https://github.com/Comfy-Org/ComfyUI/pull/17582): The Qwen 3.5 text encoder accepts video, so text generation answers from the clip instead of the prompt alone
* [**Solid mode for Extract Mesh from Splat**](https://github.com/Comfy-Org/ComfyUI/pull/17497): Fill behind the surface contour to get a closed mesh instead of an open shell

**Improvements**
* [**SeedVR2 VAE slicing**](https://github.com/Comfy-Org/ComfyUI/pull/17508): The slice length follows free memory, so long videos fit on smaller cards
* [**Greedy decoding penalties**](https://github.com/Comfy-Org/ComfyUI/pull/17635): Repetition and presence penalty now apply to greedy sampling
* [**Resolution Selector precision**](https://github.com/Comfy-Org/ComfyUI/pull/17555): Megapixel targets land on the requested size
* [**Get Video Components memory**](https://github.com/Comfy-Org/ComfyUI/pull/17562): Decoded frames are held once instead of twice
* [**Local web sign-in by default**](https://github.com/Comfy-Org/ComfyUI/pull/17540): `local_web_sso` and `partner_node_token` are enabled by default

</Update>

<Update label="v0.44.0" description="November 4, 2026">

**Partner Node Updates**
Expand Down
34 changes: 33 additions & 1 deletion es/changelog/index.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -2,9 +2,10 @@
title: "Registro de cambios"
description: "Historial de ComfyUI: novedades, mejoras y correcciones de cada versión, actualizado a medida que se publican."
icon: "clock-rotate-left"
translationSourceHash: 2889b328
translationSourceHash: bee98d7b
translationFrom: changelog/index.mdx
translationBlockHashes:
"v0.45.0": bcbd3a93
"v0.44.0": 0673a7fe
"v0.43.1": 4598e422
"v0.43.0": ea595fe3
Expand Down Expand Up @@ -152,6 +153,37 @@ translationBlockHashes:
"v0.3.40": 24608eaf
---

<Update label="v0.45.0" description="10 de noviembre de 2026">

**Actualizaciones de nodos de socios**
* [**Pika H3 Flex**](https://github.com/Comfy-Org/ComfyUI/pull/17509): el MiniMax H3 ajustado de Pika como nodo de video y como nodo de referencia a video
* [**Nodos de audio de Pika**](https://github.com/Comfy-Org/ComfyUI/pull/17328): Pika Soundtrack escribe una pista sincronizada para un clip, Pika SFX crea un solo efecto de sonido y Pika Music compone una canción a partir de una descripción con letras opcionales
* [**Kling 4.0 Flash**](https://github.com/Comfy-Org/ComfyUI/pull/17536): un modelo Kling 4.0 más rápido en los nodos Kling 4.0, además de una opción para generar el audio original
* [**Breeze TTS 2**](https://github.com/Comfy-Org/ComfyUI/pull/17611): nodos Text to Speech, Voice Design, Voice Clone y Voice Selector
* [**Seedream 5.5**](https://github.com/Comfy-Org/ComfyUI/pull/17631): una nueva opción de modelo en el nodo ByteDance Seedream y en Layer Separation
* [**Seedance 2.5 4K**](https://github.com/Comfy-Org/ComfyUI/pull/17579): una resolución de salida 4k, disponible para cuentas de empresa
* [**Gemini Omni Flash 1.2**](https://github.com/Comfy-Org/ComfyUI/pull/17560): una nueva opción de modelo en el nodo de video Gemini Omni, con la duración del clip definida en el prompt hasta 15 segundos
* [**GPT Image 2.6 GIF y SVG**](https://github.com/Comfy-Org/ComfyUI/pull/17591): genera un GIF animado, devuelto como cuadros más una frecuencia de cuadros, o un SVG a partir de un prompt
* [**Tencent HY**](https://github.com/Comfy-Org/ComfyUI/pull/17534): generación de texto con el modelo Hy4 de Tencent, a partir de un prompt e imágenes opcionales
* [**Qwen Image en Comfy Router**](https://github.com/Comfy-Org/ComfyUI/pull/17527): los nodos Qwen Image ahora se ejecutan a través de Comfy Router

**Soporte de modelos de código abierto**
* [**MiniMax-H3-Image**](https://github.com/Comfy-Org/ComfyUI/pull/17655): el modelo de imagen abierto de MiniMax, con texto a imagen y edición en un solo checkpoint de unos 16 MP
* [**RelayVSR**](https://github.com/Comfy-Org/ComfyUI/pull/17511): ampliación de video que ejecuta el modelo en cada 15.º cuadro y restaura los cuadros restantes en el decodificador VSR
* [**WeVisOne-VL**](https://github.com/Comfy-Org/ComfyUI/pull/17632): detección de grounding, segmentación, puntos y conteo a partir de un codificador de texto WeVisOne-VL
* [**UltraTex**](https://github.com/Comfy-Org/ComfyUI/pull/17636): pinta el aspecto de una imagen de referencia sobre una malla como textura, renderizado como una imagen por vista
* [**Entrada de video en Qwen 3.5**](https://github.com/Comfy-Org/ComfyUI/pull/17582): el codificador de texto Qwen 3.5 acepta video, así que la generación de texto responde a partir del clip y no solo del prompt
* [**Modo Solid para Extraer malla de Splat**](https://github.com/Comfy-Org/ComfyUI/pull/17497): rellena detrás del contorno de la superficie para obtener una malla cerrada en lugar de una carcasa abierta

**Mejoras**
* [**Segmentación de VAE en SeedVR2**](https://github.com/Comfy-Org/ComfyUI/pull/17508): la longitud del corte sigue la memoria libre, así los videos largos caben en tarjetas más pequeñas
* [**Penalizaciones de decodificación greedy**](https://github.com/Comfy-Org/ComfyUI/pull/17635): las penalizaciones por repetición y por presencia ahora se aplican al muestreo greedy
* [**Precisión del Selector de resolución**](https://github.com/Comfy-Org/ComfyUI/pull/17555): los objetivos en megapíxeles llegan al tamaño solicitado
* [**Memoria en Obtener componentes de video**](https://github.com/Comfy-Org/ComfyUI/pull/17562): los cuadros decodificados se guardan una vez en lugar de dos
* [**Inicio de sesión web local de forma predeterminada**](https://github.com/Comfy-Org/ComfyUI/pull/17540): `local_web_sso` y `partner_node_token` están habilitados de forma predeterminada

</Update>

<Update label="v0.44.0" description="4 de noviembre de 2026">

**Actualizaciones de nodos de socios**
Expand Down
34 changes: 33 additions & 1 deletion ja/changelog/index.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -2,9 +2,10 @@
title: "変更履歴"
description: "ComfyUI の変更ログ:各リリースの最新機能、改善点、バグ修正を、新バージョンのリリースに合わせて更新。"
icon: "clock-rotate-left"
translationSourceHash: 2889b328
translationSourceHash: bee98d7b
translationFrom: changelog/index.mdx
translationBlockHashes:
"v0.45.0": bcbd3a93
"v0.44.0": 0673a7fe
"v0.43.1": 4598e422
"v0.43.0": ea595fe3
Expand Down Expand Up @@ -152,6 +153,37 @@ translationBlockHashes:
"v0.3.40": 24608eaf
---

<Update label="v0.45.0" description="2026年11月10日">

**パートナーノードの更新**
* [**Pika H3 Flex**](https://github.com/Comfy-Org/ComfyUI/pull/17509): Pikaが調整したMiniMax H3をビデオノードとして、また参照からビデオへのノードとして提供
* [**Pikaオーディオノード**](https://github.com/Comfy-Org/ComfyUI/pull/17328): Pika Soundtrackはクリップに同期したトラックを書き出し、Pika SFXは単一のサウンドエフェクトを生成し、Pika Musicは説明から任意の歌詞付きで楽曲を書き出します
* [**Kling 4.0 Flash**](https://github.com/Comfy-Org/ComfyUI/pull/17536): Kling 4.0ノードでより高速なKling 4.0モデルを利用可能。オリジナルのオーディオを生成するオプションも追加
* [**Breeze TTS 2**](https://github.com/Comfy-Org/ComfyUI/pull/17611): Text to Speech、Voice Design、Voice Clone、Voice Selectorノード
* [**Seedream 5.5**](https://github.com/Comfy-Org/ComfyUI/pull/17631): ByteDance SeedreamノードとLayer Separationに新しいモデルオプションを追加
* [**Seedance 2.5 4K**](https://github.com/Comfy-Org/ComfyUI/pull/17579): エンタープライズアカウントで利用できる4K出力解像度
* [**Gemini Omni Flash 1.2**](https://github.com/Comfy-Org/ComfyUI/pull/17560): Gemini Omniビデオノードに新しいモデルオプションを追加。クリップの長さはプロンプトで最大15秒まで設定可能
* [**GPT Image 2.6 GIFとSVG**](https://github.com/Comfy-Org/ComfyUI/pull/17591): アニメーションGIF(フレームとフレームレートとして返される)またはSVGをプロンプトから生成
* [**Tencent HY**](https://github.com/Comfy-Org/ComfyUI/pull/17534): TencentのHy4モデルによるテキスト生成。プロンプトと任意の画像から生成
* [**Comfy Router上のQwen Image**](https://github.com/Comfy-Org/ComfyUI/pull/17527): Qwen ImageノードがComfy Router経由で実行されるようになりました

**オープンソースモデルのサポート**
* [**MiniMax-H3-Image**](https://github.com/Comfy-Org/ComfyUI/pull/17655): MiniMaxのオープン画像モデル。テキストから画像と編集を1つのcheckpointで約16MPで実行
* [**RelayVSR**](https://github.com/Comfy-Org/ComfyUI/pull/17511): 15フレームごとにモデルを実行し、残りのフレームをVSRデコーダーで復元するビデオアップスケーリング
* [**WeVisOne-VL**](https://github.com/Comfy-Org/ComfyUI/pull/17632): WeVisOne-VLテキストエンコーダーによるグラウンディング検出、セグメンテーション、ポイント、カウント
* [**UltraTex**](https://github.com/Comfy-Org/ComfyUI/pull/17636): 参照画像のルックをメッシュにテクスチャとしてペイントし、ビューごとに1枚の画像としてレンダリング
* [**Qwen 3.5のビデオ入力**](https://github.com/Comfy-Org/ComfyUI/pull/17582): Qwen 3.5のテキストエンコーダーがビデオを受け付けるため、テキスト生成はプロンプトだけでなくクリップの内容にも基づいて回答します
* [**スプラットからメッシュを抽出: ソリッドモード**](https://github.com/Comfy-Org/ComfyUI/pull/17497): 表面輪郭の背後を埋めて、開いたシェルではなく閉じたメッシュを取得

**改善**
* [**SeedVR2 VAEスライス**](https://github.com/Comfy-Org/ComfyUI/pull/17508): スライス長が空きメモリに応じて決まるため、長いビデオもより小さなカードに収まります
* [**Greedyデコーディングのペナルティ**](https://github.com/Comfy-Org/ComfyUI/pull/17635): 繰り返しペナルティとプレゼンスペナルティがgreedyサンプリングにも適用されるようになりました
* [**解像度セレクターの精度**](https://github.com/Comfy-Org/ComfyUI/pull/17555): メガピクセル指定が要求したサイズに一致
* [**ビデオコンポーネントの取得: メモリ**](https://github.com/Comfy-Org/ComfyUI/pull/17562): デコードされたフレームが2回ではなく1回だけ保持される
* [**ローカルWebサインインをデフォルトに**](https://github.com/Comfy-Org/ComfyUI/pull/17540): `local_web_sso` と `partner_node_token` がデフォルトで有効になりました

</Update>

<Update label="v0.44.0" description="2026年11月4日">

**パートナーノードの更新**
Expand Down
34 changes: 33 additions & 1 deletion ko/changelog/index.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -2,9 +2,10 @@
title: "변경 로그"
description: "ComfyUI 변경 로그: 각 릴리스의 최신 기능, 개선 사항, 버그 수정을 새 버전 출시에 맞춰 업데이트합니다."
icon: "clock-rotate-left"
translationSourceHash: 2889b328
translationSourceHash: bee98d7b
translationFrom: changelog/index.mdx
translationBlockHashes:
"v0.45.0": bcbd3a93
"v0.44.0": 0673a7fe
"v0.43.1": 4598e422
"v0.43.0": ea595fe3
Expand Down Expand Up @@ -152,6 +153,37 @@ translationBlockHashes:
"v0.3.40": 24608eaf
---

<Update label="v0.45.0" description="2026년 11월 10일">

**파트너 노드 업데이트**
* [**Pika H3 Flex**](https://github.com/Comfy-Org/ComfyUI/pull/17509): Pika가 튜닝한 MiniMax H3를 비디오 노드와 레퍼런스 기반 비디오 생성 노드로 제공
* [**Pika 오디오 노드**](https://github.com/Comfy-Org/ComfyUI/pull/17328): Pika Soundtrack은 클립에 동기화된 트랙을 작성하고, Pika SFX는 단일 사운드 효과를 만들며, Pika Music은 선택적 가사와 함께 설명으로부터 곡을 작성
* [**Kling 4.0 Flash**](https://github.com/Comfy-Org/ComfyUI/pull/17536): Kling 4.0 노드에서 더 빠른 Kling 4.0 모델을 사용할 수 있으며, 원본 오디오를 생성하는 옵션도 추가
* [**Breeze TTS 2**](https://github.com/Comfy-Org/ComfyUI/pull/17611): Text to Speech, Voice Design, Voice Clone, Voice Selector 노드
* [**Seedream 5.5**](https://github.com/Comfy-Org/ComfyUI/pull/17631): ByteDance Seedream 노드와 Layer Separation에 추가된 새 모델 옵션
* [**Seedance 2.5 4K**](https://github.com/Comfy-Org/ComfyUI/pull/17579): 4k 출력 해상도, 엔터프라이즈 계정에서 사용 가능
* [**Gemini Omni Flash 1.2**](https://github.com/Comfy-Org/ComfyUI/pull/17560): Gemini Omni 비디오 노드의 새 모델 옵션으로, 최대 15초까지 프롬프트에서 클립 길이를 설정
* [**GPT Image 2.6 GIF 및 SVG**](https://github.com/Comfy-Org/ComfyUI/pull/17591): 프레임과 프레임 레이트로 반환되는 애니메이션 GIF 또는 프롬프트로부터 SVG를 생성
* [**Tencent HY**](https://github.com/Comfy-Org/ComfyUI/pull/17534): 프롬프트와 선택적 이미지로부터 Tencent의 Hy4 모델을 사용한 텍스트 생성
* [**Comfy Router의 Qwen Image**](https://github.com/Comfy-Org/ComfyUI/pull/17527): Qwen 이미지 노드가 이제 Comfy Router를 통해 실행

**오픈소스 모델 지원**
* [**MiniMax-H3-Image**](https://github.com/Comfy-Org/ComfyUI/pull/17655): MiniMax의 오픈 이미지 모델로, 하나의 checkpoint에서 약 16 MP로 텍스트 기반 이미지 생성과 편집 지원
* [**RelayVSR**](https://github.com/Comfy-Org/ComfyUI/pull/17511): 15프레임마다 모델을 실행하고 나머지 프레임을 VSR 디코더에서 복원하는 비디오 업스케일링
* [**WeVisOne-VL**](https://github.com/Comfy-Org/ComfyUI/pull/17632): WeVisOne-VL 텍스트 인코더로부터의 그라운딩 감지, 세그멘테이션, 포인트 및 카운팅
* [**UltraTex**](https://github.com/Comfy-Org/ComfyUI/pull/17636): 레퍼런스 이미지의 모습을 메시에 텍스처로 칠하고, 뷰마다 하나의 이미지로 렌더링
* [**Qwen 3.5 비디오 입력**](https://github.com/Comfy-Org/ComfyUI/pull/17582): Qwen 3.5 텍스트 인코더가 비디오를 받으므로 텍스트 생성이 프롬프트만이 아니라 클립 내용을 바탕으로 답변합니다
* [**Splat에서 메시 추출의 Solid 모드**](https://github.com/Comfy-Org/ComfyUI/pull/17497): 표면 윤곽 뒤를 채워 열린 셸 대신 닫힌 메시를 얻음

**개선 사항**
* [**SeedVR2 VAE 슬라이싱**](https://github.com/Comfy-Org/ComfyUI/pull/17508): 슬라이스 길이가 여유 메모리를 따르므로 긴 비디오도 더 작은 카드에서 처리 가능
* [**그리디 디코딩 페널티**](https://github.com/Comfy-Org/ComfyUI/pull/17635): 반복 및 존재 페널티가 이제 그리디 샘플링에도 적용
* [**해상도 선택기 정밀도**](https://github.com/Comfy-Org/ComfyUI/pull/17555): 메가픽셀 목표가 요청한 크기에 맞게 설정
* [**비디오 컴포넌트 추출 메모리**](https://github.com/Comfy-Org/ComfyUI/pull/17562): 디코딩된 프레임이 두 번이 아닌 한 번만 유지
* [**기본적으로 로컬 웹 로그인**](https://github.com/Comfy-Org/ComfyUI/pull/17540): `local_web_sso` 및 `partner_node_token`이 기본적으로 활성화

</Update>

<Update label="v0.44.0" description="2026년 11월 4일">

**파트너 노드 업데이트**
Expand Down
Loading
Loading