From 87270d690bc8656d4225bd5c419c741d233b9f5c Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Fri, 13 Nov 2026 13:17:49 +0800 Subject: [PATCH 1/3] docs(pika): add the Pika Music and Pika SFX page --- docs.json | 15 ++++-- .../partner-nodes/pika/pika-audio.mdx | 52 +++++++++++++++++++ .../partner-nodes/pika/pika-audio.mdx | 52 +++++++++++++++++++ .../partner-nodes/pika/pika-audio.mdx | 52 +++++++++++++++++++ tutorials/partner-nodes/pika/pika-audio.mdx | 50 ++++++++++++++++++ .../partner-nodes/pika/pika-audio.mdx | 52 +++++++++++++++++++ 6 files changed, 268 insertions(+), 5 deletions(-) create mode 100644 es/tutorials/partner-nodes/pika/pika-audio.mdx create mode 100644 ja/tutorials/partner-nodes/pika/pika-audio.mdx create mode 100644 ko/tutorials/partner-nodes/pika/pika-audio.mdx create mode 100644 tutorials/partner-nodes/pika/pika-audio.mdx create mode 100644 zh/tutorials/partner-nodes/pika/pika-audio.mdx diff --git a/docs.json b/docs.json index 1faf918fcb..6760b3d52c 100644 --- a/docs.json +++ b/docs.json @@ -841,7 +841,8 @@ { "group": "Pika", "pages": [ - "tutorials/partner-nodes/pika/pika-h3-flex" + "tutorials/partner-nodes/pika/pika-h3-flex", + "tutorials/partner-nodes/pika/pika-audio" ] }, { @@ -5075,7 +5076,8 @@ { "group": "Pika", "pages": [ - "zh/tutorials/partner-nodes/pika/pika-h3-flex" + "zh/tutorials/partner-nodes/pika/pika-h3-flex", + "zh/tutorials/partner-nodes/pika/pika-audio" ] }, { @@ -9260,7 +9262,8 @@ { "group": "Pika", "pages": [ - "ja/tutorials/partner-nodes/pika/pika-h3-flex" + "ja/tutorials/partner-nodes/pika/pika-h3-flex", + "ja/tutorials/partner-nodes/pika/pika-audio" ] }, { @@ -13462,7 +13465,8 @@ { "group": "Pika", "pages": [ - "ko/tutorials/partner-nodes/pika/pika-h3-flex" + "ko/tutorials/partner-nodes/pika/pika-h3-flex", + "ko/tutorials/partner-nodes/pika/pika-audio" ] }, { @@ -17602,7 +17606,8 @@ { "group": "Pika", "pages": [ - "es/tutorials/partner-nodes/pika/pika-h3-flex" + "es/tutorials/partner-nodes/pika/pika-h3-flex", + "es/tutorials/partner-nodes/pika/pika-audio" ] }, { diff --git a/es/tutorials/partner-nodes/pika/pika-audio.mdx b/es/tutorials/partner-nodes/pika/pika-audio.mdx new file mode 100644 index 0000000000..2a4f24531c --- /dev/null +++ b/es/tutorials/partner-nodes/pika/pika-audio.mdx @@ -0,0 +1,52 @@ +--- +title: "Pika Music y Pika SFX en ComfyUI: canciones y efectos de sonido" +description: "Genera una canción desde un prompt de estilo y letras opcionales, versiona una pista o crea efectos de sonido de 1 a 20 segundos con los nodos de audio de Pika." +sidebarTitle: "Pika Music y SFX" +translationSourceHash: 60f1ba51 +translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx +--- + +import ReqHint from "/snippets/es/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/es/tutorials/update-reminder.mdx"; + +Los nodos de audio de Pika convierten una descripción de texto en audio finalizado sin necesidad de una GPU local. **Pika Music** crea una canción a partir de un prompt de estilo y letras opcionales, con una duración que eliges entre 10 y 360 segundos, y también puede versionar una canción subida con la duración propia de esa canción. **Pika SFX** devuelve un único efecto de sonido de 1 a 20 segundos, que encaja en foley, blips de UI y golpes de tráiler. + +Ninguno de los dos nodos descarga pesos. Ambos se ejecutan en tu cuenta de Comfy y se cobran en créditos como los demás nodos asociados. + + + + +## Pika Music: texto a música {#api_pika_music} + +Escribe los arreglos, el ambiente y la instrumentación en el prompt. Añade letras con etiquetas de sección como `[Verse]` y `[Chorus]` para una pista con voz, o deja el campo de letras vacío para una instrumental. + +Pika Music text to music workflow preview + + + + Descarga el JSON o busca "Pika Music" en la Biblioteca de plantillas + + + +1. Elige `text to music` en el nodo `Pika Music` y describe la pista: género, tempo, instrumentos y ambiente. +2. Establece `duration` entre 10 y 360 segundos. Activa `rewrite_prompt` si solo tienes una idea breve y quieres que el modelo la expanda primero. +3. Ejecuta el flujo de trabajo. `Save Audio (Advanced)` escribe un MP3 en `ComfyUI/output/audio/`. + +Cambia el modo de ese nodo a `cover` para cambiar el estilo de una canción existente. Conecta un archivo de audio a `source_audio`, describe el nuevo estilo y el resultado conserva la duración de la fuente, ajustada al rango de 5 a 360 segundos. + +## Pika SFX: texto a efectos de sonido {#api_pika_sfx} + +Nombra el material, la acción, el espacio y la intensidad en una sola línea. Una descripción cercana y seca se lee mejor que una lista de adjetivos. + +Pika SFX text to sound effect workflow preview + + + + Descarga el JSON o busca "Pika SFX" en la Biblioteca de plantillas + + + +1. Escribe el sonido en el prompt de `Pika SFX`, por ejemplo: una vieja puerta de madera que se abre chirriando sobre bisagras oxidadas y luego un pestillo pesado que se cierra de golpe. +2. Establece `duration` entre 1 y 20 segundos y ejecuta el flujo de trabajo. El MP3 queda en `ComfyUI/output/audio/`. + +Cada ejecución produce un único efecto de sonido mono. Para una secuencia de sonidos, ejecuta el nodo una vez por cada sonido y colócalos tú mismo en la línea de tiempo. \ No newline at end of file diff --git a/ja/tutorials/partner-nodes/pika/pika-audio.mdx b/ja/tutorials/partner-nodes/pika/pika-audio.mdx new file mode 100644 index 0000000000..d27061a731 --- /dev/null +++ b/ja/tutorials/partner-nodes/pika/pika-audio.mdx @@ -0,0 +1,52 @@ +--- +title: "ComfyUI の Pika Music と Pika SFX: 楽曲とサウンドエフェクト" +description: "スタイルプロンプトと任意の歌詞から楽曲全体を生成するか、既存のトラックをカバーするか、Pika オーディオノードで 1〜20 秒のサウンドエフェクトを作成します。" +sidebarTitle: "Pika Music & SFX" +translationSourceHash: 60f1ba51 +translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx +--- + +import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; + +Pika のオーディオノードは、ローカル GPU なしでテキストの説明から完成したオーディオを作り出します。**Pika Music** はスタイルプロンプトと任意の歌詞から、10〜360 秒の範囲で選んだ長さの楽曲を書き出し、アップロードした楽曲をその曲本来の長さでカバーすることもできます。**Pika SFX** は 1〜20 秒の単一のサウンドエフェクトを返し、フォーリー、UI の効果音、トレーラーのヒット音に適しています。 + +どちらのノードも重みをダウンロードしません。どちらも Comfy アカウント上で実行され、他のパートナーノードと同様にクレジットで課金されます。 + + + + +## Pika Music: テキストから楽曲へ {#api_pika_music} + +アレンジ、雰囲気、楽器編成をプロンプトに書き込みます。ボーカルトラックにしたい場合は `[Verse]` や `[Chorus]` などのタグを使ったセクション付きの歌詞を追加し、インストゥルメンタルにしたい場合は歌詞フィールドを空のままにします。 + +Pika Music text to music workflow preview + + + + JSON をダウンロードするか、テンプレートライブラリで「Pika Music」を検索してください + + + +1. `Pika Music` ノードで `text to music` を選択し、トラックを説明します。ジャンル、テンポ、楽器、雰囲気を指定します。 +2. `duration` を 10〜360 秒の範囲で設定します。短いアイデアしかなく、モデルにまずそれを展開してほしい場合は `rewrite_prompt` をオンにします。 +3. ワークフローを実行します。`Save Audio (Advanced)` が MP3 を `ComfyUI/output/audio/` に書き出します。 + +代わりに既存の楽曲をスタイル変更するには、同じノードのモードを `cover` に切り替えます。オーディオファイルを `source_audio` に接続し、新しいスタイルを説明すると、結果はソースの長さを保ち、5〜360 秒に制限されます。 + +## Pika SFX: テキストからサウンドエフェクトへ {#api_pika_sfx} + +マテリアル、動作、空間、強度を 1 行で指定します。形容詞を並べるよりも、近くで乾いた描写のほうがうまく伝わります。 + +Pika SFX text to sound effect workflow preview + + + + JSON をダウンロードするか、テンプレートライブラリで「Pika SFX」を検索してください + + + +1. `Pika SFX` のプロンプトにサウンドを書きます。たとえば、古い木の扉が錆びた蝶番で軋みながら開き、その後重い掛け金がカチッと閉まる様子などです。 +2. `duration` を 1〜20 秒の範囲で設定し、ワークフローを実行します。MP3 は `ComfyUI/output/audio/` に出力されます。 + +実行ごとに 1 つのモノラルのサウンドエフェクトが生成されます。一連のサウンドが必要な場合は、サウンドごとにノードを 1 回実行し、タイムライン上に自分で配置してください。 \ No newline at end of file diff --git a/ko/tutorials/partner-nodes/pika/pika-audio.mdx b/ko/tutorials/partner-nodes/pika/pika-audio.mdx new file mode 100644 index 0000000000..281069a664 --- /dev/null +++ b/ko/tutorials/partner-nodes/pika/pika-audio.mdx @@ -0,0 +1,52 @@ +--- +title: "ComfyUI의 Pika Music 및 Pika SFX: 노래와 사운드 효과" +description: "스타일 프롬프트와 선택적 가사로 전체 곡을 생성하거나, 기존 트랙을 커버하거나, ComfyUI의 Pika 오디오 노드로 1~20초 길이의 사운드 효과를 만들 수 있습니다." +sidebarTitle: "Pika Music 및 SFX" +translationSourceHash: 60f1ba51 +translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx +--- + +import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; + +Pika의 오디오 노드는 로컬 GPU 없이 텍스트 설명을 완성된 오디오로 바꿉니다. **Pika Music**은 스타일 프롬프트와 선택적 가사로 노래를 작성하며, 길이는 10~360초에서 선택할 수 있고, 업로드한 노래를 해당 노래의 원래 길이로 커버할 수도 있습니다. **Pika SFX**는 1~20초의 단일 사운드 효과를 반환하며, 폴리, UI 블립, 예고편 임팩트에 적합합니다. + +두 노드 모두 가중치를 다운로드하지 않습니다. 둘 다 Comfy 계정에서 실행되며 다른 파트너 노드와 마찬가지로 크레딧으로 청구됩니다. + + + + +## Pika Music: 텍스트에서 음악으로 {#api_pika_music} + +프롬프트에 편곡, 분위기, 악기 구성을 작성하세요. 보컬 트랙에는 `[Verse]`, `[Chorus]` 같은 태그로 섹션을 표시한 가사를 추가하고, 연주곡을 원하면 가사 필드를 비워 두세요. + +Pika Music 텍스트에서 음악으로 워크플로 미리보기 + + + + JSON을 다운로드하거나 템플릿 라이브러리에서 "Pika Music"을 검색하세요 + + + +1. `Pika Music` 노드에서 `text to music`을 선택하고 트랙을 설명하세요: 장르, 템포, 악기, 분위기. +2. `duration`을 10~360초로 설정하세요. 짧은 아이디어만 있고 모델이 먼저 이를 확장하길 원하면 `rewrite_prompt`를 켜세요. +3. 워크플로를 실행하세요. `Save Audio (Advanced)`가 MP3를 `ComfyUI/output/audio/`에 저장합니다. + +기존 노래의 스타일을 바꾸려면 같은 노드의 모드를 `cover`로 전환하세요. 오디오 파일을 `source_audio`에 연결하고 새 스타일을 설명하면, 결과는 소스 길이를 유지하며 5~360초로 제한됩니다. + +## Pika SFX: 텍스트에서 사운드 효과로 {#api_pika_sfx} + +재질, 동작, 공간, 강도를 한 줄로 지정하세요. 구체적이고 담백한 설명이 형용사 목록보다 더 잘 해석됩니다. + +Pika SFX 텍스트에서 사운드 효과로 워크플로 미리보기 + + + + JSON을 다운로드하거나 템플릿 라이브러리에서 "Pika SFX"를 검색하세요 + + + +1. `Pika SFX` 프롬프트에 사운드를 작성하세요. 예를 들어 녹슨 경첩이 삐걱이며 열리는 오래된 목재 문, 이어서 무거운 걸쇠가 철컥 닫히는 소리입니다. +2. `duration`을 1~20초로 설정한 다음 워크플로를 실행하세요. MP3는 `ComfyUI/output/audio/`에 저장됩니다. + +각 실행은 하나의 모노 사운드 효과를 생성합니다. 연속된 사운드가 필요하면 사운드마다 노드를 한 번씩 실행하고 타임라인에 직접 배치하세요. \ No newline at end of file diff --git a/tutorials/partner-nodes/pika/pika-audio.mdx b/tutorials/partner-nodes/pika/pika-audio.mdx new file mode 100644 index 0000000000..a3db3d2617 --- /dev/null +++ b/tutorials/partner-nodes/pika/pika-audio.mdx @@ -0,0 +1,50 @@ +--- +title: "Pika Music and Pika SFX in ComfyUI: songs and sound effects" +description: "Generate a full song from a style prompt plus optional lyrics, cover an existing track, or make a 1 to 20 second sound effect with the Pika audio nodes in ComfyUI." +sidebarTitle: "Pika Music & SFX" +--- + +import ReqHint from "/snippets/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; + +Pika's audio nodes turn a text description into finished audio without a local GPU. **Pika Music** writes a song from a style prompt and optional lyrics, in a length you choose from 10 to 360 seconds, and it can also cover an uploaded song at that song's own length. **Pika SFX** returns a single sound effect of 1 to 20 seconds, which fits foley, UI blips and trailer hits. + +Neither node downloads weights. Both run on your Comfy account and are billed in credits like the other partner nodes. + + + + +## Pika Music: Text to Music {#api_pika_music} + +Write the arrangement, mood and instrumentation into the prompt. Add section-tagged lyrics with tags such as `[Verse]` and `[Chorus]` for a vocal track, or leave the lyrics field empty for an instrumental. + +Pika Music text to music workflow preview + + + + Download JSON or search "Pika Music" in Template Library + + + +1. Pick `text to music` in the `Pika Music` node and describe the track: genre, tempo, instruments and mood. +2. Set `duration` between 10 and 360 seconds. Turn on `rewrite_prompt` if you only have a short idea and want the model to expand it first. +3. Run the workflow. `Save Audio (Advanced)` writes an MP3 to `ComfyUI/output/audio/`. + +Switch the same node's mode to `cover` to restyle an existing song instead. Connect an audio file to `source_audio`, describe the new style, and the result keeps the source length, clamped to 5 to 360 seconds. + +## Pika SFX: Text to Sound Effects {#api_pika_sfx} + +Name the material, the action, the space and the intensity in one line. A close, dry description reads better than a list of adjectives. + +Pika SFX text to sound effect workflow preview + + + + Download JSON or search "Pika SFX" in Template Library + + + +1. Write the sound in the `Pika SFX` prompt, for example an old timber door creaking open on rusty hinges, then a heavy latch clicking shut. +2. Set `duration` between 1 and 20 seconds, then run the workflow. The MP3 lands in `ComfyUI/output/audio/`. + +Each run produces one mono sound effect. For a sequence of sounds, run the node once per sound and place them on the timeline yourself. diff --git a/zh/tutorials/partner-nodes/pika/pika-audio.mdx b/zh/tutorials/partner-nodes/pika/pika-audio.mdx new file mode 100644 index 0000000000..42608cce33 --- /dev/null +++ b/zh/tutorials/partner-nodes/pika/pika-audio.mdx @@ -0,0 +1,52 @@ +--- +title: "ComfyUI 中的 Pika Music 与 Pika SFX:歌曲与音效" +description: "使用 ComfyUI 中的 Pika 音频节点,从风格提示词加上可选歌词生成一首完整歌曲、翻唱已有曲目,或制作 1 到 20 秒的音效。" +sidebarTitle: "Pika Music 与 SFX" +translationSourceHash: 60f1ba51 +translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx +--- + +import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; +import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; + +Pika 的音频节点无需本地 GPU 就能把文字描述变成成品音频。**Pika Music** 根据风格提示词和可选歌词创作歌曲,时长由你在 10 到 360 秒之间自行选择,它还能以歌曲自身的长度翻唱上传的歌曲。**Pika SFX** 返回一段 1 到 20 秒的单一音效,适用于拟音、UI 提示音和预告片冲击音。 + +两个节点都不下载权重。它们都运行在你的 Comfy 账户上,与其他合作节点一样按积分计费。 + + + + +## Pika Music:文本生成音乐 {#api_pika_music} + +把编曲、情绪和配器写进提示词。如果要有人声,就添加带分段标签的歌词,标签例如 `[Verse]` 和 `[Chorus]`;如果要纯音乐,就把歌词字段留空。 + +Pika Music 文本生成音乐工作流预览 + + + + 下载 JSON,或在模板库中搜索 "Pika Music" + + + +1. 在 `Pika Music` 节点中选择 `text to music`,并描述曲目:流派、速度、乐器和情绪。 +2. 将 `duration` 设为 10 到 360 秒之间。如果你只有一个简短的想法,希望模型先把它展开,请打开 `rewrite_prompt`。 +3. 运行工作流。`Save Audio (Advanced)` 会把 MP3 写入 `ComfyUI/output/audio/`。 + +把同一个节点的模式切换为 `cover`,即可改为重新演绎一首已有的歌曲。将音频文件连接到 `source_audio`,描述新的风格,结果会沿用来源长度,并限制在 5 到 360 秒之间。 + +## Pika SFX:文本生成音效 {#api_pika_sfx} + +用一句话写明材质、动作、空间和强度。贴近、干练的描述比堆砌形容词效果更好。 + +Pika SFX 文本生成音效工作流预览 + + + + 下载 JSON,或在模板库中搜索 "Pika SFX" + + + +1. 在 `Pika SFX` 的提示词中写出声音,例如:一扇老旧木门在生锈的合页上吱呀打开,随后一声沉重的门闩咔哒合上。 +2. 将 `duration` 设为 1 到 20 秒之间,然后运行工作流。MP3 会落在 `ComfyUI/output/audio/`。 + +每次运行生成一段单声道音效。如果需要一组连续的声音,请为每个声音各运行一次节点,并自行把它们排到时间线上。 \ No newline at end of file From daf2c8e4b57a68379bd26d8d0650dacb9a2c0773 Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Fri, 13 Nov 2026 18:32:24 +0800 Subject: [PATCH 2/3] docs(pika): expand the Pika Music and Pika SFX page The first pass described both nodes in a few paragraphs. This adds the material a reader and a search engine both need, in all five languages: - Node controls tables for text to music, cover and SFX, with the real ranges from nodes_pika.py (prompt 4,000, lyrics 8,000, duration 10-360 and 1-20, rewrite_prompt off by default, seed default 42) - A music prompt formula with four worked examples, plus what changes the result (arrangement arc, negatives, vocal delivery, length, iteration) - Lyrics and section tags, including an empty field meaning instrumental and breaks driving phrasing - A sound design section with six example prompts by goal, and rules for one event per run and matching duration to the event - Duration, limits and credits table, plus when seed stops reproducing (rewrite_prompt on) - Output and saving via Save Audio (Advanced), and how the AUDIO output composes with Create Video - Music vs SFX vs Soundtrack routing table, and related page links - Cross-link from the Pika H3 Flex page in every locale Fixes a factual error in the first pass: Pika SFX renders 44.1 kHz stereo, not mono, per Pika's model docs and blog. --- .../partner-nodes/pika/pika-audio.mdx | 195 ++++++++++++++++-- .../partner-nodes/pika/pika-h3-flex.mdx | 2 + .../partner-nodes/pika/pika-audio.mdx | 193 +++++++++++++++-- .../partner-nodes/pika/pika-h3-flex.mdx | 2 + .../partner-nodes/pika/pika-audio.mdx | 195 ++++++++++++++++-- .../partner-nodes/pika/pika-h3-flex.mdx | 2 + tutorials/partner-nodes/pika/pika-audio.mdx | 179 ++++++++++++++-- tutorials/partner-nodes/pika/pika-h3-flex.mdx | 2 + .../partner-nodes/pika/pika-audio.mdx | 193 +++++++++++++++-- .../partner-nodes/pika/pika-h3-flex.mdx | 2 + 10 files changed, 898 insertions(+), 67 deletions(-) diff --git a/es/tutorials/partner-nodes/pika/pika-audio.mdx b/es/tutorials/partner-nodes/pika/pika-audio.mdx index 2a4f24531c..35f757ea7c 100644 --- a/es/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/es/tutorials/partner-nodes/pika/pika-audio.mdx @@ -1,24 +1,51 @@ --- -title: "Pika Music y Pika SFX en ComfyUI: canciones y efectos de sonido" -description: "Genera una canción desde un prompt de estilo y letras opcionales, versiona una pista o crea efectos de sonido de 1 a 20 segundos con los nodos de audio de Pika." +title: "Pika Music y Pika SFX en ComfyUI: canciones y efectos de sonido con IA" +description: "Genera una canción completa desde un prompt de estilo y letras opcionales, versiona una pista o crea efectos de sonido de 1 a 20 segundos con los nodos Pika Music y Pika SFX de ComfyUI." sidebarTitle: "Pika Music y SFX" -translationSourceHash: 60f1ba51 +translationSourceHash: c04fcf5c translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx +translationBlockHashes: + "_intro": dc3cd241 + "Get started": fd80ca7d + "Pika Music: text to music {#api_pika_music}": ae1738bd + "Pika Music: cover a song {#api_pika_music_cover}": 0793be8c + "Pika SFX: text to sound effects {#api_pika_sfx}": 81088686 + "Node controls": 2c4dd7da + "Writing a music prompt": a24287f2 + "Lyrics and section tags": af8f7f9e + "Writing a sound effect prompt": 20f01804 + "Duration, limits and credits": 763b11c7 + "Output and saving": 4abf0bf7 + "Which Pika audio node to use": 8b598461 + "Related pages": 96e4c1fe --- import ReqHint from "/snippets/es/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/es/tutorials/update-reminder.mdx"; -Los nodos de audio de Pika convierten una descripción de texto en audio finalizado sin necesidad de una GPU local. **Pika Music** crea una canción a partir de un prompt de estilo y letras opcionales, con una duración que eliges entre 10 y 360 segundos, y también puede versionar una canción subida con la duración propia de esa canción. **Pika SFX** devuelve un único efecto de sonido de 1 a 20 segundos, que encaja en foley, blips de UI y golpes de tráiler. +Los nodos de audio de Pika convierten una descripción de texto en audio terminado sin necesidad de una GPU local. **Pika Music** escribe una canción a partir de un prompt de estilo y letras opcionales, con la duración que elijas entre 10 y 360 segundos, y también puede versionar una pista subida con la duración propia de esa pista. **Pika SFX** devuelve un efecto de sonido de 1 a 20 segundos, para foley, blips de UI, risers y golpes de tráiler. -Ninguno de los dos nodos descarga pesos. Ambos se ejecutan en tu cuenta de Comfy y se cobran en créditos como los demás nodos asociados. +Ninguno de los dos nodos descarga pesos. Ambos devuelven una única salida `AUDIO`, se ejecutan en tu cuenta de Comfy y se facturan en créditos por segundo de audio generado. + +**Para qué sirven estos nodos** + +- Una sintonía, un jingle o una base instrumental cortada a la duración exacta de un montaje +- Una canción a partir de letras que ya escribiste, con un estilo que puedes cambiar sin tocar las palabras +- Una versión rehecha de una pista que ya tienes: otro género, otro tempo, la misma canción debajo +- Foley, impactos, whooshes, tono de sala y sonidos de eventos para un plano ya terminado +## Primeros pasos + +1. Actualiza ComfyUI a 0.45.0 o superior +2. Ve a la Biblioteca de plantillas y busca `Pika Music` o `Pika SFX` +3. Escribe el prompt, ajusta `duration` y ejecuta el flujo de trabajo + ## Pika Music: texto a música {#api_pika_music} -Escribe los arreglos, el ambiente y la instrumentación en el prompt. Añade letras con etiquetas de sección como `[Verse]` y `[Chorus]` para una pista con voz, o deja el campo de letras vacío para una instrumental. +Escribe los arreglos, el ambiente y la instrumentación en el prompt, y añade letras si quieres voz. El nodo devuelve una pista terminada, mezclada por el modelo. Pika Music text to music workflow preview @@ -28,15 +55,23 @@ Escribe los arreglos, el ambiente y la instrumentación en el prompt. Añade let -1. Elige `text to music` en el nodo `Pika Music` y describe la pista: género, tempo, instrumentos y ambiente. -2. Establece `duration` entre 10 y 360 segundos. Activa `rewrite_prompt` si solo tienes una idea breve y quieres que el modelo la expanda primero. -3. Ejecuta el flujo de trabajo. `Save Audio (Advanced)` escribe un MP3 en `ComfyUI/output/audio/`. +1. Elige `text to music` en el nodo `Pika Music`. +2. Escribe el brief en `prompt`: género y época, BPM y tonalidad, los instrumentos que llevan el gancho, el carácter de la voz y cómo avanza la canción de una sección a otra. Al modelo le llegan hasta 4.000 caracteres. +3. Añade `lyrics` con etiquetas de sección como `[Verse]` y `[Chorus]`, o deja el campo vacío para una instrumental. Hasta 8.000 caracteres. +4. Ajusta `duration` entre 10 y 360 segundos. Activa `rewrite_prompt` solo cuando tu prompt sean unas pocas palabras y quieras que un modelo lo expanda antes. +5. Ejecuta el flujo de trabajo. `Save Audio (Advanced)` escribe un MP3 en `ComfyUI/output/audio/`. + +## Pika Music: versionar una canción {#api_pika_music_cover} + +Cambia el modo del mismo nodo a `cover` para rehacer una canción que ya tienes en lugar de escribir una nueva. Conecta un archivo de audio a `source_audio` y escribe en el prompt el arreglo que quieres en su lugar: otro género, otro tempo, otra banda. -Cambia el modo de ese nodo a `cover` para cambiar el estilo de una canción existente. Conecta un archivo de audio a `source_audio`, describe el nuevo estilo y el resultado conserva la duración de la fuente, ajustada al rango de 5 a 360 segundos. +El cover ignora `duration`. El resultado conserva la duración de la fuente, ajustada al rango de 5 a 360 segundos, así que una subida de tres minutos vuelve con tres minutos. La entrada `lyrics` sigue aplicándose, lo que significa que puedes mantener la letra original o darle al modelo una nueva para el mismo arreglo. + +El cover parte de una mezcla, así que separar voz e instrumental es tarea del modelo. Una voz principal centrada y un arreglo sencillo sobreviven mejor a la rehacer que un máster denso. ## Pika SFX: texto a efectos de sonido {#api_pika_sfx} -Nombra el material, la acción, el espacio y la intensidad en una sola línea. Una descripción cercana y seca se lee mejor que una lista de adjetivos. +Nombra el sonido en una línea y el nodo devuelve un único efecto. Pika sigue el brief sin añadir voz, música ni ruido de fondo que no hayas pedido, así que el efecto entra en la línea de tiempo sin una base que tengas que quitar. Pika SFX text to sound effect workflow preview @@ -46,7 +81,139 @@ Nombra el material, la acción, el espacio y la intensidad en una sola línea. U -1. Escribe el sonido en el prompt de `Pika SFX`, por ejemplo: una vieja puerta de madera que se abre chirriando sobre bisagras oxidadas y luego un pestillo pesado que se cierra de golpe. -2. Establece `duration` entre 1 y 20 segundos y ejecuta el flujo de trabajo. El MP3 queda en `ComfyUI/output/audio/`. +1. Escribe el sonido en el prompt de `Pika SFX`, por ejemplo una vieja puerta de madera que se abre chirriando sobre bisagras oxidadas y luego un pestillo pesado que se cierra de golpe. +2. Ajusta `duration` entre 1 y 20 segundos y ejecuta el flujo de trabajo. El MP3 queda en `ComfyUI/output/audio/`. + +Cada ejecución produce un efecto en estéreo a 44,1 kHz. Para una secuencia de sonidos, ejecuta el nodo una vez por cada sonido y colócalos tú mismo en la línea de tiempo. Pika SFX no escribe música ni voces cantadas; para eso está Pika Music. + +## Controles del nodo + +`Pika Music`, modo text to music: + +| Control | Valores | Efecto | +| --- | --- | --- | +| `prompt` | texto, de 1 a 4.000 caracteres | El brief: género, BPM, instrumentación, carácter de la voz, arreglo | +| `lyrics` | opcional, hasta 8.000 caracteres | La letra que se canta. Las etiquetas de sección fijan la estructura. Vacío significa instrumental | +| `duration` | de 10 a 360 segundos, por defecto 30 | Duración de la pista | +| `rewrite_prompt` | desactivado por defecto | Expande con un modelo un prompt corto o vago antes de generar | +| `seed` | entero, por defecto 42 | Cámbialo para explorar variaciones del mismo brief | + +`Pika Music`, modo cover: + +| Control | Valores | Efecto | +| --- | --- | --- | +| `source_audio` | entrada de audio | La canción a versionar. La salida conserva su duración, ajustada a 5–360 segundos | +| `prompt` | texto, de 1 a 4.000 caracteres | El nuevo arreglo con el que se reconstruye la canción | +| `lyrics` | opcional, hasta 8.000 caracteres | Sustituye la letra o mantén la original | +| `rewrite_prompt` | desactivado por defecto | Expande con un modelo un prompt corto o vago antes de generar | +| `seed` | entero, por defecto 42 | Cámbialo para explorar variaciones del mismo cover | + +`Pika SFX`: + +| Control | Valores | Efecto | +| --- | --- | --- | +| `prompt` | texto, de 1 a 4.000 caracteres | El sonido: material, acción, espacio, perspectiva, intensidad | +| `duration` | de 1 a 20 segundos, por defecto 5 | Duración del efecto | +| `seed` | entero, por defecto 42 | Cámbialo para explorar variaciones del mismo brief | + +## Cómo escribir un prompt de música + +Construye el brief en este orden: **género y época, BPM y tonalidad, los instrumentos que llevan el gancho, carácter de la voz, desarrollo del arreglo, notas de mezcla**. El modelo lee los 4.000 caracteres, así que el detalle concreto aporta y una pila de adjetivos no. + +El prompt por defecto de la plantilla muestra la forma: + +```text +Slow-burning indie soul ballad at 72 BPM: warm Rhodes electric piano, brushed drums, upright bass and a soft horn section that swells in the chorus. Intimate late-night lounge mood, analog tape saturation, wide stereo reverb. +``` + +Más puntos de partida: + +- `Mainstream dance pop, four-on-the-floor beat at 124 BPM in A minor, bright female lead in an upper-mid register, sidechained warm pads, plucked synth hook, filtered verse lifting into a festival-sized chorus, glossy radio mix.` +- `Sparse solo grand piano at 68 BPM, felt hammers and close mic, rubato intro resolving into a steady left-hand arpeggio, no drums, no vocals.` +- `Cinematic orchestral build at 90 BPM, taiko and low-string ostinato, brass swell around twenty seconds, choir pad underneath, wide trailer mix, instrumental.` +- `Lofi hip hop at 82 BPM, dusty boom-bap drums, vinyl crackle, warm electric piano loop, upright bass, no vocals, loopable.` + +Lo que cambia el resultado: + +- **Pide el desarrollo, no solo la paleta.** "Filtered verse lifting into a full chorus" da a la pista un lugar hacia el que ir; una lista larga de instrumentos suele volver como una textura estática. +- **Di lo que no quieres.** `no vocals`, `no ad-libs`, `dry close-mic lead` se leen como dirección. +- **Nombra la interpretación.** El registro, el género de la voz y cuánto se adelanta en la mezcla cambian la canción más que un instrumento extra. +- **Ajusta la duración al uso.** Un sting de 10 segundos y una pieza de 180 piden briefs distintos, así que genera la duración que necesitas en vez de recortar una canción larga. +- **Itera una dimensión a la vez.** Mantén el mismo seed mientras editas el prompt, para que la diferencia que oigas venga del cambio que hiciste. + +## Letras y etiquetas de sección + +`lyrics` es una entrada separada del prompt, así que las mismas palabras pueden volver como balada, sencillo de dance o interpretación de rock cambiando solo el brief. + +- Un campo `lyrics` vacío devuelve una instrumental con el mismo brief, sin voz. +- Las etiquetas fijan la estructura. Se leen como frases y límites de sección, no como texto que cantar. `[Verse]`, `[Pre-Chorus]` y `[Chorus]` cubren la mayoría de las canciones; añade `[Verse 2]`, `[Bridge]` o `[Final Chorus]` para marcar repeticiones y el final. +- Una línea de letra por línea. El modelo frasea a partir de los saltos de línea. + +```text +[Verse] +Streetlights hum on the empty avenue +I keep the radio low and the windows down +[Pre-Chorus] +Every red light is a chance to turn around +[Chorus] +I'm not lost, I'm just taking the long way home +``` + +## Cómo escribir un prompt de efecto de sonido + +Una línea con cuatro cosas: **cuál es el material, qué le ocurre, en qué espacio sucede y a qué distancia está el micrófono**. La intensidad va al final. Una descripción cercana y seca se lee mejor que una lista de adjetivos. + +| Objetivo | Prompt | +| --- | --- | +| Foley: puerta | `Old timber door creaks open slowly on rusty hinges, then a heavy latch clicks shut, close perspective, dry, no music, no voices` | +| Blip de UI | `Short bright digital blip, one hundred milliseconds, clean tone, no reverb` | +| Impacto de tráiler | `Deep sub-bass impact with a metal debris tail, close and heavy, wide stereo, cinematic` | +| Ambiente de multitud | `Reverberant stadium crowd cheering, chants and claps, steady bed with no music` | +| Riser | `Rising noise sweep with a pitch climb across eight seconds, tension building to a hard stop` | +| Detalle mecánico | `Close-up foley inside an antique mechanical clock, steady pendulum tick-tock, tiny brass gears turning, escapement clicking` | + +- **Un evento por ejecución.** Una puerta que se abre y un pestillo que se cierra comparten prompt solo si ocurren en ese orden dentro de la duración que fijaste. +- **Di `no music, no voices`** cuando el efecto tiene que ir bajo diálogo. Mantiene la mezcla limpia. +- **Ajusta la duración al evento.** Un clic es un segundo; un riser tiene que llenar el hueco delante del corte. +- **Escucha variantes.** Vuelve a ejecutar con otro seed y cambia una dimensión cada vez: primero material, luego espacio, luego intensidad. + +## Duración, límites y créditos + +| Límite | Pika Music | Pika SFX | +| --- | --- | --- | +| Prompt | de 1 a 4.000 caracteres | de 1 a 4.000 caracteres | +| Letras | hasta 8.000 caracteres, opcional | no compatible | +| Duración | de 10 a 360 segundos, por defecto 30 | de 1 a 20 segundos, por defecto 5 | +| Duración del cover | sigue a `source_audio`, ajustada a 5–360 segundos | no compatible | +| Salida | una pista en estéreo | un efecto en estéreo a 44,1 kHz | + +Ambos nodos se facturan en créditos por segundo de audio generado, así que un sting de 10 segundos cuesta una fracción de una canción de 3 minutos. Consulta [Precios de los nodos asociados](/es/tutorials/partner-nodes/pricing) para la tarifa actual. + +Con `rewrite_prompt` desactivado, el mismo prompt, letras, duración y seed reproducen el mismo resultado. Con `rewrite_prompt` activado, un modelo reescribe el brief antes de generar, así que el seed por sí solo ya no fija la salida. Déjalo desactivado cuando quieras un flujo reproducible. + +## Salida y guardado + +Ambos nodos devuelven `AUDIO`, y las plantillas lo guardan con `Save Audio (Advanced)`: prefijo `audio/Pika_Music` o `audio/Pika_SFX`, formato `mp3` con calidad `V0`, que queda en `ComfyUI/output/audio/`. Cambia el formato a `flac` cuando quieras un archivo sin pérdida para masterizar o recortar. + +La salida es un flujo de audio normal, así que se combina con el resto del grafo: + +- Conéctala a `Create Video` para pegar la pista a un conjunto de fotogramas +- Ejecuta `Pika SFX` una vez por impacto y superpón los resultados en tu editor, porque una ejecución da un solo efecto +- Ajusta la duración de la pista a la del corte en lugar de recortar una canción más larga en la línea de tiempo + +## Qué nodo de audio de Pika usar + +| Nodo | Entrada | Salida | Cuándo usarlo | +| --- | --- | --- | --- | +| **Pika Music** | prompt, letras, canción fuente opcional | Una canción completa, de 10 a 360 segundos | Necesitas música: una base, un jingle, una canción o un cover | +| **Pika SFX** | prompt | Un efecto, de 1 a 20 segundos | Necesitas un sonido concreto a partir de una descripción | +| **Pika Soundtrack** | un video, instrucción opcional | El audio del clip sustituido por una banda sonora sincronizada | El sonido tiene que seguir a la imagen, hasta 10 minutos | + +Para un efecto sincronizado con la acción en pantalla, la ruta de video es [Pika Soundtrack en la página de Pika H3 Flex](/es/tutorials/partner-nodes/pika/pika-h3-flex), que lee el clip en lugar de una descripción. + +## Páginas relacionadas -Cada ejecución produce un único efecto de sonido mono. Para una secuencia de sonidos, ejecuta el nodo una vez por cada sonido y colócalos tú mismo en la línea de tiempo. \ No newline at end of file +- [Pika H3 Flex](/es/tutorials/partner-nodes/pika/pika-h3-flex): generación de video con audio sincronizado +- [Mirelo SFX 2.0](/es/tutorials/partner-nodes/mirelo/mirelo-sfx): foley sincronizado con un clip y texto a SFX +- [ACE-Step 1.5](/es/tutorials/audio/ace-step/ace-step-v1-5): un modelo de música de pesos abiertos que corre en tu propia GPU +- [Precios de los nodos asociados](/es/tutorials/partner-nodes/pricing) y [límites de concurrencia](/es/tutorials/partner-nodes/concurrency-limits) diff --git a/es/tutorials/partner-nodes/pika/pika-h3-flex.mdx b/es/tutorials/partner-nodes/pika/pika-h3-flex.mdx index a6ce3e4999..ec66413472 100644 --- a/es/tutorials/partner-nodes/pika/pika-h3-flex.mdx +++ b/es/tutorials/partner-nodes/pika/pika-h3-flex.mdx @@ -68,6 +68,8 @@ Tres nodos de audio se ejecutan en la misma API de Pika y se facturan por segund - **Pika SFX**: genera un efecto de sonido a partir de una descripción de texto, de 1 a 20 segundos de duración. - **Pika Music**: genera una canción a partir de un prompt de texto y letras opcionales, de 10 a 360 segundos, o versiona una canción original con su propia duración, limitada a 5 a 360 segundos. +Para ver los controles, los patrones de prompt y los límites de estos dos nodos de audio, consulta [Pika Music y Pika SFX](/es/tutorials/partner-nodes/pika/pika-audio). + ## Formato del prompt Los prompts de Pika H3 Flex siguen la estructura del modo de texto de MiniMax H3, donde una sola descripción reúne la imagen y el sonido: diff --git a/ja/tutorials/partner-nodes/pika/pika-audio.mdx b/ja/tutorials/partner-nodes/pika/pika-audio.mdx index d27061a731..c142e00a4c 100644 --- a/ja/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/ja/tutorials/partner-nodes/pika/pika-audio.mdx @@ -1,24 +1,51 @@ --- -title: "ComfyUI の Pika Music と Pika SFX: 楽曲とサウンドエフェクト" -description: "スタイルプロンプトと任意の歌詞から楽曲全体を生成するか、既存のトラックをカバーするか、Pika オーディオノードで 1〜20 秒のサウンドエフェクトを作成します。" +title: "ComfyUI の Pika Music と Pika SFX: AI 楽曲とサウンドエフェクト" +description: "ComfyUI の Pika Music と Pika SFX ノードで、スタイルプロンプトと歌詞から楽曲全体を生成し、既存のトラックをカバーし、1〜20 秒のサウンドエフェクトを作成します。" sidebarTitle: "Pika Music & SFX" -translationSourceHash: 60f1ba51 +translationSourceHash: c04fcf5c translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx +translationBlockHashes: + "_intro": dc3cd241 + "Get started": fd80ca7d + "Pika Music: text to music {#api_pika_music}": ae1738bd + "Pika Music: cover a song {#api_pika_music_cover}": 0793be8c + "Pika SFX: text to sound effects {#api_pika_sfx}": 81088686 + "Node controls": 2c4dd7da + "Writing a music prompt": a24287f2 + "Lyrics and section tags": af8f7f9e + "Writing a sound effect prompt": 20f01804 + "Duration, limits and credits": 763b11c7 + "Output and saving": 4abf0bf7 + "Which Pika audio node to use": 8b598461 + "Related pages": 96e4c1fe --- import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; -Pika のオーディオノードは、ローカル GPU なしでテキストの説明から完成したオーディオを作り出します。**Pika Music** はスタイルプロンプトと任意の歌詞から、10〜360 秒の範囲で選んだ長さの楽曲を書き出し、アップロードした楽曲をその曲本来の長さでカバーすることもできます。**Pika SFX** は 1〜20 秒の単一のサウンドエフェクトを返し、フォーリー、UI の効果音、トレーラーのヒット音に適しています。 +Pika のオーディオノードは、ローカル GPU なしでテキストの説明から完成したオーディオを作り出します。**Pika Music** はスタイルプロンプトと任意の歌詞から、10〜360 秒の範囲で選んだ長さの楽曲を書き出し、アップロードしたトラックをその曲本来の長さでカバーすることもできます。**Pika SFX** は 1〜20 秒の単一のサウンドエフェクトを返し、フォーリー、UI の効果音、ライザー、トレーラーのヒット音に使えます。 -どちらのノードも重みをダウンロードしません。どちらも Comfy アカウント上で実行され、他のパートナーノードと同様にクレジットで課金されます。 +どちらのノードも重みをダウンロードしません。出力は `AUDIO` 1 本だけで、Comfy アカウント上で実行され、生成されたオーディオの秒数に応じてクレジットで課金されます。 + +**このノードで作れるもの** + +- 動画編集の尺にそのまま合わせた楽曲、ジングル、インストゥルメンタルの下地用トラック +- 書き上げた歌詞を歌にすること。言葉を変えずにスタイルだけ差し替えられます +- 手持ちのトラックの作り直し: ジャンルもテンポも変えて、下にあるのは同じ曲 +- 完成した映像のためのフォーリー、インパクト、ホイッシュ、ルームトーン、ゲームイベント用の効果音 +## はじめに + +1. ComfyUI を 0.45.0 以降に更新します +2. テンプレートライブラリで `Pika Music` または `Pika SFX` を検索します +3. プロンプトを書き、`duration` を設定してワークフローを実行します + ## Pika Music: テキストから楽曲へ {#api_pika_music} -アレンジ、雰囲気、楽器編成をプロンプトに書き込みます。ボーカルトラックにしたい場合は `[Verse]` や `[Chorus]` などのタグを使ったセクション付きの歌詞を追加し、インストゥルメンタルにしたい場合は歌詞フィールドを空のままにします。 +アレンジ、雰囲気、楽器編成をプロンプトに書き、ボーカルが必要なら歌詞を加えます。ノードはミックス済みの完成トラックを 1 本返します。 Pika Music text to music workflow preview @@ -28,15 +55,23 @@ Pika のオーディオノードは、ローカル GPU なしでテキストの -1. `Pika Music` ノードで `text to music` を選択し、トラックを説明します。ジャンル、テンポ、楽器、雰囲気を指定します。 -2. `duration` を 10〜360 秒の範囲で設定します。短いアイデアしかなく、モデルにまずそれを展開してほしい場合は `rewrite_prompt` をオンにします。 -3. ワークフローを実行します。`Save Audio (Advanced)` が MP3 を `ComfyUI/output/audio/` に書き出します。 +1. `Pika Music` ノードで `text to music` を選択します。 +2. `prompt` に方向性を書きます。ジャンルと年代、BPM とキー、フックを担う楽器、ボーカルの性格、そして曲がセクションごとにどう進むか。最大 4,000 文字がモデルに届きます。 +3. `[Verse]` や `[Chorus]` などのセクションタグ付きの `lyrics` を追加します。インストゥルメンタルにする場合は空のままにします。上限は 8,000 文字です。 +4. `duration` を 10〜360 秒の範囲で設定します。`rewrite_prompt` は、プロンプトが数語だけで、まずモデルに展開してほしいときだけオンにします。 +5. ワークフローを実行します。`Save Audio (Advanced)` が MP3 を `ComfyUI/output/audio/` に書き出します。 + +## Pika Music: 曲をカバーする {#api_pika_music_cover} + +同じノードのモードを `cover` に切り替えると、新しく書き下ろす代わりに手持ちの曲を作り直せます。オーディオファイルを `source_audio` に接続し、代わりに欲しいアレンジをプロンプトに書きます。別のジャンル、別のテンポ、別のバンドです。 -代わりに既存の楽曲をスタイル変更するには、同じノードのモードを `cover` に切り替えます。オーディオファイルを `source_audio` に接続し、新しいスタイルを説明すると、結果はソースの長さを保ち、5〜360 秒に制限されます。 +カバーは `duration` を無視します。出力はソースの長さを保ち、5〜360 秒に制限されるため、3 分の素材は 3 分のまま返ります。`lyrics` は引き続き有効で、元の歌詞を残すことも、同じアレンジに新しい歌詞を与えることもできます。 + +カバーはミックス済みの音源から出発するので、ボーカルと伴奏の分離はモデルの仕事です。センターに定位したリードボーカルとシンプルなアレンジのほうが、密度の高いマスターよりも作り直しに耐えます。 ## Pika SFX: テキストからサウンドエフェクトへ {#api_pika_sfx} -マテリアル、動作、空間、強度を 1 行で指定します。形容詞を並べるよりも、近くで乾いた描写のほうがうまく伝わります。 +サウンドを 1 行で指定すると、ノードはエフェクトを 1 本返します。Pika は指示に従い、求めていない音声・音楽・背景ノイズを加えないので、余分な下地を剥がさずにそのままタイムラインへ置けます。 Pika SFX text to sound effect workflow preview @@ -46,7 +81,139 @@ Pika のオーディオノードは、ローカル GPU なしでテキストの -1. `Pika SFX` のプロンプトにサウンドを書きます。たとえば、古い木の扉が錆びた蝶番で軋みながら開き、その後重い掛け金がカチッと閉まる様子などです。 +1. `Pika SFX` のプロンプトにサウンドを書きます。たとえば、古い木の扉が錆びた蝶番でゆっくり軋みながら開き、その後で重い掛け金がカチッと閉まる様子です。 2. `duration` を 1〜20 秒の範囲で設定し、ワークフローを実行します。MP3 は `ComfyUI/output/audio/` に出力されます。 -実行ごとに 1 つのモノラルのサウンドエフェクトが生成されます。一連のサウンドが必要な場合は、サウンドごとにノードを 1 回実行し、タイムライン上に自分で配置してください。 \ No newline at end of file +1 回の実行で 44.1 kHz ステレオのエフェクトが 1 本生成されます。連続したサウンドが必要な場合は、サウンドごとにノードを 1 回実行し、タイムライン上に自分で配置してください。Pika SFX は音楽や歌唱を生成しません。それは Pika Music の担当です。 + +## ノードのコントロール + +`Pika Music`、text to music モード: + +| コントロール | 値 | 効果 | +| --- | --- | --- | +| `prompt` | テキスト、1〜4,000 文字 | 方向性: ジャンル、BPM、楽器編成、ボーカルの性格、アレンジ | +| `lyrics` | 任意、最大 8,000 文字 | 歌う歌詞。セクションタグが構造を決めます。空ならインストゥルメンタル | +| `duration` | 10〜360 秒、既定 30 | トラックの長さ | +| `rewrite_prompt` | 既定はオフ | 生成前に短い・曖昧なプロンプトをモデルで展開します | +| `seed` | 整数、既定 42 | 同じ方向性のバリエーションを試すときに変えます | + +`Pika Music`、cover モード: + +| コントロール | 値 | 効果 | +| --- | --- | --- | +| `source_audio` | オーディオ入力 | カバーする曲。出力はその長さを保ち、5〜360 秒に制限されます | +| `prompt` | テキスト、1〜4,000 文字 | その曲をどんな新しいアレンジに作り直すか | +| `lyrics` | 任意、最大 8,000 文字 | 歌詞を差し替えるか、元の詞を残すか | +| `rewrite_prompt` | 既定はオフ | 生成前に短い・曖昧なプロンプトをモデルで展開します | +| `seed` | 整数、既定 42 | 同じカバーのバリエーションを試すときに変えます | + +`Pika SFX`: + +| コントロール | 値 | 効果 | +| --- | --- | --- | +| `prompt` | テキスト、1〜4,000 文字 | サウンドそのもの: 素材、動作、空間、マイクの距離感、強度 | +| `duration` | 1〜20 秒、既定 5 | エフェクトの長さ | +| `seed` | 整数、既定 42 | 同じ方向性のバリエーションを試すときに変えます | + +## 楽曲プロンプトの書き方 + +**ジャンルと年代、BPM とキー、フックを担う楽器、ボーカルの性格、アレンジの展開、ミックスの方向性** の順で組み立てます。4,000 文字すべてがモデルに届くので、具体的な情報は効きますが、形容詞を積み重ねても効きません。 + +テンプレートの既定プロンプトが、この形の実例です: + +```text +Slow-burning indie soul ballad at 72 BPM: warm Rhodes electric piano, brushed drums, upright bass and a soft horn section that swells in the chorus. Intimate late-night lounge mood, analog tape saturation, wide stereo reverb. +``` + +さらにいくつかの出発点: + +- `Mainstream dance pop, four-on-the-floor beat at 124 BPM in A minor, bright female lead in an upper-mid register, sidechained warm pads, plucked synth hook, filtered verse lifting into a festival-sized chorus, glossy radio mix.` +- `Sparse solo grand piano at 68 BPM, felt hammers and close mic, rubato intro resolving into a steady left-hand arpeggio, no drums, no vocals.` +- `Cinematic orchestral build at 90 BPM, taiko and low-string ostinato, brass swell around twenty seconds, choir pad underneath, wide trailer mix, instrumental.` +- `Lofi hip hop at 82 BPM, dusty boom-bap drums, vinyl crackle, warm electric piano loop, upright bass, no vocals, loopable.` + +結果が変わるポイント: + +- **音色だけでなく展開を書く。** `Filtered verse lifting into a full chorus` と書くと曲に進む先ができます。楽器名を長く並べただけだと、静的な音の塊が返りがちです。 +- **不要なものを明示する。** `no vocals`、`no ad-libs`、`dry close-mic lead` はいずれも方向性として読まれます。 +- **歌い方を指定する。** 音域、性別、ミックス内でのボーカルの前に出方のほうが、楽器を 1 つ足すより曲を変えます。 +- **用途に合わせて長さを決める。** 10 秒のステイングと 180 秒のBGMでは必要な方向性が違います。長い曲を作って切るのではなく、必要な長さで生成します。 +- **一度に 1 つだけ変える。** プロンプトを編集する間は seed を固定すると、聞こえる違いは自分が加えた変更によるものになります。 + +## 歌詞とセクションタグ + +`lyrics` はプロンプトとは別の入力なので、同じ歌詞のままプロンプトだけを変えて、バラードにもダンスシングルにもロックの演奏にもできます。 + +- `lyrics` が空のときは、同じ方向性でボーカルのないインストゥルメンタルが返ります。 +- タグは構造を決めます。歌う文字ではなく、フレーズの切り方とセクションの境界として読まれます。`[Verse]`、`[Pre-Chorus]`、`[Chorus]` でほとんどの曲は足ります。繰り返しや終わりを指定するときは `[Verse 2]`、`[Bridge]`、`[Final Chorus]` を加えます。 +- 歌詞は 1 行に 1 フレーズ。モデルは改行を区切りとしてフレーズを作ります。 + +```text +[Verse] +Streetlights hum on the empty avenue +I keep the radio low and the windows down +[Pre-Chorus] +Every red light is a chance to turn around +[Chorus] +I'm not lost, I'm just taking the long way home +``` + +## サウンドエフェクトのプロンプトの書き方 + +1 行に 4 つを入れます: **素材は何か、それに何が起きるか、どんな空間か、マイクがどれだけ近いか**。強度は最後です。形容詞を並べるより、近くて乾いた描写のほうがよく伝わります。 + +| 目的 | プロンプト | +| --- | --- | +| フォーリー: 扉 | `Old timber door creaks open slowly on rusty hinges, then a heavy latch clicks shut, close perspective, dry, no music, no voices` | +| UI の効果音 | `Short bright digital blip, one hundred milliseconds, clean tone, no reverb` | +| トレーラーのインパクト | `Deep sub-bass impact with a metal debris tail, close and heavy, wide stereo, cinematic` | +| 群衆の環境音 | `Reverberant stadium crowd cheering, chants and claps, steady bed with no music` | +| ライザー | `Rising noise sweep with a pitch climb across eight seconds, tension building to a hard stop` | +| 機械のディテール | `Close-up foley inside an antique mechanical clock, steady pendulum tick-tock, tiny brass gears turning, escapement clicking` | + +- **1 回につき 1 イベント。** 扉が開いて掛け金が閉まるのは、その順番で設定した長さの中に収まるときだけ、同じプロンプトに書けます。 +- **台詞の下に敷くなら `no music, no voices`。** 生成物がきれいに残ります。 +- **長さはイベントに合わせる。** クリック音は 1 秒で足りますが、ライザーはカット前の間を埋める必要があります。 +- **複数試す。** 新しい seed で再実行し、素材、空間、強度の順に 1 つずつ変えます。 + +## 長さ、制限、クレジット + +| 制限 | Pika Music | Pika SFX | +| --- | --- | --- | +| プロンプト | 1〜4,000 文字 | 1〜4,000 文字 | +| 歌詞 | 任意、最大 8,000 文字 | 非対応 | +| 長さ | 10〜360 秒、既定 30 | 1〜20 秒、既定 5 | +| カバーの長さ | `source_audio` に従い、5〜360 秒に制限 | 非対応 | +| 出力 | ステレオのトラック 1 本 | 44.1 kHz ステレオのエフェクト 1 本 | + +どちらのノードも生成されたオーディオの秒数でクレジット課金されるため、10 秒のステイングは 3 分の曲のごく一部の費用で済みます。現在のレートは [パートナーノードの料金](/ja/tutorials/partner-nodes/pricing) を参照してください。 + +`rewrite_prompt` をオフにした状態では、同じプロンプト、歌詞、長さ、seed で同じ結果が再現されます。`rewrite_prompt` をオンにすると生成前にモデルがプロンプトを書き換えるため、seed だけでは出力を固定できません。再現性のあるワークフローが必要なときはオフのままにしてください。 + +## 出力と保存 + +どちらのノードも `AUDIO` を出力し、テンプレートは `Save Audio (Advanced)` で保存します。ファイル名のプレフィックスは `audio/Pika_Music` または `audio/Pika_SFX`、形式は `mp3` で品質 `V0`、保存先は `ComfyUI/output/audio/` です。マスターや再編集用のロスレスが必要なら、形式を `flac` に切り替えます。 + +出力は通常のオーディオストリームなので、グラフの他のノードとそのまま組み合わせられます: + +- `Create Video` に接続して、一連のフレームにトラックを付けられます +- ヒットごとに `Pika SFX` を 1 回ずつ実行し、編集ソフトで重ねます。1 回の実行でエフェクトは 1 本だけです +- タイムライン上で長い曲を切るのではなく、トラックの長さをカットの尺に合わせて設定します + +## どの Pika オーディオノードを使うか + +| ノード | 入力 | 出力 | 使う場面 | +| --- | --- | --- | --- | +| **Pika Music** | プロンプト、歌詞、任意の元曲 | 10〜360 秒の楽曲 | 音楽が必要なとき: 下地、ジングル、歌、カバー | +| **Pika SFX** | プロンプト | 1〜20 秒のエフェクト 1 本 | 説明から特定の 1 つの音がほしいとき | +| **Pika Soundtrack** | ビデオ、任意の指示 | クリップのオーディオを同期したサウンドトラックに置き換え | 音が映像に合わせる必要があるとき。最大 10 分 | + +画面上の動きに合わせて音を鳴らしたい場合は、映像側の経路を使います: [Pika H3 Flex ページの Pika Soundtrack](/ja/tutorials/partner-nodes/pika/pika-h3-flex)。説明ではなくクリップを読み取ります。 + +## 関連ページ + +- [Pika H3 Flex](/ja/tutorials/partner-nodes/pika/pika-h3-flex): 同期したオーディオ付きのビデオ生成 +- [Mirelo SFX 2.0](/ja/tutorials/partner-nodes/mirelo/mirelo-sfx): クリップに合わせたフォーリーと、テキストから SFX +- [ACE-Step 1.5](/ja/tutorials/audio/ace-step/ace-step-v1-5): ローカルの GPU で動くオープンウェイトの楽曲モデル +- [パートナーノードの料金](/ja/tutorials/partner-nodes/pricing) と [同時実行の制限](/ja/tutorials/partner-nodes/concurrency-limits) diff --git a/ja/tutorials/partner-nodes/pika/pika-h3-flex.mdx b/ja/tutorials/partner-nodes/pika/pika-h3-flex.mdx index e1cfc20bfb..39020489b7 100644 --- a/ja/tutorials/partner-nodes/pika/pika-h3-flex.mdx +++ b/ja/tutorials/partner-nodes/pika/pika-h3-flex.mdx @@ -68,6 +68,8 @@ Pika H3 Flexは、[MiniMax H3](/ja/tutorials/video/minimax/minimax-h3)をPikaが - **Pika SFX**: テキストの説明から、長さ1~20秒のサウンドエフェクトを生成します - **Pika Music**: テキストプロンプトと任意の歌詞から10~360秒の楽曲を生成するか、元の曲をその長さのままカバーします(5~360秒に制限されます) +テキストから生成するこの2つのオーディオノードのコントロール、プロンプトの書き方、制限については [Pika Music と Pika SFX](/ja/tutorials/partner-nodes/pika/pika-audio) を参照してください。 + ## プロンプト形式 Pika H3 Flex のプロンプトは MiniMax H3 のテキストモードの構造に従い、1つの説明で映像とサウンドの両方を表現します。 diff --git a/ko/tutorials/partner-nodes/pika/pika-audio.mdx b/ko/tutorials/partner-nodes/pika/pika-audio.mdx index 281069a664..b1fa7ba5c9 100644 --- a/ko/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/ko/tutorials/partner-nodes/pika/pika-audio.mdx @@ -1,24 +1,51 @@ --- -title: "ComfyUI의 Pika Music 및 Pika SFX: 노래와 사운드 효과" -description: "스타일 프롬프트와 선택적 가사로 전체 곡을 생성하거나, 기존 트랙을 커버하거나, ComfyUI의 Pika 오디오 노드로 1~20초 길이의 사운드 효과를 만들 수 있습니다." +title: "ComfyUI의 Pika Music 및 Pika SFX: AI 노래와 사운드 효과" +description: "ComfyUI의 Pika Music 및 Pika SFX 노드로 스타일 프롬프트와 가사에서 노래 전체를 생성하고, 기존 트랙을 커버하며, 1~20초 사운드 효과를 만들 수 있습니다." sidebarTitle: "Pika Music 및 SFX" -translationSourceHash: 60f1ba51 +translationSourceHash: c04fcf5c translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx +translationBlockHashes: + "_intro": dc3cd241 + "Get started": fd80ca7d + "Pika Music: text to music {#api_pika_music}": ae1738bd + "Pika Music: cover a song {#api_pika_music_cover}": 0793be8c + "Pika SFX: text to sound effects {#api_pika_sfx}": 81088686 + "Node controls": 2c4dd7da + "Writing a music prompt": a24287f2 + "Lyrics and section tags": af8f7f9e + "Writing a sound effect prompt": 20f01804 + "Duration, limits and credits": 763b11c7 + "Output and saving": 4abf0bf7 + "Which Pika audio node to use": 8b598461 + "Related pages": 96e4c1fe --- import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; -Pika의 오디오 노드는 로컬 GPU 없이 텍스트 설명을 완성된 오디오로 바꿉니다. **Pika Music**은 스타일 프롬프트와 선택적 가사로 노래를 작성하며, 길이는 10~360초에서 선택할 수 있고, 업로드한 노래를 해당 노래의 원래 길이로 커버할 수도 있습니다. **Pika SFX**는 1~20초의 단일 사운드 효과를 반환하며, 폴리, UI 블립, 예고편 임팩트에 적합합니다. +Pika의 오디오 노드는 로컬 GPU 없이 텍스트 설명을 완성된 오디오로 바꿉니다. **Pika Music**은 스타일 프롬프트와 선택적 가사로 노래를 작성하며, 길이는 10~360초에서 선택할 수 있고, 업로드한 트랙을 그 곡의 원래 길이로 커버할 수도 있습니다. **Pika SFX**는 1~20초의 사운드 효과를 하나 반환하며, 폴리, UI 블립, 라이저, 예고편 임팩트에 쓸 수 있습니다. -두 노드 모두 가중치를 다운로드하지 않습니다. 둘 다 Comfy 계정에서 실행되며 다른 파트너 노드와 마찬가지로 크레딧으로 청구됩니다. +두 노드 모두 가중치를 다운로드하지 않습니다. 출력은 `AUDIO` 하나이고, Comfy 계정에서 실행되며, 생성된 오디오의 초 단위로 크레딧이 청구됩니다. + +**이 노드로 만드는 것** + +- 영상 편집 길이에 정확히 맞춘 큐, 징글, 인스트루멘털 베드 +- 이미 써 둔 가사를 노래로. 단어는 그대로 두고 스타일만 바꿀 수 있습니다 +- 가지고 있는 트랙의 재해석: 장르와 템포를 바꾸고, 그 아래에는 같은 곡 +- 완성된 화면을 위한 폴리, 임팩트, 휘슬 사운드, 룸 톤, 게임 이벤트 효과음 +## 시작하기 + +1. ComfyUI를 0.45.0 이상으로 업데이트합니다 +2. 템플릿 라이브러리에서 `Pika Music` 또는 `Pika SFX`를 검색합니다 +3. 프롬프트를 쓰고 `duration`을 설정한 다음 워크플로를 실행합니다 + ## Pika Music: 텍스트에서 음악으로 {#api_pika_music} -프롬프트에 편곡, 분위기, 악기 구성을 작성하세요. 보컬 트랙에는 `[Verse]`, `[Chorus]` 같은 태그로 섹션을 표시한 가사를 추가하고, 연주곡을 원하면 가사 필드를 비워 두세요. +편곡, 분위기, 악기 구성을 프롬프트에 쓰고, 보컬이 필요하면 가사를 더합니다. 노드는 모델이 믹스한 완성 트랙 하나를 반환합니다. Pika Music 텍스트에서 음악으로 워크플로 미리보기 @@ -28,15 +55,23 @@ Pika의 오디오 노드는 로컬 GPU 없이 텍스트 설명을 완성된 오 -1. `Pika Music` 노드에서 `text to music`을 선택하고 트랙을 설명하세요: 장르, 템포, 악기, 분위기. -2. `duration`을 10~360초로 설정하세요. 짧은 아이디어만 있고 모델이 먼저 이를 확장하길 원하면 `rewrite_prompt`를 켜세요. -3. 워크플로를 실행하세요. `Save Audio (Advanced)`가 MP3를 `ComfyUI/output/audio/`에 저장합니다. +1. `Pika Music` 노드에서 `text to music`을 선택합니다. +2. `prompt`에 방향을 씁니다: 장르와 시대, BPM과 키, 훅을 담당하는 악기, 보컬의 성격, 그리고 곡이 섹션마다 어떻게 진행되는지. 최대 4,000자가 모델에 전달됩니다. +3. `[Verse]`, `[Chorus]` 같은 섹션 태그가 있는 `lyrics`를 추가합니다. 연주곡을 원하면 이 필드를 비워 둡니다. 최대 8,000자입니다. +4. `duration`을 10~360초로 설정합니다. `rewrite_prompt`는 프롬프트가 몇 단어뿐이고 모델이 먼저 확장해 주길 원할 때만 켜세요. +5. 워크플로를 실행합니다. `Save Audio (Advanced)`가 MP3를 `ComfyUI/output/audio/`에 저장합니다. + +## Pika Music: 곡 커버하기 {#api_pika_music_cover} + +같은 노드의 모드를 `cover`로 바꾸면 새로 쓰는 대신 가진 곡을 다시 만들 수 있습니다. 오디오 파일을 `source_audio`에 연결하고, 대신 원하는 편곡을 프롬프트에 씁니다. 다른 장르, 다른 템포, 다른 밴드입니다. -기존 노래의 스타일을 바꾸려면 같은 노드의 모드를 `cover`로 전환하세요. 오디오 파일을 `source_audio`에 연결하고 새 스타일을 설명하면, 결과는 소스 길이를 유지하며 5~360초로 제한됩니다. +커버는 `duration`을 무시합니다. 출력은 소스 길이를 유지하며 5~360초로 제한되므로, 3분짜리 소재는 3분 그대로 돌아옵니다. `lyrics`는 여전히 적용되므로 원래 가사를 유지하거나 같은 편곡에 새 가사를 줄 수 있습니다. + +커버는 믹스된 음원에서 시작하므로 보컬과 반주를 분리하는 일은 모델의 몫입니다. 보컬이 중앙에 있고 편곡이 단순한 소재가 밀도 높은 마스터보다 재해석에 더 잘 견딥니다. ## Pika SFX: 텍스트에서 사운드 효과로 {#api_pika_sfx} -재질, 동작, 공간, 강도를 한 줄로 지정하세요. 구체적이고 담백한 설명이 형용사 목록보다 더 잘 해석됩니다. +사운드를 한 줄로 적으면 노드가 효과음 하나를 반환합니다. Pika는 요청한 설명을 따르며 요청하지 않은 음성, 음악, 배경 노이즈를 더하지 않으므로, 걷어낼 깔개 없이 바로 타임라인에 올릴 수 있습니다. Pika SFX 텍스트에서 사운드 효과로 워크플로 미리보기 @@ -46,7 +81,139 @@ Pika의 오디오 노드는 로컬 GPU 없이 텍스트 설명을 완성된 오 -1. `Pika SFX` 프롬프트에 사운드를 작성하세요. 예를 들어 녹슨 경첩이 삐걱이며 열리는 오래된 목재 문, 이어서 무거운 걸쇠가 철컥 닫히는 소리입니다. -2. `duration`을 1~20초로 설정한 다음 워크플로를 실행하세요. MP3는 `ComfyUI/output/audio/`에 저장됩니다. +1. `Pika SFX` 프롬프트에 사운드를 씁니다. 예를 들어 녹슨 경첩에서 삐걱이며 천천히 열리는 낡은 목재 문, 이어서 무거운 걸쇠가 철컥 닫히는 소리입니다. +2. `duration`을 1~20초로 설정한 다음 워크플로를 실행합니다. MP3는 `ComfyUI/output/audio/`에 저장됩니다. + +실행 한 번에 44.1 kHz 스테레오 효과음 하나가 생성됩니다. 연속된 사운드가 필요하면 사운드마다 노드를 한 번씩 실행하고 타임라인에 직접 배치하세요. Pika SFX는 음악이나 노래를 만들지 않습니다. 그 부분은 Pika Music이 담당합니다. + +## 노드 컨트롤 + +`Pika Music`, text to music 모드: + +| 컨트롤 | 값 | 효과 | +| --- | --- | --- | +| `prompt` | 텍스트, 1~4,000자 | 방향: 장르, BPM, 악기 구성, 보컬 성격, 편곡 | +| `lyrics` | 선택, 최대 8,000자 | 부를 가사. 섹션 태그가 구조를 정합니다. 비우면 연주곡 | +| `duration` | 10~360초, 기본 30 | 트랙 길이 | +| `rewrite_prompt` | 기본 꺼짐 | 생성 전에 짧거나 모호한 프롬프트를 모델로 확장합니다 | +| `seed` | 정수, 기본 42 | 같은 방향의 변형을 탐색할 때 바꿉니다 | + +`Pika Music`, cover 모드: + +| 컨트롤 | 값 | 효과 | +| --- | --- | --- | +| `source_audio` | 오디오 입력 | 커버할 곡. 출력은 그 길이를 유지하며 5~360초로 제한됩니다 | +| `prompt` | 텍스트, 1~4,000자 | 그 곡을 어떤 새 편곡으로 다시 만들지 | +| `lyrics` | 선택, 최대 8,000자 | 가사를 바꾸거나 원래 가사를 유지합니다 | +| `rewrite_prompt` | 기본 꺼짐 | 생성 전에 짧거나 모호한 프롬프트를 모델로 확장합니다 | +| `seed` | 정수, 기본 42 | 같은 커버의 변형을 탐색할 때 바꿉니다 | + +`Pika SFX`: + +| 컨트롤 | 값 | 효과 | +| --- | --- | --- | +| `prompt` | 텍스트, 1~4,000자 | 사운드 자체: 재질, 동작, 공간, 마이크 거리감, 강도 | +| `duration` | 1~20초, 기본 5 | 효과음 길이 | +| `seed` | 정수, 기본 42 | 같은 방향의 변형을 탐색할 때 바꿉니다 | + +## 음악 프롬프트 작성법 + +**장르와 시대, BPM과 키, 훅을 담당하는 악기, 보컬 성격, 편곡의 진행, 믹스 방향** 순서로 방향을 세웁니다. 4,000자 전체가 모델에 전달되므로 구체적인 정보는 값을 하고, 형용사만 쌓으면 그렇지 않습니다. + +템플릿의 기본 프롬프트가 이 형태의 실례입니다: + +```text +Slow-burning indie soul ballad at 72 BPM: warm Rhodes electric piano, brushed drums, upright bass and a soft horn section that swells in the chorus. Intimate late-night lounge mood, analog tape saturation, wide stereo reverb. +``` + +더 많은 출발점: + +- `Mainstream dance pop, four-on-the-floor beat at 124 BPM in A minor, bright female lead in an upper-mid register, sidechained warm pads, plucked synth hook, filtered verse lifting into a festival-sized chorus, glossy radio mix.` +- `Sparse solo grand piano at 68 BPM, felt hammers and close mic, rubato intro resolving into a steady left-hand arpeggio, no drums, no vocals.` +- `Cinematic orchestral build at 90 BPM, taiko and low-string ostinato, brass swell around twenty seconds, choir pad underneath, wide trailer mix, instrumental.` +- `Lofi hip hop at 82 BPM, dusty boom-bap drums, vinyl crackle, warm electric piano loop, upright bass, no vocals, loopable.` + +결과를 바꾸는 것들: + +- **음색만이 아니라 진행을 쓰세요.** `Filtered verse lifting into a full chorus`라고 쓰면 곡에 갈 곳이 생깁니다. 악기 이름만 길게 나열하면 정적인 음향 덩어리가 돌아오기 쉽습니다. +- **원하지 않는 것을 말하세요.** `no vocals`, `no ad-libs`, `dry close-mic lead` 모두 방향으로 읽힙니다. +- **창법을 지정하세요.** 음역, 성별, 믹스에서 보컬이 앞에 나오는 정도가 악기를 하나 더 넣는 것보다 곡을 더 바꿉니다. +- **용도에 맞춰 길이를 정하세요.** 10초 스팅과 180초 배경 음악은 필요한 방향이 다릅니다. 긴 곡을 만들어 자르지 말고 필요한 길이로 생성하세요. +- **한 번에 하나만 바꾸세요.** 프롬프트를 고치는 동안 seed를 고정하면, 들리는 차이는 내가 가한 변경에서 나옵니다. + +## 가사와 섹션 태그 + +`lyrics`는 프롬프트와 분리된 입력이므로, 같은 가사에 프롬프트만 바꿔 발라드, 댄스 싱글, 록 연주로 다시 받을 수 있습니다. + +- `lyrics`를 비우면 같은 방향의 보컬 없는 연주곡이 돌아옵니다. +- 태그는 구조를 정합니다. 부를 텍스트가 아니라 프레이즈를 끊는 방식과 섹션 경계로 읽힙니다. `[Verse]`, `[Pre-Chorus]`, `[Chorus]`면 대부분의 곡에 충분하고, 반복과 엔딩을 지정하려면 `[Verse 2]`, `[Bridge]`, `[Final Chorus]`를 더합니다. +- 가사는 한 줄에 한 프레이즈. 모델은 줄바꿈을 기준으로 프레이즈를 만듭니다. + +```text +[Verse] +Streetlights hum on the empty avenue +I keep the radio low and the windows down +[Pre-Chorus] +Every red light is a chance to turn around +[Chorus] +I'm not lost, I'm just taking the long way home +``` + +## 사운드 효과 프롬프트 작성법 + +한 줄에 네 가지를 넣습니다: **재질이 무엇인지, 그에게 무슨 일이 일어나는지, 어떤 공간인지, 마이크가 얼마나 가까운지**. 강도는 마지막입니다. 형용사를 나열하기보다 가깝고 건조한 묘사가 더 잘 전달됩니다. + +| 목적 | 프롬프트 | +| --- | --- | +| 폴리: 문 | `Old timber door creaks open slowly on rusty hinges, then a heavy latch clicks shut, close perspective, dry, no music, no voices` | +| UI 블립 | `Short bright digital blip, one hundred milliseconds, clean tone, no reverb` | +| 예고편 임팩트 | `Deep sub-bass impact with a metal debris tail, close and heavy, wide stereo, cinematic` | +| 군중 앰비언스 | `Reverberant stadium crowd cheering, chants and claps, steady bed with no music` | +| 라이저 | `Rising noise sweep with a pitch climb across eight seconds, tension building to a hard stop` | +| 기계 디테일 | `Close-up foley inside an antique mechanical clock, steady pendulum tick-tock, tiny brass gears turning, escapement clicking` | + +- **실행 한 번에 이벤트 하나.** 문이 열리고 걸쇠가 닫히는 소리는 그 순서로 설정한 길이 안에 들어갈 때만 한 프롬프트에 함께 씁니다. +- **대사 아래에 깔 거라면 `no music, no voices`.** 결과물이 깔끔하게 남습니다. +- **길이는 이벤트에 맞추세요.** 클릭은 1초면 충분하지만, 라이저는 컷 앞의 빈 공간을 채워야 합니다. +- **여러 번 들어 보세요.** 새 seed로 다시 실행하고 재질, 공간, 강도 순서로 하나씩 바꿉니다. + +## 길이, 제한, 크레딧 + +| 제한 | Pika Music | Pika SFX | +| --- | --- | --- | +| 프롬프트 | 1~4,000자 | 1~4,000자 | +| 가사 | 선택, 최대 8,000자 | 지원하지 않음 | +| 길이 | 10~360초, 기본 30 | 1~20초, 기본 5 | +| 커버 길이 | `source_audio`를 따르며 5~360초로 제한 | 지원하지 않음 | +| 출력 | 스테레오 트랙 하나 | 44.1 kHz 스테레오 효과음 하나 | + +두 노드 모두 생성된 오디오의 초 단위로 크레딧이 청구되므로, 10초 스팅은 3분 곡의 일부만큼만 듭니다. 현재 요율은 [파트너 노드 가격](/ko/tutorials/partner-nodes/pricing)을 참고하세요. + +`rewrite_prompt`가 꺼진 상태에서는 같은 프롬프트, 가사, 길이, seed로 같은 결과가 재현됩니다. `rewrite_prompt`를 켜면 생성 전에 모델이 프롬프트를 다시 쓰므로 seed만으로는 출력이 고정되지 않습니다. 재현 가능한 워크플로가 필요하면 꺼 두세요. + +## 출력과 저장 + +두 노드 모두 `AUDIO`를 출력하고, 템플릿은 `Save Audio (Advanced)`로 저장합니다. 파일명 접두사는 `audio/Pika_Music` 또는 `audio/Pika_SFX`, 형식은 `mp3`, 품질은 `V0`이며 `ComfyUI/output/audio/`에 저장됩니다. 마스터나 재편집용 무손실 파일이 필요하면 형식을 `flac`으로 바꾸세요. + +출력은 일반 오디오 스트림이라 그래프의 다른 노드와 그대로 조합됩니다: + +- `Create Video`에 연결해 프레임 묶음에 트랙을 붙일 수 있습니다 +- 임팩트마다 `Pika SFX`를 한 번씩 실행하고 편집기에서 겹치세요. 한 번 실행에 효과음 하나입니다 +- 타임라인에서 긴 곡을 자르는 대신 트랙 길이를 컷 길이에 맞춰 설정하세요 + +## 어떤 Pika 오디오 노드를 쓸까 + +| 노드 | 입력 | 출력 | 사용할 때 | +| --- | --- | --- | --- | +| **Pika Music** | 프롬프트, 가사, 선택적 원곡 | 10~360초의 곡 | 음악이 필요할 때: 베드, 징글, 노래, 커버 | +| **Pika SFX** | 프롬프트 | 1~20초 효과음 하나 | 설명에서 특정한 소리 하나를 얻고 싶을 때 | +| **Pika Soundtrack** | 비디오, 선택적 지시 | 클립 오디오를 동기화된 사운드트랙으로 교체 | 소리가 화면을 따라야 할 때, 최대 10분 | + +화면의 움직임에 맞춰 소리를 내려면 영상 쪽 경로를 씁니다: [Pika H3 Flex 페이지의 Pika Soundtrack](/ko/tutorials/partner-nodes/pika/pika-h3-flex). 설명이 아니라 클립을 읽습니다. + +## 관련 페이지 -각 실행은 하나의 모노 사운드 효과를 생성합니다. 연속된 사운드가 필요하면 사운드마다 노드를 한 번씩 실행하고 타임라인에 직접 배치하세요. \ No newline at end of file +- [Pika H3 Flex](/ko/tutorials/partner-nodes/pika/pika-h3-flex): 동기화된 오디오가 있는 비디오 생성 +- [Mirelo SFX 2.0](/ko/tutorials/partner-nodes/mirelo/mirelo-sfx): 클립에 맞춘 폴리와 텍스트에서 SFX +- [ACE-Step 1.5](/ko/tutorials/audio/ace-step/ace-step-v1-5): 로컬 GPU에서 실행되는 오픈 웨이트 음악 모델 +- [파트너 노드 가격](/ko/tutorials/partner-nodes/pricing) 및 [동시 실행 제한](/ko/tutorials/partner-nodes/concurrency-limits) diff --git a/ko/tutorials/partner-nodes/pika/pika-h3-flex.mdx b/ko/tutorials/partner-nodes/pika/pika-h3-flex.mdx index ad111db411..79a1330c92 100644 --- a/ko/tutorials/partner-nodes/pika/pika-h3-flex.mdx +++ b/ko/tutorials/partner-nodes/pika/pika-h3-flex.mdx @@ -68,6 +68,8 @@ Pika H3 Flex는 [MiniMax H3](/ko/tutorials/video/minimax/minimax-h3)를 Pika가 - **Pika SFX**: 텍스트 설명을 바탕으로 1~20초 길이의 사운드 효과를 생성합니다 - **Pika Music**: 텍스트 프롬프트와 선택적 가사로 10~360초 길이의 노래를 생성하거나, 원곡 길이 그대로 커버하며 5~360초로 제한됩니다 +텍스트로 구동되는 이 두 오디오 노드의 컨트롤과 프롬프트 작성법, 제한은 [Pika Music 및 Pika SFX](/ko/tutorials/partner-nodes/pika/pika-audio)를 참고하세요. + ## 프롬프트 형식 Pika H3 Flex 프롬프트는 MiniMax H3 텍스트 모드 구조를 따르며, 여기서 하나의 설명이 화면과 사운드를 모두 담습니다: diff --git a/tutorials/partner-nodes/pika/pika-audio.mdx b/tutorials/partner-nodes/pika/pika-audio.mdx index a3db3d2617..acd772769b 100644 --- a/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/tutorials/partner-nodes/pika/pika-audio.mdx @@ -1,22 +1,35 @@ --- -title: "Pika Music and Pika SFX in ComfyUI: songs and sound effects" -description: "Generate a full song from a style prompt plus optional lyrics, cover an existing track, or make a 1 to 20 second sound effect with the Pika audio nodes in ComfyUI." +title: "Pika Music and Pika SFX in ComfyUI: AI songs and sound effects" +description: "Generate a full song from a style prompt and lyrics, cover an existing track, or build 1 to 20 second sound effects with the Pika Music and Pika SFX nodes in ComfyUI." sidebarTitle: "Pika Music & SFX" --- import ReqHint from "/snippets/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; -Pika's audio nodes turn a text description into finished audio without a local GPU. **Pika Music** writes a song from a style prompt and optional lyrics, in a length you choose from 10 to 360 seconds, and it can also cover an uploaded song at that song's own length. **Pika SFX** returns a single sound effect of 1 to 20 seconds, which fits foley, UI blips and trailer hits. +Pika's audio nodes turn a written description into finished audio without a local GPU. **Pika Music** writes a song from a style prompt plus optional lyrics, at a length you choose from 10 to 360 seconds, and it can cover an uploaded track at that track's own length. **Pika SFX** returns one sound effect from 1 to 20 seconds, for foley, UI blips, risers and trailer hits. -Neither node downloads weights. Both run on your Comfy account and are billed in credits like the other partner nodes. +Neither node downloads weights. Both hand back a single `AUDIO` output, run on your Comfy account, and are billed in credits by the second of generated audio. + +**What these nodes are for** + +- A cue, jingle or instrumental bed cut to the exact length of a video edit +- A song from lyrics you already wrote, in a style you can change without touching the words +- A restyled version of a track you own: new genre, new tempo, the same song underneath +- Foley, impacts, whooshes, room tone and game event sounds for a finished picture -## Pika Music: Text to Music {#api_pika_music} +## Get started + +1. Update ComfyUI to 0.45.0 or later +2. Go to Template Library, search for `Pika Music` or `Pika SFX` +3. Write the prompt, set `duration`, then run the workflow -Write the arrangement, mood and instrumentation into the prompt. Add section-tagged lyrics with tags such as `[Verse]` and `[Chorus]` for a vocal track, or leave the lyrics field empty for an instrumental. +## Pika Music: text to music {#api_pika_music} + +Write the arrangement, the mood and the instrumentation into the prompt, then add lyrics if you want a vocal. The node returns one finished track, mixed by the model. Pika Music text to music workflow preview @@ -26,15 +39,23 @@ Write the arrangement, mood and instrumentation into the prompt. Add section-tag -1. Pick `text to music` in the `Pika Music` node and describe the track: genre, tempo, instruments and mood. -2. Set `duration` between 10 and 360 seconds. Turn on `rewrite_prompt` if you only have a short idea and want the model to expand it first. -3. Run the workflow. `Save Audio (Advanced)` writes an MP3 to `ComfyUI/output/audio/`. +1. Choose `text to music` in the `Pika Music` node. +2. Write the brief into `prompt`: genre and era, BPM and key, the instruments that carry the hook, the vocal character, and how the song moves from section to section. Up to 4,000 characters reach the model. +3. Add `lyrics` with section tags such as `[Verse]` and `[Chorus]`, or leave the field empty for an instrumental. Up to 8,000 characters. +4. Set `duration` between 10 and 360 seconds. Turn on `rewrite_prompt` only when your prompt is a few words and you want a model to expand it first. +5. Run the workflow. `Save Audio (Advanced)` writes an MP3 to `ComfyUI/output/audio/`. + +## Pika Music: cover a song {#api_pika_music_cover} + +Switch the same node's mode to `cover` to restyle a song you already have instead of writing a new one. Connect an audio file to `source_audio`, then write the brief as the arrangement you want instead: a different genre, a different tempo, a different band. -Switch the same node's mode to `cover` to restyle an existing song instead. Connect an audio file to `source_audio`, describe the new style, and the result keeps the source length, clamped to 5 to 360 seconds. +Cover ignores `duration`. The result keeps the length of the source, clamped to 5 to 360 seconds, so a three-minute upload comes back three minutes long. The `lyrics` input still applies, which means you can keep the original words or hand the model a new set for the same arrangement. -## Pika SFX: Text to Sound Effects {#api_pika_sfx} +A cover works from a mix, so separation is the model's job. A centered lead vocal and a simple arrangement survive a restyle more reliably than a dense master. -Name the material, the action, the space and the intensity in one line. A close, dry description reads better than a list of adjectives. +## Pika SFX: text to sound effects {#api_pika_sfx} + +Name the sound in one line and the node returns a single effect. Pika follows the brief without adding speech, music or background noise you did not ask for, so an effect drops into a timeline without a bed you have to strip out. Pika SFX text to sound effect workflow preview @@ -47,4 +68,136 @@ Name the material, the action, the space and the intensity in one line. A close, 1. Write the sound in the `Pika SFX` prompt, for example an old timber door creaking open on rusty hinges, then a heavy latch clicking shut. 2. Set `duration` between 1 and 20 seconds, then run the workflow. The MP3 lands in `ComfyUI/output/audio/`. -Each run produces one mono sound effect. For a sequence of sounds, run the node once per sound and place them on the timeline yourself. +Each run produces one effect at 44.1 kHz stereo. For a sequence of sounds, run the node once per sound and place them on the timeline yourself. Pika SFX does not write music or sung vocals; use Pika Music for that. + +## Node controls + +`Pika Music`, text to music mode: + +| Control | Values | Effect | +| --- | --- | --- | +| `prompt` | text, 1 to 4,000 characters | The brief: genre, BPM, instrumentation, vocal character, arrangement | +| `lyrics` | optional, up to 8,000 characters | The words to sing. Section tags set the structure. Empty means instrumental | +| `duration` | 10 to 360 seconds, default 30 | Length of the track | +| `rewrite_prompt` | off by default | Expands a short or vague prompt with a model before generating | +| `seed` | integer, default 42 | Change it to explore variations of the same brief | + +`Pika Music`, cover mode: + +| Control | Values | Effect | +| --- | --- | --- | +| `source_audio` | audio input | The song to cover. The output keeps its length, clamped to 5 to 360 seconds | +| `prompt` | text, 1 to 4,000 characters | The new arrangement to rebuild the song as | +| `lyrics` | optional, up to 8,000 characters | Replace the words, or keep the original ones | +| `rewrite_prompt` | off by default | Expands a short or vague prompt with a model before generating | +| `seed` | integer, default 42 | Change it to explore variations of the same cover | + +`Pika SFX`: + +| Control | Values | Effect | +| --- | --- | --- | +| `prompt` | text, 1 to 4,000 characters | The sound: material, action, space, perspective, intensity | +| `duration` | 1 to 20 seconds, default 5 | Length of the effect | +| `seed` | integer, default 42 | Change it to explore variations of the same brief | + +## Writing a music prompt + +Build the brief in this order: **genre and era, BPM and key, the instruments that carry the hook, vocal character, arrangement arc, mix notes**. The model reads all 4,000 characters, so detail earns its place; a stack of adjectives does not. + +The default prompt in the template shows the shape: + +```text +Slow-burning indie soul ballad at 72 BPM: warm Rhodes electric piano, brushed drums, upright bass and a soft horn section that swells in the chorus. Intimate late-night lounge mood, analog tape saturation, wide stereo reverb. +``` + +More starting points: + +- `Mainstream dance pop, four-on-the-floor beat at 124 BPM in A minor, bright female lead in an upper-mid register, sidechained warm pads, plucked synth hook, filtered verse lifting into a festival-sized chorus, glossy radio mix.` +- `Sparse solo grand piano at 68 BPM, felt hammers and close mic, rubato intro resolving into a steady left-hand arpeggio, no drums, no vocals.` +- `Cinematic orchestral build at 90 BPM, taiko and low-string ostinato, brass swell around twenty seconds, choir pad underneath, wide trailer mix, instrumental.` +- `Lofi hip hop at 82 BPM, dusty boom-bap drums, vinyl crackle, warm electric piano loop, upright bass, no vocals, loopable.` + +What changes the result: + +- **Ask for the arc, not just the palette.** "Filtered verse lifting into a full chorus" gives the track somewhere to go; a long instrument list on its own tends to come back as a static texture. +- **Say what you do not want.** `no vocals`, `no ad-libs`, `dry close-mic lead` are all read as direction. +- **Name the delivery.** Register, gender and how forward the vocal sits in the mix move the track more than one more instrument does. +- **Match the length to the job.** A 10-second sting and a 180-second cue want different briefs, so generate the length you need instead of cutting a long song down. +- **Iterate one dimension at a time.** Keep the seed while you edit the prompt, so the difference you hear comes from the change you made. + +## Lyrics and section tags + +`lyrics` is a separate input from the prompt, so the same words can come back as a ballad, a dance single or a rock performance when only the brief changes. + +- An empty `lyrics` field returns an instrumental on the same brief, with no vocal. +- Tags set the structure. They are read as phrasing and section boundaries rather than text to sing. `[Verse]`, `[Pre-Chorus]` and `[Chorus]` cover most songs; add `[Verse 2]`, `[Bridge]` or `[Final Chorus]` to name repeats and the ending. +- Keep one lyric line per line. Line breaks are what the model phrases against. + +```text +[Verse] +Streetlights hum on the empty avenue +I keep the radio low and the windows down +[Pre-Chorus] +Every red light is a chance to turn around +[Chorus] +I'm not lost, I'm just taking the long way home +``` + +## Writing a sound effect prompt + +One line, and four things in it: **what the material is, what happens to it, the space it happens in, and how close the mic is**. Intensity goes last. A close, dry description reads better than a list of adjectives. + +| Goal | Prompt | +| --- | --- | +| Door, foley | `Old timber door creaks open slowly on rusty hinges, then a heavy latch clicks shut, close perspective, dry, no music, no voices` | +| UI blip | `Short bright digital blip, one hundred milliseconds, clean tone, no reverb` | +| Trailer impact | `Deep sub-bass impact with a metal debris tail, close and heavy, wide stereo, cinematic` | +| Crowd ambience | `Reverberant stadium crowd cheering, chants and claps, steady bed with no music` | +| Riser | `Rising noise sweep with a pitch climb across eight seconds, tension building to a hard stop` | +| Mechanical detail | `Close-up foley inside an antique mechanical clock, steady pendulum tick-tock, tiny brass gears turning, escapement clicking` | + +- **One event per run.** A door opening and a latch closing share one prompt only when they land in that order inside the duration you set. +- **Say `no music, no voices`** when the effect has to sit under dialogue. It keeps the render clean. +- **Match duration to the event.** A click is one second; a riser has to fill the gap in front of the cut. +- **Audition variants.** Re-run with a new seed, then change one dimension: material first, then space, then intensity. + +## Duration, limits and credits + +| Limit | Pika Music | Pika SFX | +| --- | --- | --- | +| Prompt | 1 to 4,000 characters | 1 to 4,000 characters | +| Lyrics | up to 8,000 characters, optional | not supported | +| Duration | 10 to 360 seconds, default 30 | 1 to 20 seconds, default 5 | +| Cover length | follows `source_audio`, clamped to 5 to 360 seconds | not supported | +| Output | one track, stereo | one effect, 44.1 kHz stereo | + +Both nodes bill in credits by the second of generated audio, so a 10-second sting costs a fraction of a 3-minute song. See [Partner node pricing](/tutorials/partner-nodes/pricing) for the current rate. + +The same prompt, lyrics, duration and seed with `rewrite_prompt` off reproduce the same result. With `rewrite_prompt` on, a model rewrites the brief before generation, so the seed alone no longer pins the output. Leave it off when you want a reproducible workflow. + +## Output and saving + +Both nodes output `AUDIO`, and the templates save it with `Save Audio (Advanced)`: prefix `audio/Pika_Music` or `audio/Pika_SFX`, format `mp3` at `V0`, which lands in `ComfyUI/output/audio/`. Switch the format to `flac` when you want a lossless file to master or re-cut. + +The output is an ordinary audio stream, so it composes with the rest of the graph: + +- Connect it to `Create Video` to attach the track to a set of frames +- Run `Pika SFX` once per hit and layer the results in an editor, since one run gives one effect +- Set the track duration to the length of the cut instead of trimming a longer song in the timeline + +## Which Pika audio node to use + +| Node | Input | Output | Use it when | +| --- | --- | --- | --- | +| **Pika Music** | prompt, lyrics, optional source song | A full song, 10 to 360 seconds | You need music: a bed, a jingle, a song, or a cover | +| **Pika SFX** | prompt | One effect, 1 to 20 seconds | You need one specific sound from a description | +| **Pika Soundtrack** | a video, optional instruction | The clip's audio replaced with a synchronized soundtrack | The sound has to follow the picture, up to 10 minutes | + +For an effect timed to action on screen, the video path is [Pika Soundtrack on the Pika H3 Flex page](/tutorials/partner-nodes/pika/pika-h3-flex), which reads the clip instead of a description. + +## Related pages + +- [Pika H3 Flex](/tutorials/partner-nodes/pika/pika-h3-flex): video generation with synchronized audio +- [Mirelo SFX 2.0](/tutorials/partner-nodes/mirelo/mirelo-sfx): foley timed to a clip, and text to SFX +- [ACE-Step 1.5](/tutorials/audio/ace-step/ace-step-v1-5): an open-weights music model that runs locally on your own GPU +- [Partner node pricing](/tutorials/partner-nodes/pricing) and [concurrency limits](/tutorials/partner-nodes/concurrency-limits) diff --git a/tutorials/partner-nodes/pika/pika-h3-flex.mdx b/tutorials/partner-nodes/pika/pika-h3-flex.mdx index 108754a9d8..c26a177ebf 100644 --- a/tutorials/partner-nodes/pika/pika-h3-flex.mdx +++ b/tutorials/partner-nodes/pika/pika-h3-flex.mdx @@ -57,6 +57,8 @@ Three audio nodes run on the same Pika API and are billed per second of generate - **Pika SFX**: generates a sound effect from a text description, 1 to 20 seconds long - **Pika Music**: generates a song from a text prompt and optional lyrics, 10 to 360 seconds, or covers a source song at its own length, clamped to 5 to 360 seconds +For node controls, prompt patterns and limits on the two text-driven audio nodes, see [Pika Music and Pika SFX](/tutorials/partner-nodes/pika/pika-audio). + ## Prompt format Pika H3 Flex prompts follow the MiniMax H3 text mode structure, where one description carries both the picture and the sound: diff --git a/zh/tutorials/partner-nodes/pika/pika-audio.mdx b/zh/tutorials/partner-nodes/pika/pika-audio.mdx index 42608cce33..7b48d7946e 100644 --- a/zh/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/zh/tutorials/partner-nodes/pika/pika-audio.mdx @@ -1,24 +1,51 @@ --- -title: "ComfyUI 中的 Pika Music 与 Pika SFX:歌曲与音效" -description: "使用 ComfyUI 中的 Pika 音频节点,从风格提示词加上可选歌词生成一首完整歌曲、翻唱已有曲目,或制作 1 到 20 秒的音效。" +title: "ComfyUI 中的 Pika Music 与 Pika SFX:AI 歌曲与音效" +description: "在 ComfyUI 中用 Pika Music 与 Pika SFX 节点生成完整歌曲:风格提示词加可选歌词、翻唱已有曲目,或制作 1 到 20 秒的音效。" sidebarTitle: "Pika Music 与 SFX" -translationSourceHash: 60f1ba51 +translationSourceHash: c04fcf5c translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx +translationBlockHashes: + "_intro": dc3cd241 + "Get started": fd80ca7d + "Pika Music: text to music {#api_pika_music}": ae1738bd + "Pika Music: cover a song {#api_pika_music_cover}": 0793be8c + "Pika SFX: text to sound effects {#api_pika_sfx}": 81088686 + "Node controls": 2c4dd7da + "Writing a music prompt": a24287f2 + "Lyrics and section tags": af8f7f9e + "Writing a sound effect prompt": 20f01804 + "Duration, limits and credits": 763b11c7 + "Output and saving": 4abf0bf7 + "Which Pika audio node to use": 8b598461 + "Related pages": 96e4c1fe --- import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; -Pika 的音频节点无需本地 GPU 就能把文字描述变成成品音频。**Pika Music** 根据风格提示词和可选歌词创作歌曲,时长由你在 10 到 360 秒之间自行选择,它还能以歌曲自身的长度翻唱上传的歌曲。**Pika SFX** 返回一段 1 到 20 秒的单一音效,适用于拟音、UI 提示音和预告片冲击音。 +Pika 的音频节点无需本地 GPU 就能把文字描述变成成品音频。**Pika Music** 根据风格提示词和可选歌词创作歌曲,时长由你在 10 到 360 秒之间自行选择,它还能以歌曲自身的长度翻唱上传的曲目。**Pika SFX** 返回一段 1 到 20 秒的单一音效,适用于拟音、UI 提示音、上升音和预告片冲击音。 -两个节点都不下载权重。它们都运行在你的 Comfy 账户上,与其他合作节点一样按积分计费。 +两个节点都不下载权重。它们都只输出一个 `AUDIO`,运行在你的 Comfy 账户上,并按生成音频的秒数用积分计费。 + +**这些节点能做什么** + +- 为一段视频剪辑配乐、做短标音乐或纯器乐铺底,长度与剪辑完全对齐 +- 把已经写好的歌词唱成歌,改风格时不用动一个字 +- 重新演绎你已有的曲目:换流派、换速度,底下还是同一首歌 +- 为成片制作拟音、冲击音、呼啸声、房间底噪和游戏事件音效 +## 快速开始 + +1. 将 ComfyUI 更新到 0.45.0 或更高版本 +2. 打开模板库,搜索 `Pika Music` 或 `Pika SFX` +3. 写好提示词,设置 `duration`,然后运行工作流 + ## Pika Music:文本生成音乐 {#api_pika_music} -把编曲、情绪和配器写进提示词。如果要有人声,就添加带分段标签的歌词,标签例如 `[Verse]` 和 `[Chorus]`;如果要纯音乐,就把歌词字段留空。 +把编曲、情绪和配器写进提示词,需要人声时再添加歌词。节点返回一首完整成品,混音由模型完成。 Pika Music 文本生成音乐工作流预览 @@ -28,15 +55,23 @@ Pika 的音频节点无需本地 GPU 就能把文字描述变成成品音频。* -1. 在 `Pika Music` 节点中选择 `text to music`,并描述曲目:流派、速度、乐器和情绪。 -2. 将 `duration` 设为 10 到 360 秒之间。如果你只有一个简短的想法,希望模型先把它展开,请打开 `rewrite_prompt`。 -3. 运行工作流。`Save Audio (Advanced)` 会把 MP3 写入 `ComfyUI/output/audio/`。 +1. 在 `Pika Music` 节点中选择 `text to music`。 +2. 把创作方向写进 `prompt`:流派与年代、BPM 与调性、承载记忆点的乐器、人声特征,以及歌曲在各段之间如何推进。最多 4,000 个字符会送进模型。 +3. 添加带分段标签的 `lyrics`,标签如 `[Verse]`、`[Chorus]`;要做纯音乐就把歌词字段留空。上限 8,000 个字符。 +4. 将 `duration` 设为 10 到 360 秒之间。只有当你只写了几个词、希望模型先扩写时,才打开 `rewrite_prompt`。 +5. 运行工作流。`Save Audio (Advanced)` 会把 MP3 写入 `ComfyUI/output/audio/`。 + +## Pika Music:翻唱一首歌 {#api_pika_music_cover} + +把同一个节点的模式切换为 `cover`,就可以重新演绎你已有的歌曲,而不是从零写一首。把音频文件连接到 `source_audio`,然后把你想要的编曲写进提示词:换流派、换速度、换一支乐队。 -把同一个节点的模式切换为 `cover`,即可改为重新演绎一首已有的歌曲。将音频文件连接到 `source_audio`,描述新的风格,结果会沿用来源长度,并限制在 5 到 360 秒之间。 +翻唱会忽略 `duration`。输出沿用来源长度,并限制在 5 到 360 秒之间,所以三分钟的素材回来还是三分钟。`lyrics` 仍然生效,也就是说你可以保留原词,也可以给同一套编曲换一版新词。 + +翻唱是从混音成品出发的,所以分离人声和伴奏是模型的工作。人声居中、编曲简单的素材,比厚重的母带更容易在重新演绎后站得住。 ## Pika SFX:文本生成音效 {#api_pika_sfx} -用一句话写明材质、动作、空间和强度。贴近、干练的描述比堆砌形容词效果更好。 +用一句话写明声音,节点返回一段音效。Pika 会照着描述生成,不会添加你没有要求的语音、音乐或背景噪声,所以音效可以直接落到时间线上,不必再剥一层底。 Pika SFX 文本生成音效工作流预览 @@ -46,7 +81,139 @@ Pika 的音频节点无需本地 GPU 就能把文字描述变成成品音频。* -1. 在 `Pika SFX` 的提示词中写出声音,例如:一扇老旧木门在生锈的合页上吱呀打开,随后一声沉重的门闩咔哒合上。 +1. 在 `Pika SFX` 的提示词里写出声音,例如一扇老旧木门在生锈的合页上缓慢吱呀打开,随后一声沉重的门闩咔哒合上。 2. 将 `duration` 设为 1 到 20 秒之间,然后运行工作流。MP3 会落在 `ComfyUI/output/audio/`。 -每次运行生成一段单声道音效。如果需要一组连续的声音,请为每个声音各运行一次节点,并自行把它们排到时间线上。 \ No newline at end of file +每次运行生成一段 44.1 kHz 立体声音效。如果需要一组连续的声音,请为每个声音各运行一次节点,并自行把它们排到时间线上。Pika SFX 不写音乐、也不唱歌,那部分交给 Pika Music。 + +## 节点控件 + +`Pika Music`,text to music 模式: + +| 控件 | 取值 | 作用 | +| --- | --- | --- | +| `prompt` | 文本,1 到 4,000 个字符 | 创作方向:流派、BPM、配器、人声特征、编曲 | +| `lyrics` | 可选,最多 8,000 个字符 | 要唱的歌词。分段标签决定结构,留空即纯音乐 | +| `duration` | 10 到 360 秒,默认 30 | 曲目长度 | +| `rewrite_prompt` | 默认关闭 | 生成前用模型扩写简短或含糊的提示词 | +| `seed` | 整数,默认 42 | 换一个值可以探索同一方向下的变化 | + +`Pika Music`,cover 模式: + +| 控件 | 取值 | 作用 | +| --- | --- | --- | +| `source_audio` | 音频输入 | 要翻唱的歌曲。输出沿用它的长度,限制在 5 到 360 秒 | +| `prompt` | 文本,1 到 4,000 个字符 | 要把这首歌重建成什么样的新编曲 | +| `lyrics` | 可选,最多 8,000 个字符 | 换掉歌词,或保留原词 | +| `rewrite_prompt` | 默认关闭 | 生成前用模型扩写简短或含糊的提示词 | +| `seed` | 整数,默认 42 | 换一个值可以探索同一次翻唱的变化 | + +`Pika SFX`: + +| 控件 | 取值 | 作用 | +| --- | --- | --- | +| `prompt` | 文本,1 到 4,000 个字符 | 声音本身:材质、动作、空间、听感距离、强度 | +| `duration` | 1 到 20 秒,默认 5 | 音效长度 | +| `seed` | 整数,默认 42 | 换一个值可以探索同一方向下的变化 | + +## 如何写音乐提示词 + +按这个顺序搭建方向:**流派与年代、BPM 与调性、承载记忆点的乐器、人声特征、编曲推进、混音取向**。4,000 个字符模型都会读,所以细节有价值,堆砌形容词没有。 + +模板里的默认提示词就是这种写法: + +```text +Slow-burning indie soul ballad at 72 BPM: warm Rhodes electric piano, brushed drums, upright bass and a soft horn section that swells in the chorus. Intimate late-night lounge mood, analog tape saturation, wide stereo reverb. +``` + +更多起点: + +- `Mainstream dance pop, four-on-the-floor beat at 124 BPM in A minor, bright female lead in an upper-mid register, sidechained warm pads, plucked synth hook, filtered verse lifting into a festival-sized chorus, glossy radio mix.` +- `Sparse solo grand piano at 68 BPM, felt hammers and close mic, rubato intro resolving into a steady left-hand arpeggio, no drums, no vocals.` +- `Cinematic orchestral build at 90 BPM, taiko and low-string ostinato, brass swell around twenty seconds, choir pad underneath, wide trailer mix, instrumental.` +- `Lofi hip hop at 82 BPM, dusty boom-bap drums, vinyl crackle, warm electric piano loop, upright bass, no vocals, loopable.` + +会改变结果的几点: + +- **要写推进,不只写音色。** `Filtered verse lifting into a full chorus` 给曲目留下了走向;只列一长串乐器,返回的往往是静态的音色块。 +- **把不要的东西说出来。** `no vocals`、`no ad-libs`、`dry close-mic lead` 都会被当作方向。 +- **写清演唱方式。** 音区、性别、人声在混音里的位置,比再加一件乐器更能改变整首歌。 +- **长度按用途定。** 10 秒的短标和 180 秒的配乐需要不同的方向,该生成多长就生成多长,而不是先做一首长歌再剪。 +- **一次只改一个维度。** 改提示词时保持同一个 seed,你听到的差别才会来自你这次的改动。 + +## 歌词与分段标签 + +`lyrics` 是与提示词分开的输入,所以同一套词只换提示词,就能变成抒情曲、舞曲单曲或摇滚现场。 + +- `lyrics` 留空时,返回同一方向下的纯音乐,没有人声。 +- 标签用来定结构。它们被当作分段边界和断句依据,而不是要唱出来的文字。`[Verse]`、`[Pre-Chorus]`、`[Chorus]` 覆盖大部分歌曲;加上 `[Verse 2]`、`[Bridge]` 或 `[Final Chorus]` 可以标出重复段和结尾。 +- 一行一句歌词。模型是按换行来断句的。 + +```text +[Verse] +Streetlights hum on the empty avenue +I keep the radio low and the windows down +[Pre-Chorus] +Every red light is a chance to turn around +[Chorus] +I'm not lost, I'm just taking the long way home +``` + +## 如何写音效提示词 + +一行,写进四件事:**材质是什么、对它做了什么、发生在什么空间里、麦克风有多近**。强度放在最后。贴近、干练的描述比堆形容词更好用。 + +| 目标 | 提示词 | +| --- | --- | +| 拟音:门 | `Old timber door creaks open slowly on rusty hinges, then a heavy latch clicks shut, close perspective, dry, no music, no voices` | +| UI 提示音 | `Short bright digital blip, one hundred milliseconds, clean tone, no reverb` | +| 预告片冲击音 | `Deep sub-bass impact with a metal debris tail, close and heavy, wide stereo, cinematic` | +| 人群环境声 | `Reverberant stadium crowd cheering, chants and claps, steady bed with no music` | +| 上升音 | `Rising noise sweep with a pitch climb across eight seconds, tension building to a hard stop` | +| 机械细节 | `Close-up foley inside an antique mechanical clock, steady pendulum tick-tock, tiny brass gears turning, escapement clicking` | + +- **一次一个事件。** 只有当开门和落闩确实按这个顺序发生、并且都在你设的时长之内时,才写进同一条提示词。 +- **要压在台词下就说 `no music, no voices`。** 这样出来的素材更干净。 +- **时长跟着事件走。** 咔哒声一秒就够,上升音却要填满剪辑前的空档。 +- **多试几版。** 换个 seed 重跑,然后一次只改一个维度:先材质,再空间,最后强度。 + +## 时长、限制与计费 + +| 限制项 | Pika Music | Pika SFX | +| --- | --- | --- | +| 提示词 | 1 到 4,000 个字符 | 1 到 4,000 个字符 | +| 歌词 | 可选,最多 8,000 个字符 | 不支持 | +| 时长 | 10 到 360 秒,默认 30 | 1 到 20 秒,默认 5 | +| 翻唱长度 | 跟随 `source_audio`,限制在 5 到 360 秒 | 不支持 | +| 输出 | 一首立体声曲目 | 一段 44.1 kHz 立体声音效 | + +两个节点都按生成音频的秒数计费,所以 10 秒的短标只花三分钟歌曲的一小部分。当前费率见 [合作伙伴节点定价](/zh/tutorials/partner-nodes/pricing)。 + +在 `rewrite_prompt` 关闭时,相同的提示词、歌词、时长和 seed 会复现相同的结果。打开 `rewrite_prompt` 后,生成前会先由模型改写提示词,这时只靠 seed 已经无法锁定输出;需要可复现的工作流就把它关掉。 + +## 输出与保存 + +两个节点都输出 `AUDIO`,模板用 `Save Audio (Advanced)` 保存:文件名前缀 `audio/Pika_Music` 或 `audio/Pika_SFX`,格式 `mp3`、质量 `V0`,落在 `ComfyUI/output/audio/`。需要无损母带或再剪辑时,把格式换成 `flac`。 + +输出就是普通的音频流,所以可以和流程里其他节点自由组合: + +- 连接到 `Create Video`,把音轨配到一组画面上 +- 每个打击声各跑一次 `Pika SFX`,再在剪辑软件里叠起来,因为一次只出一段音效 +- 曲目长度按剪辑长度来设,而不是在时间线上把长歌剪短 + +## 该用哪个 Pika 音频节点 + +| 节点 | 输入 | 输出 | 适用场景 | +| --- | --- | --- | --- | +| **Pika Music** | 提示词、歌词、可选源曲 | 一首完整歌曲,10 到 360 秒 | 你需要音乐:铺底、短标、歌曲或翻唱 | +| **Pika SFX** | 提示词 | 一段音效,1 到 20 秒 | 你需要根据描述得到某个具体声音 | +| **Pika Soundtrack** | 一段视频、可选指令 | 把片段音频替换为同步配乐 | 声音必须跟着画面走,最长 10 分钟 | + +要让音效卡着画面上的动作,走视频那条路:[Pika H3 Flex 页面上的 Pika Soundtrack](/zh/tutorials/partner-nodes/pika/pika-h3-flex),它读的是片段而不是描述。 + +## 相关页面 + +- [Pika H3 Flex](/zh/tutorials/partner-nodes/pika/pika-h3-flex):带同步音频的视频生成 +- [Mirelo SFX 2.0](/zh/tutorials/partner-nodes/mirelo/mirelo-sfx):跟着片段卡点的拟音,以及文本生成音效 +- [ACE-Step 1.5](/zh/tutorials/audio/ace-step/ace-step-v1-5):可在本地 GPU 上运行的开放权重音乐模型 +- [合作伙伴节点定价](/zh/tutorials/partner-nodes/pricing) 与 [并发限制](/zh/tutorials/partner-nodes/concurrency-limits) diff --git a/zh/tutorials/partner-nodes/pika/pika-h3-flex.mdx b/zh/tutorials/partner-nodes/pika/pika-h3-flex.mdx index 027defc527..226781cc3f 100644 --- a/zh/tutorials/partner-nodes/pika/pika-h3-flex.mdx +++ b/zh/tutorials/partner-nodes/pika/pika-h3-flex.mdx @@ -68,6 +68,8 @@ Pika H3 Flex 是 Pika 基于 [MiniMax H3](/zh/tutorials/video/minimax/minimax-h3 - **Pika SFX**:根据文字描述生成音效,时长为 1 至 20 秒 - **Pika Music**:根据文本提示词和可选歌词生成歌曲,时长为 10 至 360 秒,或按原曲长度制作翻唱(时长限制在 5 至 360 秒) +这两个文本驱动的音频节点的控件、提示词写法和限制,见 [Pika Music 与 Pika SFX](/zh/tutorials/partner-nodes/pika/pika-audio)。 + ## 提示词格式 Pika H3 Flex 的提示词遵循 MiniMax H3 文本模式的结构,即由一段描述同时承载画面与声音: From 65b6971f6b030878ff63f0580a6d66d6858fb25e Mon Sep 17 00:00:00 2001 From: "Daxiong (Lin)" Date: Fri, 13 Nov 2026 18:43:15 +0800 Subject: [PATCH 3/3] docs(pika): fold the node controls into the workflow steps --- .../partner-nodes/pika/pika-audio.mdx | 45 +++---------------- .../partner-nodes/pika/pika-audio.mdx | 45 +++---------------- .../partner-nodes/pika/pika-audio.mdx | 45 +++---------------- tutorials/partner-nodes/pika/pika-audio.mdx | 36 ++------------- .../partner-nodes/pika/pika-audio.mdx | 45 +++---------------- 5 files changed, 31 insertions(+), 185 deletions(-) diff --git a/es/tutorials/partner-nodes/pika/pika-audio.mdx b/es/tutorials/partner-nodes/pika/pika-audio.mdx index 35f757ea7c..7877ad6c92 100644 --- a/es/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/es/tutorials/partner-nodes/pika/pika-audio.mdx @@ -2,19 +2,18 @@ title: "Pika Music y Pika SFX en ComfyUI: canciones y efectos de sonido con IA" description: "Genera una canción completa desde un prompt de estilo y letras opcionales, versiona una pista o crea efectos de sonido de 1 a 20 segundos con los nodos Pika Music y Pika SFX de ComfyUI." sidebarTitle: "Pika Music y SFX" -translationSourceHash: c04fcf5c +translationSourceHash: ebd0cf37 translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx translationBlockHashes: "_intro": dc3cd241 "Get started": fd80ca7d - "Pika Music: text to music {#api_pika_music}": ae1738bd + "Pika Music: text to music {#api_pika_music}": b6238bbc "Pika Music: cover a song {#api_pika_music_cover}": 0793be8c - "Pika SFX: text to sound effects {#api_pika_sfx}": 81088686 - "Node controls": 2c4dd7da + "Pika SFX: text to sound effects {#api_pika_sfx}": d4966763 "Writing a music prompt": a24287f2 "Lyrics and section tags": af8f7f9e "Writing a sound effect prompt": 20f01804 - "Duration, limits and credits": 763b11c7 + "Duration, limits and credits": 7190bc82 "Output and saving": 4abf0bf7 "Which Pika audio node to use": 8b598461 "Related pages": 96e4c1fe @@ -58,7 +57,7 @@ Escribe los arreglos, el ambiente y la instrumentación en el prompt, y añade l 1. Elige `text to music` en el nodo `Pika Music`. 2. Escribe el brief en `prompt`: género y época, BPM y tonalidad, los instrumentos que llevan el gancho, el carácter de la voz y cómo avanza la canción de una sección a otra. Al modelo le llegan hasta 4.000 caracteres. 3. Añade `lyrics` con etiquetas de sección como `[Verse]` y `[Chorus]`, o deja el campo vacío para una instrumental. Hasta 8.000 caracteres. -4. Ajusta `duration` entre 10 y 360 segundos. Activa `rewrite_prompt` solo cuando tu prompt sean unas pocas palabras y quieras que un modelo lo expanda antes. +4. Ajusta `duration` entre 10 y 360 segundos (30 por defecto). Activa `rewrite_prompt` solo cuando tu prompt sean unas pocas palabras y quieras que un modelo lo expanda antes. 5. Ejecuta el flujo de trabajo. `Save Audio (Advanced)` escribe un MP3 en `ComfyUI/output/audio/`. ## Pika Music: versionar una canción {#api_pika_music_cover} @@ -82,40 +81,10 @@ Nombra el sonido en una línea y el nodo devuelve un único efecto. Pika sigue e 1. Escribe el sonido en el prompt de `Pika SFX`, por ejemplo una vieja puerta de madera que se abre chirriando sobre bisagras oxidadas y luego un pestillo pesado que se cierra de golpe. -2. Ajusta `duration` entre 1 y 20 segundos y ejecuta el flujo de trabajo. El MP3 queda en `ComfyUI/output/audio/`. +2. Ajusta `duration` entre 1 y 20 segundos (5 por defecto) y ejecuta el flujo de trabajo. El MP3 queda en `ComfyUI/output/audio/`. Cada ejecución produce un efecto en estéreo a 44,1 kHz. Para una secuencia de sonidos, ejecuta el nodo una vez por cada sonido y colócalos tú mismo en la línea de tiempo. Pika SFX no escribe música ni voces cantadas; para eso está Pika Music. -## Controles del nodo - -`Pika Music`, modo text to music: - -| Control | Valores | Efecto | -| --- | --- | --- | -| `prompt` | texto, de 1 a 4.000 caracteres | El brief: género, BPM, instrumentación, carácter de la voz, arreglo | -| `lyrics` | opcional, hasta 8.000 caracteres | La letra que se canta. Las etiquetas de sección fijan la estructura. Vacío significa instrumental | -| `duration` | de 10 a 360 segundos, por defecto 30 | Duración de la pista | -| `rewrite_prompt` | desactivado por defecto | Expande con un modelo un prompt corto o vago antes de generar | -| `seed` | entero, por defecto 42 | Cámbialo para explorar variaciones del mismo brief | - -`Pika Music`, modo cover: - -| Control | Valores | Efecto | -| --- | --- | --- | -| `source_audio` | entrada de audio | La canción a versionar. La salida conserva su duración, ajustada a 5–360 segundos | -| `prompt` | texto, de 1 a 4.000 caracteres | El nuevo arreglo con el que se reconstruye la canción | -| `lyrics` | opcional, hasta 8.000 caracteres | Sustituye la letra o mantén la original | -| `rewrite_prompt` | desactivado por defecto | Expande con un modelo un prompt corto o vago antes de generar | -| `seed` | entero, por defecto 42 | Cámbialo para explorar variaciones del mismo cover | - -`Pika SFX`: - -| Control | Valores | Efecto | -| --- | --- | --- | -| `prompt` | texto, de 1 a 4.000 caracteres | El sonido: material, acción, espacio, perspectiva, intensidad | -| `duration` | de 1 a 20 segundos, por defecto 5 | Duración del efecto | -| `seed` | entero, por defecto 42 | Cámbialo para explorar variaciones del mismo brief | - ## Cómo escribir un prompt de música Construye el brief en este orden: **género y época, BPM y tonalidad, los instrumentos que llevan el gancho, carácter de la voz, desarrollo del arreglo, notas de mezcla**. El modelo lee los 4.000 caracteres, así que el detalle concreto aporta y una pila de adjetivos no. @@ -189,7 +158,7 @@ Una línea con cuatro cosas: **cuál es el material, qué le ocurre, en qué esp Ambos nodos se facturan en créditos por segundo de audio generado, así que un sting de 10 segundos cuesta una fracción de una canción de 3 minutos. Consulta [Precios de los nodos asociados](/es/tutorials/partner-nodes/pricing) para la tarifa actual. -Con `rewrite_prompt` desactivado, el mismo prompt, letras, duración y seed reproducen el mismo resultado. Con `rewrite_prompt` activado, un modelo reescribe el brief antes de generar, así que el seed por sí solo ya no fija la salida. Déjalo desactivado cuando quieras un flujo reproducible. +Con `rewrite_prompt` desactivado, el mismo prompt, letras, duración y seed reproducen el mismo resultado. Con `rewrite_prompt` activado, un modelo reescribe el brief antes de generar, así que el seed por sí solo ya no fija la salida. Déjalo desactivado cuando quieras un flujo reproducible. Ambos nodos parten de `seed` 42: mantenlo mientras editas el prompt y cámbialo solo para escuchar otra toma del mismo brief. ## Salida y guardado diff --git a/ja/tutorials/partner-nodes/pika/pika-audio.mdx b/ja/tutorials/partner-nodes/pika/pika-audio.mdx index c142e00a4c..8e72dfa167 100644 --- a/ja/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/ja/tutorials/partner-nodes/pika/pika-audio.mdx @@ -2,19 +2,18 @@ title: "ComfyUI の Pika Music と Pika SFX: AI 楽曲とサウンドエフェクト" description: "ComfyUI の Pika Music と Pika SFX ノードで、スタイルプロンプトと歌詞から楽曲全体を生成し、既存のトラックをカバーし、1〜20 秒のサウンドエフェクトを作成します。" sidebarTitle: "Pika Music & SFX" -translationSourceHash: c04fcf5c +translationSourceHash: ebd0cf37 translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx translationBlockHashes: "_intro": dc3cd241 "Get started": fd80ca7d - "Pika Music: text to music {#api_pika_music}": ae1738bd + "Pika Music: text to music {#api_pika_music}": b6238bbc "Pika Music: cover a song {#api_pika_music_cover}": 0793be8c - "Pika SFX: text to sound effects {#api_pika_sfx}": 81088686 - "Node controls": 2c4dd7da + "Pika SFX: text to sound effects {#api_pika_sfx}": d4966763 "Writing a music prompt": a24287f2 "Lyrics and section tags": af8f7f9e "Writing a sound effect prompt": 20f01804 - "Duration, limits and credits": 763b11c7 + "Duration, limits and credits": 7190bc82 "Output and saving": 4abf0bf7 "Which Pika audio node to use": 8b598461 "Related pages": 96e4c1fe @@ -58,7 +57,7 @@ Pika のオーディオノードは、ローカル GPU なしでテキストの 1. `Pika Music` ノードで `text to music` を選択します。 2. `prompt` に方向性を書きます。ジャンルと年代、BPM とキー、フックを担う楽器、ボーカルの性格、そして曲がセクションごとにどう進むか。最大 4,000 文字がモデルに届きます。 3. `[Verse]` や `[Chorus]` などのセクションタグ付きの `lyrics` を追加します。インストゥルメンタルにする場合は空のままにします。上限は 8,000 文字です。 -4. `duration` を 10〜360 秒の範囲で設定します。`rewrite_prompt` は、プロンプトが数語だけで、まずモデルに展開してほしいときだけオンにします。 +4. `duration` を 10〜360 秒の範囲(既定 30)で設定します。`rewrite_prompt` は、プロンプトが数語だけで、まずモデルに展開してほしいときだけオンにします。 5. ワークフローを実行します。`Save Audio (Advanced)` が MP3 を `ComfyUI/output/audio/` に書き出します。 ## Pika Music: 曲をカバーする {#api_pika_music_cover} @@ -82,40 +81,10 @@ Pika のオーディオノードは、ローカル GPU なしでテキストの 1. `Pika SFX` のプロンプトにサウンドを書きます。たとえば、古い木の扉が錆びた蝶番でゆっくり軋みながら開き、その後で重い掛け金がカチッと閉まる様子です。 -2. `duration` を 1〜20 秒の範囲で設定し、ワークフローを実行します。MP3 は `ComfyUI/output/audio/` に出力されます。 +2. `duration` を 1〜20 秒の範囲(既定 5)で設定し、ワークフローを実行します。MP3 は `ComfyUI/output/audio/` に出力されます。 1 回の実行で 44.1 kHz ステレオのエフェクトが 1 本生成されます。連続したサウンドが必要な場合は、サウンドごとにノードを 1 回実行し、タイムライン上に自分で配置してください。Pika SFX は音楽や歌唱を生成しません。それは Pika Music の担当です。 -## ノードのコントロール - -`Pika Music`、text to music モード: - -| コントロール | 値 | 効果 | -| --- | --- | --- | -| `prompt` | テキスト、1〜4,000 文字 | 方向性: ジャンル、BPM、楽器編成、ボーカルの性格、アレンジ | -| `lyrics` | 任意、最大 8,000 文字 | 歌う歌詞。セクションタグが構造を決めます。空ならインストゥルメンタル | -| `duration` | 10〜360 秒、既定 30 | トラックの長さ | -| `rewrite_prompt` | 既定はオフ | 生成前に短い・曖昧なプロンプトをモデルで展開します | -| `seed` | 整数、既定 42 | 同じ方向性のバリエーションを試すときに変えます | - -`Pika Music`、cover モード: - -| コントロール | 値 | 効果 | -| --- | --- | --- | -| `source_audio` | オーディオ入力 | カバーする曲。出力はその長さを保ち、5〜360 秒に制限されます | -| `prompt` | テキスト、1〜4,000 文字 | その曲をどんな新しいアレンジに作り直すか | -| `lyrics` | 任意、最大 8,000 文字 | 歌詞を差し替えるか、元の詞を残すか | -| `rewrite_prompt` | 既定はオフ | 生成前に短い・曖昧なプロンプトをモデルで展開します | -| `seed` | 整数、既定 42 | 同じカバーのバリエーションを試すときに変えます | - -`Pika SFX`: - -| コントロール | 値 | 効果 | -| --- | --- | --- | -| `prompt` | テキスト、1〜4,000 文字 | サウンドそのもの: 素材、動作、空間、マイクの距離感、強度 | -| `duration` | 1〜20 秒、既定 5 | エフェクトの長さ | -| `seed` | 整数、既定 42 | 同じ方向性のバリエーションを試すときに変えます | - ## 楽曲プロンプトの書き方 **ジャンルと年代、BPM とキー、フックを担う楽器、ボーカルの性格、アレンジの展開、ミックスの方向性** の順で組み立てます。4,000 文字すべてがモデルに届くので、具体的な情報は効きますが、形容詞を積み重ねても効きません。 @@ -189,7 +158,7 @@ I'm not lost, I'm just taking the long way home どちらのノードも生成されたオーディオの秒数でクレジット課金されるため、10 秒のステイングは 3 分の曲のごく一部の費用で済みます。現在のレートは [パートナーノードの料金](/ja/tutorials/partner-nodes/pricing) を参照してください。 -`rewrite_prompt` をオフにした状態では、同じプロンプト、歌詞、長さ、seed で同じ結果が再現されます。`rewrite_prompt` をオンにすると生成前にモデルがプロンプトを書き換えるため、seed だけでは出力を固定できません。再現性のあるワークフローが必要なときはオフのままにしてください。 +`rewrite_prompt` をオフにした状態では、同じプロンプト、歌詞、長さ、seed で同じ結果が再現されます。`rewrite_prompt` をオンにすると生成前にモデルがプロンプトを書き換えるため、seed だけでは出力を固定できません。再現性のあるワークフローが必要なときはオフのままにしてください。どちらのノードも `seed` の既定は 42 なので、プロンプトを編集する間はそのままにし、同じ方向の別テイクを聴きたいときだけ変えます。 ## 出力と保存 diff --git a/ko/tutorials/partner-nodes/pika/pika-audio.mdx b/ko/tutorials/partner-nodes/pika/pika-audio.mdx index b1fa7ba5c9..97cbe0821b 100644 --- a/ko/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/ko/tutorials/partner-nodes/pika/pika-audio.mdx @@ -2,19 +2,18 @@ title: "ComfyUI의 Pika Music 및 Pika SFX: AI 노래와 사운드 효과" description: "ComfyUI의 Pika Music 및 Pika SFX 노드로 스타일 프롬프트와 가사에서 노래 전체를 생성하고, 기존 트랙을 커버하며, 1~20초 사운드 효과를 만들 수 있습니다." sidebarTitle: "Pika Music 및 SFX" -translationSourceHash: c04fcf5c +translationSourceHash: ebd0cf37 translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx translationBlockHashes: "_intro": dc3cd241 "Get started": fd80ca7d - "Pika Music: text to music {#api_pika_music}": ae1738bd + "Pika Music: text to music {#api_pika_music}": b6238bbc "Pika Music: cover a song {#api_pika_music_cover}": 0793be8c - "Pika SFX: text to sound effects {#api_pika_sfx}": 81088686 - "Node controls": 2c4dd7da + "Pika SFX: text to sound effects {#api_pika_sfx}": d4966763 "Writing a music prompt": a24287f2 "Lyrics and section tags": af8f7f9e "Writing a sound effect prompt": 20f01804 - "Duration, limits and credits": 763b11c7 + "Duration, limits and credits": 7190bc82 "Output and saving": 4abf0bf7 "Which Pika audio node to use": 8b598461 "Related pages": 96e4c1fe @@ -58,7 +57,7 @@ Pika의 오디오 노드는 로컬 GPU 없이 텍스트 설명을 완성된 오 1. `Pika Music` 노드에서 `text to music`을 선택합니다. 2. `prompt`에 방향을 씁니다: 장르와 시대, BPM과 키, 훅을 담당하는 악기, 보컬의 성격, 그리고 곡이 섹션마다 어떻게 진행되는지. 최대 4,000자가 모델에 전달됩니다. 3. `[Verse]`, `[Chorus]` 같은 섹션 태그가 있는 `lyrics`를 추가합니다. 연주곡을 원하면 이 필드를 비워 둡니다. 최대 8,000자입니다. -4. `duration`을 10~360초로 설정합니다. `rewrite_prompt`는 프롬프트가 몇 단어뿐이고 모델이 먼저 확장해 주길 원할 때만 켜세요. +4. `duration`을 10~360초(기본 30)로 설정합니다. `rewrite_prompt`는 프롬프트가 몇 단어뿐이고 모델이 먼저 확장해 주길 원할 때만 켜세요. 5. 워크플로를 실행합니다. `Save Audio (Advanced)`가 MP3를 `ComfyUI/output/audio/`에 저장합니다. ## Pika Music: 곡 커버하기 {#api_pika_music_cover} @@ -82,40 +81,10 @@ Pika의 오디오 노드는 로컬 GPU 없이 텍스트 설명을 완성된 오 1. `Pika SFX` 프롬프트에 사운드를 씁니다. 예를 들어 녹슨 경첩에서 삐걱이며 천천히 열리는 낡은 목재 문, 이어서 무거운 걸쇠가 철컥 닫히는 소리입니다. -2. `duration`을 1~20초로 설정한 다음 워크플로를 실행합니다. MP3는 `ComfyUI/output/audio/`에 저장됩니다. +2. `duration`을 1~20초(기본 5)로 설정한 다음 워크플로를 실행합니다. MP3는 `ComfyUI/output/audio/`에 저장됩니다. 실행 한 번에 44.1 kHz 스테레오 효과음 하나가 생성됩니다. 연속된 사운드가 필요하면 사운드마다 노드를 한 번씩 실행하고 타임라인에 직접 배치하세요. Pika SFX는 음악이나 노래를 만들지 않습니다. 그 부분은 Pika Music이 담당합니다. -## 노드 컨트롤 - -`Pika Music`, text to music 모드: - -| 컨트롤 | 값 | 효과 | -| --- | --- | --- | -| `prompt` | 텍스트, 1~4,000자 | 방향: 장르, BPM, 악기 구성, 보컬 성격, 편곡 | -| `lyrics` | 선택, 최대 8,000자 | 부를 가사. 섹션 태그가 구조를 정합니다. 비우면 연주곡 | -| `duration` | 10~360초, 기본 30 | 트랙 길이 | -| `rewrite_prompt` | 기본 꺼짐 | 생성 전에 짧거나 모호한 프롬프트를 모델로 확장합니다 | -| `seed` | 정수, 기본 42 | 같은 방향의 변형을 탐색할 때 바꿉니다 | - -`Pika Music`, cover 모드: - -| 컨트롤 | 값 | 효과 | -| --- | --- | --- | -| `source_audio` | 오디오 입력 | 커버할 곡. 출력은 그 길이를 유지하며 5~360초로 제한됩니다 | -| `prompt` | 텍스트, 1~4,000자 | 그 곡을 어떤 새 편곡으로 다시 만들지 | -| `lyrics` | 선택, 최대 8,000자 | 가사를 바꾸거나 원래 가사를 유지합니다 | -| `rewrite_prompt` | 기본 꺼짐 | 생성 전에 짧거나 모호한 프롬프트를 모델로 확장합니다 | -| `seed` | 정수, 기본 42 | 같은 커버의 변형을 탐색할 때 바꿉니다 | - -`Pika SFX`: - -| 컨트롤 | 값 | 효과 | -| --- | --- | --- | -| `prompt` | 텍스트, 1~4,000자 | 사운드 자체: 재질, 동작, 공간, 마이크 거리감, 강도 | -| `duration` | 1~20초, 기본 5 | 효과음 길이 | -| `seed` | 정수, 기본 42 | 같은 방향의 변형을 탐색할 때 바꿉니다 | - ## 음악 프롬프트 작성법 **장르와 시대, BPM과 키, 훅을 담당하는 악기, 보컬 성격, 편곡의 진행, 믹스 방향** 순서로 방향을 세웁니다. 4,000자 전체가 모델에 전달되므로 구체적인 정보는 값을 하고, 형용사만 쌓으면 그렇지 않습니다. @@ -189,7 +158,7 @@ I'm not lost, I'm just taking the long way home 두 노드 모두 생성된 오디오의 초 단위로 크레딧이 청구되므로, 10초 스팅은 3분 곡의 일부만큼만 듭니다. 현재 요율은 [파트너 노드 가격](/ko/tutorials/partner-nodes/pricing)을 참고하세요. -`rewrite_prompt`가 꺼진 상태에서는 같은 프롬프트, 가사, 길이, seed로 같은 결과가 재현됩니다. `rewrite_prompt`를 켜면 생성 전에 모델이 프롬프트를 다시 쓰므로 seed만으로는 출력이 고정되지 않습니다. 재현 가능한 워크플로가 필요하면 꺼 두세요. +`rewrite_prompt`가 꺼진 상태에서는 같은 프롬프트, 가사, 길이, seed로 같은 결과가 재현됩니다. `rewrite_prompt`를 켜면 생성 전에 모델이 프롬프트를 다시 쓰므로 seed만으로는 출력이 고정되지 않습니다. 재현 가능한 워크플로가 필요하면 꺼 두세요. 두 노드 모두 `seed` 기본값은 42이므로 프롬프트를 수정하는 동안은 그대로 두고, 같은 방향의 다른 테이크를 들을 때만 바꿉니다. ## 출력과 저장 diff --git a/tutorials/partner-nodes/pika/pika-audio.mdx b/tutorials/partner-nodes/pika/pika-audio.mdx index acd772769b..f8607aded1 100644 --- a/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/tutorials/partner-nodes/pika/pika-audio.mdx @@ -42,7 +42,7 @@ Write the arrangement, the mood and the instrumentation into the prompt, then ad 1. Choose `text to music` in the `Pika Music` node. 2. Write the brief into `prompt`: genre and era, BPM and key, the instruments that carry the hook, the vocal character, and how the song moves from section to section. Up to 4,000 characters reach the model. 3. Add `lyrics` with section tags such as `[Verse]` and `[Chorus]`, or leave the field empty for an instrumental. Up to 8,000 characters. -4. Set `duration` between 10 and 360 seconds. Turn on `rewrite_prompt` only when your prompt is a few words and you want a model to expand it first. +4. Set `duration` between 10 and 360 seconds, 30 by default. Turn on `rewrite_prompt` only when your prompt is a few words and you want a model to expand it first. 5. Run the workflow. `Save Audio (Advanced)` writes an MP3 to `ComfyUI/output/audio/`. ## Pika Music: cover a song {#api_pika_music_cover} @@ -66,40 +66,10 @@ Name the sound in one line and the node returns a single effect. Pika follows th 1. Write the sound in the `Pika SFX` prompt, for example an old timber door creaking open on rusty hinges, then a heavy latch clicking shut. -2. Set `duration` between 1 and 20 seconds, then run the workflow. The MP3 lands in `ComfyUI/output/audio/`. +2. Set `duration` between 1 and 20 seconds, 5 by default, then run the workflow. The MP3 lands in `ComfyUI/output/audio/`. Each run produces one effect at 44.1 kHz stereo. For a sequence of sounds, run the node once per sound and place them on the timeline yourself. Pika SFX does not write music or sung vocals; use Pika Music for that. -## Node controls - -`Pika Music`, text to music mode: - -| Control | Values | Effect | -| --- | --- | --- | -| `prompt` | text, 1 to 4,000 characters | The brief: genre, BPM, instrumentation, vocal character, arrangement | -| `lyrics` | optional, up to 8,000 characters | The words to sing. Section tags set the structure. Empty means instrumental | -| `duration` | 10 to 360 seconds, default 30 | Length of the track | -| `rewrite_prompt` | off by default | Expands a short or vague prompt with a model before generating | -| `seed` | integer, default 42 | Change it to explore variations of the same brief | - -`Pika Music`, cover mode: - -| Control | Values | Effect | -| --- | --- | --- | -| `source_audio` | audio input | The song to cover. The output keeps its length, clamped to 5 to 360 seconds | -| `prompt` | text, 1 to 4,000 characters | The new arrangement to rebuild the song as | -| `lyrics` | optional, up to 8,000 characters | Replace the words, or keep the original ones | -| `rewrite_prompt` | off by default | Expands a short or vague prompt with a model before generating | -| `seed` | integer, default 42 | Change it to explore variations of the same cover | - -`Pika SFX`: - -| Control | Values | Effect | -| --- | --- | --- | -| `prompt` | text, 1 to 4,000 characters | The sound: material, action, space, perspective, intensity | -| `duration` | 1 to 20 seconds, default 5 | Length of the effect | -| `seed` | integer, default 42 | Change it to explore variations of the same brief | - ## Writing a music prompt Build the brief in this order: **genre and era, BPM and key, the instruments that carry the hook, vocal character, arrangement arc, mix notes**. The model reads all 4,000 characters, so detail earns its place; a stack of adjectives does not. @@ -173,7 +143,7 @@ One line, and four things in it: **what the material is, what happens to it, the Both nodes bill in credits by the second of generated audio, so a 10-second sting costs a fraction of a 3-minute song. See [Partner node pricing](/tutorials/partner-nodes/pricing) for the current rate. -The same prompt, lyrics, duration and seed with `rewrite_prompt` off reproduce the same result. With `rewrite_prompt` on, a model rewrites the brief before generation, so the seed alone no longer pins the output. Leave it off when you want a reproducible workflow. +The same prompt, lyrics, duration and seed with `rewrite_prompt` off reproduce the same result. With `rewrite_prompt` on, a model rewrites the brief before generation, so the seed alone no longer pins the output. Leave it off when you want a reproducible workflow. Both nodes start at `seed` 42, so hold it while you edit the prompt and change it only to audition another take. ## Output and saving diff --git a/zh/tutorials/partner-nodes/pika/pika-audio.mdx b/zh/tutorials/partner-nodes/pika/pika-audio.mdx index 7b48d7946e..ffc62318b7 100644 --- a/zh/tutorials/partner-nodes/pika/pika-audio.mdx +++ b/zh/tutorials/partner-nodes/pika/pika-audio.mdx @@ -2,19 +2,18 @@ title: "ComfyUI 中的 Pika Music 与 Pika SFX:AI 歌曲与音效" description: "在 ComfyUI 中用 Pika Music 与 Pika SFX 节点生成完整歌曲:风格提示词加可选歌词、翻唱已有曲目,或制作 1 到 20 秒的音效。" sidebarTitle: "Pika Music 与 SFX" -translationSourceHash: c04fcf5c +translationSourceHash: ebd0cf37 translationFrom: tutorials/partner-nodes/pika/pika-audio.mdx translationBlockHashes: "_intro": dc3cd241 "Get started": fd80ca7d - "Pika Music: text to music {#api_pika_music}": ae1738bd + "Pika Music: text to music {#api_pika_music}": b6238bbc "Pika Music: cover a song {#api_pika_music_cover}": 0793be8c - "Pika SFX: text to sound effects {#api_pika_sfx}": 81088686 - "Node controls": 2c4dd7da + "Pika SFX: text to sound effects {#api_pika_sfx}": d4966763 "Writing a music prompt": a24287f2 "Lyrics and section tags": af8f7f9e "Writing a sound effect prompt": 20f01804 - "Duration, limits and credits": 763b11c7 + "Duration, limits and credits": 7190bc82 "Output and saving": 4abf0bf7 "Which Pika audio node to use": 8b598461 "Related pages": 96e4c1fe @@ -58,7 +57,7 @@ Pika 的音频节点无需本地 GPU 就能把文字描述变成成品音频。* 1. 在 `Pika Music` 节点中选择 `text to music`。 2. 把创作方向写进 `prompt`:流派与年代、BPM 与调性、承载记忆点的乐器、人声特征,以及歌曲在各段之间如何推进。最多 4,000 个字符会送进模型。 3. 添加带分段标签的 `lyrics`,标签如 `[Verse]`、`[Chorus]`;要做纯音乐就把歌词字段留空。上限 8,000 个字符。 -4. 将 `duration` 设为 10 到 360 秒之间。只有当你只写了几个词、希望模型先扩写时,才打开 `rewrite_prompt`。 +4. 将 `duration` 设为 10 到 360 秒之间(默认 30)。只有当你只写了几个词、希望模型先扩写时,才打开 `rewrite_prompt`。 5. 运行工作流。`Save Audio (Advanced)` 会把 MP3 写入 `ComfyUI/output/audio/`。 ## Pika Music:翻唱一首歌 {#api_pika_music_cover} @@ -82,40 +81,10 @@ Pika 的音频节点无需本地 GPU 就能把文字描述变成成品音频。* 1. 在 `Pika SFX` 的提示词里写出声音,例如一扇老旧木门在生锈的合页上缓慢吱呀打开,随后一声沉重的门闩咔哒合上。 -2. 将 `duration` 设为 1 到 20 秒之间,然后运行工作流。MP3 会落在 `ComfyUI/output/audio/`。 +2. 将 `duration` 设为 1 到 20 秒之间(默认 5),然后运行工作流。MP3 会落在 `ComfyUI/output/audio/`。 每次运行生成一段 44.1 kHz 立体声音效。如果需要一组连续的声音,请为每个声音各运行一次节点,并自行把它们排到时间线上。Pika SFX 不写音乐、也不唱歌,那部分交给 Pika Music。 -## 节点控件 - -`Pika Music`,text to music 模式: - -| 控件 | 取值 | 作用 | -| --- | --- | --- | -| `prompt` | 文本,1 到 4,000 个字符 | 创作方向:流派、BPM、配器、人声特征、编曲 | -| `lyrics` | 可选,最多 8,000 个字符 | 要唱的歌词。分段标签决定结构,留空即纯音乐 | -| `duration` | 10 到 360 秒,默认 30 | 曲目长度 | -| `rewrite_prompt` | 默认关闭 | 生成前用模型扩写简短或含糊的提示词 | -| `seed` | 整数,默认 42 | 换一个值可以探索同一方向下的变化 | - -`Pika Music`,cover 模式: - -| 控件 | 取值 | 作用 | -| --- | --- | --- | -| `source_audio` | 音频输入 | 要翻唱的歌曲。输出沿用它的长度,限制在 5 到 360 秒 | -| `prompt` | 文本,1 到 4,000 个字符 | 要把这首歌重建成什么样的新编曲 | -| `lyrics` | 可选,最多 8,000 个字符 | 换掉歌词,或保留原词 | -| `rewrite_prompt` | 默认关闭 | 生成前用模型扩写简短或含糊的提示词 | -| `seed` | 整数,默认 42 | 换一个值可以探索同一次翻唱的变化 | - -`Pika SFX`: - -| 控件 | 取值 | 作用 | -| --- | --- | --- | -| `prompt` | 文本,1 到 4,000 个字符 | 声音本身:材质、动作、空间、听感距离、强度 | -| `duration` | 1 到 20 秒,默认 5 | 音效长度 | -| `seed` | 整数,默认 42 | 换一个值可以探索同一方向下的变化 | - ## 如何写音乐提示词 按这个顺序搭建方向:**流派与年代、BPM 与调性、承载记忆点的乐器、人声特征、编曲推进、混音取向**。4,000 个字符模型都会读,所以细节有价值,堆砌形容词没有。 @@ -189,7 +158,7 @@ I'm not lost, I'm just taking the long way home 两个节点都按生成音频的秒数计费,所以 10 秒的短标只花三分钟歌曲的一小部分。当前费率见 [合作伙伴节点定价](/zh/tutorials/partner-nodes/pricing)。 -在 `rewrite_prompt` 关闭时,相同的提示词、歌词、时长和 seed 会复现相同的结果。打开 `rewrite_prompt` 后,生成前会先由模型改写提示词,这时只靠 seed 已经无法锁定输出;需要可复现的工作流就把它关掉。 +在 `rewrite_prompt` 关闭时,相同的提示词、歌词、时长和 seed 会复现相同的结果。打开 `rewrite_prompt` 后,生成前会先由模型改写提示词,这时只靠 seed 已经无法锁定输出;需要可复现的工作流就把它关掉。两个节点的 `seed` 默认都是 42,改提示词时先保持不动,想听同一方向的另一种演绎再换值。 ## 输出与保存