Node reference

Use this reference after choosing the task you want to complete. Each entry combines an authored example and limitations with ports and controls generated from the current studio catalog. A node’s presence in the editor does not imply that it can be executed through run_node; see MCP execution boundaries.

The Russian labels in the generated catalog are the exact labels of the underlying controls. Protected instructions and legacy nodes are excluded. Read a node’s visible input preview before generating, and check the resulting artifact before using it downstream.

avg/text_input

Supply exact text without asking a model to write it.

When to use it: An existing brief, instruction or passage should enter the graph unchanged, or several texts should be joined into one.

Example: Enter “A blue paper boat crosses a quiet pond” and connect текст to the Prompt Scriptwriter’s тема input.

Check the result: Read the whole text in the result and the receiving node’s input preview.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• It does not extract the contents of an uploaded document.
• Texts connected to текст 1, текст 2… follow the field in port order, separated by a blank line; insert a text N placeholder in the field to put that wire in place instead. Type {{ in the field on the card, in the inspector or in the editor’s Edit view to pick an input from a list (connected inputs by source name, plus the next free one); typing @ in this field does not open a list.
• A text N placeholder with no wire, a connected input that arrives empty, or a placeholder inside a %rand% formula blocks the run; the card and the inspector name the reason (the same rule as Image and Video). Nodes below it cannot be run either while this node would run with them: their run button is grey and names this node. An up-to-date node that the run would reuse from cache does not block nodes below it. A wire whose source has not run yet, or whose source changed since it last ran, is not a block: running up to this node runs the source first. An Image or Video that uses a library or uploaded file does not send its prompt, so placeholders in it never block a run.
• In the inspector, the текст field opens with ✎ Editor, the same prompt editor as Image and Video prompts: formulas appear as chips and edits go straight into the field. Under the text, the Assembled — goes down the wire block shows exactly what the node sends on, formulas already rolled: texts from wires are highlighted and named by their source, connected texts without a placeholder are marked as added at the end, a source that has not run yet, or has changed since its last run, is marked as waiting for its run and shows the text the run will send, and a placeholder with no wire stays as typed, marked not connected — the run will not start. Everyone who can edit the node sees it.
• Editing the text can make connected generated results stale.

Live ports and controls

{
  "label": "Текст (ввод)",
  "inputs": [
    {
      "name": "текст 1",
      "type": "TEXT",
      "repeatable": true,
      "repeat_max": 10
    }
  ],
  "outputs": [
    {
      "name": "текст",
      "type": "TEXT"
    }
  ],
  "params": [
    {
      "key": "text",
      "label": "текст",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    }
  ]
}

avg/source_image

Bring an existing image into the graph.

When to use it: Use your own portrait, a finished scene frame or a storyboard sheet.

Example: Upload a portrait and connect картинка to an Element’s ракурс 1 input.

Check the result: The preview shows the intended file; the next node shows the same source.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• Upload or pin a file; there are no ordinary editable URL parameters.
• The receiving model decides whether this image is a reference, first frame or sheet.

Live ports and controls

{
  "label": "Картинка · файл",
  "inputs": [],
  "outputs": [
    {
      "name": "картинка",
      "type": "IMAGE_FRAME"
    }
  ],
  "params": []
}

avg/source_video

Bring an existing clip into the graph.

When to use it: Analyze a sample, edit a clip or provide a motion/video reference.

Example: Upload a clip and connect видео to Video Analysis’s видео input.

Check the result: Play the source and check its duration before analyzing or using it as a reference.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• A YouTube import first needs a quality choice from inspect_import.
• A clip connected through an Element is a different role from a video being edited or extended.

Live ports and controls

{
  "label": "Видео · загрузить",
  "inputs": [],
  "outputs": [
    {
      "name": "видео",
      "type": "VIDEO"
    }
  ],
  "params": []
}

avg/source_audio

Supply an audio file, or extract the soundtrack from a connected video.

When to use it: Reuse recorded speech, music or a clip’s existing audio.

Example: Connect a video to звук/видео, then connect звук to a finishing node.

Check the result: Listen to the output; extraction is not speech isolation.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• A video with no audio track cannot supply sound.
• Documents and unrecognized file types cannot serve as audio inputs.

Live ports and controls

{
  "label": "Звук · файл",
  "inputs": [
    {
      "name": "звук/видео",
      "type": "AUDIO,VIDEO"
    }
  ],
  "outputs": [
    {
      "name": "звук",
      "type": "AUDIO"
    }
  ],
  "params": []
}

avg/design_system

Supply a saved series identity or authored typography and graphic rules to Decoration.

When to use it: Reuse one visual identity across episodes, or manage its saved library copy separately from the film.

Example: Choose a saved authored series style from the library, then connect стиль to Decoration’s optional дизайн-система input.

Check the result: Confirm the saved edition and review the decorated text, components and motion against that identity.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• This node is optional: new Decoration V2 can invent a style and output the complete authored kit itself. Existing Editorial/Signal/Kinetic documents remain available as starting rules.
• An authored saved profile retains its CSS, component templates, motion, asset and skill pins through library reuse, graph reload and recipes.
• These are graphic overlay rules, not a character’s identity or a camera technique. Changing a design does not itself render the decorated video.

Live ports and controls

{
  "label": "Дизайн-система",
  "inputs": [],
  "outputs": [
    {
      "name": "стиль",
      "type": "DESIGN"
    }
  ],
  "params": [
    {
      "key": "title",
      "label": "название",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "design_direction",
      "label": "направление оформления",
      "kind": "select",
      "options": [
        {
          "value": "legacy",
          "label": "Прежнее оформление"
        },
        {
          "value": "authored",
          "label": "Авторский стиль серии"
        },
        {
          "value": "editorial",
          "label": "Editorial"
        },
        {
          "value": "signal",
          "label": "Signal"
        },
        {
          "value": "kinetic",
          "label": "Kinetic"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "editorial"
    },
    {
      "key": "md",
      "label": "документ (MD)",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "series_profile",
      "label": "сохранённый профиль серии",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    }
  ]
}

avg/text

Write or transform text using the connected context and an instruction.

When to use it: Create a short brief, rewrite supplied text or prepare captions.

Example: Connect a Text (input) to контекст and request three concise title options.

Check the result: Read the full generated text and verify that it preserved the supplied facts.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• This is a paid model call when no reusable result is available.
• Skills must match the node’s scope; output language is a separate setting.

Live ports and controls

{
  "label": "Текст (генерация)",
  "inputs": [
    {
      "name": "контекст",
      "type": "TEXT"
    },
    {
      "name": "картинка",
      "type": "IMAGE_FRAME"
    },
    {
      "name": "элемент 1",
      "type": "ELEMENT",
      "repeatable": true,
      "repeat_max": 10
    }
  ],
  "outputs": [
    {
      "name": "текст",
      "type": "TEXT"
    }
  ],
  "params": [
    {
      "key": "skill_ids",
      "label": "скиллы",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "provider",
      "label": "провайдер",
      "kind": "select",
      "options": [
        {
          "value": "anthropic",
          "label": "Anthropic"
        },
        {
          "value": "openai",
          "label": "OpenAI"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "anthropic"
    },
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "claude-opus-5-5",
          "label": "Claude Opus 5.5"
        },
        {
          "value": "claude-fable-5-1",
          "label": "Claude Fable 5.1"
        },
        {
          "value": "claude-opus-5",
          "label": "Claude Opus 5"
        },
        {
          "value": "claude-sonnet-5",
          "label": "Claude Sonnet 5"
        },
        {
          "value": "claude-opus-4-8",
          "label": "Claude Opus 4.8"
        }
      ],
      "dynamic": true,
      "disclosure": "essential",
      "default": "claude-opus-5-5"
    },
    {
      "key": "prompt",
      "label": "инструкция",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "max_tokens",
      "label": "лимит токенов (ответ)",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 256,
      "max": 128000,
      "disclosure": "essential",
      "default": 32000
    },
    {
      "key": "output_lang",
      "label": "язык результата",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "как на входе"
        },
        {
          "value": "ru",
          "label": "русский"
        },
        {
          "value": "en",
          "label": "английский"
        },
        {
          "value": "custom",
          "label": "свой…"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "output_lang_code",
      "label": "код языка",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    }
  ]
}

avg/prompt_model

Turn a sample breakdown and your brief into a prompt suited to the downstream video model.

When to use it: Keep selected aspects of a reference while supplying your own subject and requirements.

Example: Connect Video Analysis’s планы to контекст, your brief to бриф, and its текст output to Video’s текст 1.

Check the result: Review what is taken from the sample and what stays your own; verify the target model before generation.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• The six choices concern idea, camera/editing, action, place, light/colour and sound/rhythm.
• Your brief and Elements take precedence; source faces, handles and watermarks are not reusable creative instructions.

Live ports and controls

{
  "label": "Промпт под модель",
  "inputs": [
    {
      "name": "контекст",
      "type": "TEXT"
    },
    {
      "name": "бриф",
      "type": "TEXT"
    },
    {
      "name": "картинка",
      "type": "IMAGE_FRAME"
    },
    {
      "name": "элемент 1",
      "type": "ELEMENT",
      "repeatable": true,
      "repeat_max": 10
    }
  ],
  "outputs": [
    {
      "name": "текст",
      "type": "TEXT"
    }
  ],
  "params": [
    {
      "key": "skill_ids",
      "label": "скиллы",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "provider",
      "label": "провайдер",
      "kind": "select",
      "options": [
        {
          "value": "anthropic",
          "label": "Anthropic"
        },
        {
          "value": "openai",
          "label": "OpenAI"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "anthropic"
    },
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "claude-opus-5-5",
          "label": "Claude Opus 5.5"
        },
        {
          "value": "claude-fable-5-1",
          "label": "Claude Fable 5.1"
        },
        {
          "value": "claude-opus-5",
          "label": "Claude Opus 5"
        },
        {
          "value": "claude-sonnet-5",
          "label": "Claude Sonnet 5"
        },
        {
          "value": "claude-opus-4-8",
          "label": "Claude Opus 4.8"
        }
      ],
      "dynamic": true,
      "disclosure": "essential",
      "default": "claude-opus-5-5"
    },
    {
      "key": "take_preset",
      "label": "что берём из образца",
      "kind": "select",
      "options": [
        {
          "value": "idea",
          "label": "по идее"
        },
        {
          "value": "style",
          "label": "только как снято"
        },
        {
          "value": "story",
          "label": "сюжет"
        },
        {
          "value": "all",
          "label": "всё, кроме лиц"
        },
        {
          "value": "custom",
          "label": "свой выбор"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "take_idea",
      "label": "идея",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "по идее"
        },
        {
          "value": "take",
          "label": "берём"
        },
        {
          "value": "own",
          "label": "своё"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "take_camera",
      "label": "камера и монтаж",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "по идее"
        },
        {
          "value": "take",
          "label": "берём"
        },
        {
          "value": "own",
          "label": "своё"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "take_action",
      "label": "действие и хореография",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "по идее"
        },
        {
          "value": "take",
          "label": "берём"
        },
        {
          "value": "own",
          "label": "своё"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "take_place",
      "label": "место",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "по идее"
        },
        {
          "value": "take",
          "label": "берём"
        },
        {
          "value": "own",
          "label": "своё"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "take_style",
      "label": "свет и цвет",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "по идее"
        },
        {
          "value": "take",
          "label": "берём"
        },
        {
          "value": "own",
          "label": "своё"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "take_sound",
      "label": "звук и ритм",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "по идее"
        },
        {
          "value": "take",
          "label": "берём"
        },
        {
          "value": "own",
          "label": "своё"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    }
  ]
}

avg/image

Generate an image from a scene, text and applicable references.

When to use it: Create a finished scene frame, a design board or a storyboard sheet.

Example: Connect a written scene to сцена and run the Image node; inspect кадр before making video.

Check the result: Check the subject, framing, reference consistency and any lettering; inspect the emitted scene as well as the image.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• A model change can alter available reference inputs, quality, size and aspect choices.
• Unplaced text N inputs take precedence over scene text; use a text placeholder when the text should supplement the scene.

Live ports and controls

{
  "label": "Картинка",
  "inputs": [
    {
      "name": "сцена",
      "type": "SCENE"
    },
    {
      "name": "текст 1",
      "type": "TEXT",
      "aliases": [
        "промпт"
      ],
      "repeatable": true,
      "repeat_max": 10,
      "grows_by_address": true
    },
    {
      "name": "картинка 1",
      "type": "IMAGE_FRAME",
      "repeatable": true,
      "repeat_max": 16
    },
    {
      "name": "видео-референс",
      "type": "VIDEO"
    },
    {
      "name": "элемент 1",
      "type": "ELEMENT,IMAGE_FRAME",
      "note": "картинка здесь — референс (облик, место, стиль), а не герой. Героя ролика веди через «Элемент» с листом: портрет → лист → «Элемент», и уже его выход сюда",
      "repeatable": true,
      "repeat_max": 30,
      "grows_by_address": true
    }
  ],
  "outputs": [
    {
      "name": "кадр",
      "type": "IMAGE_FRAME"
    },
    {
      "name": "сцена",
      "type": "SCENE,TEXT,IMAGE_FRAME,VIDEO"
    }
  ],
  "params": [
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "seedream45",
          "label": "Seedream 4.5"
        },
        {
          "value": "nano_banana_21",
          "label": "Nano Banana 2.1"
        },
        {
          "value": "nano_banana_pro",
          "label": "Nano Banana Pro"
        },
        {
          "value": "gpt_image_25",
          "label": "GPT Image 2.5"
        },
        {
          "value": "flux2",
          "label": "FLUX.2 dev"
        },
        {
          "value": "flux_kontext",
          "label": "FLUX Kontext"
        },
        {
          "value": "seedream5_pro",
          "label": "Seedream 5.0 Pro"
        },
        {
          "value": "seedream5_lite",
          "label": "Seedream 5.0 Lite"
        },
        {
          "value": "grok_image_2",
          "label": "Grok Image 2.0"
        }
      ],
      "dynamic": false,
      "disclosure": "advanced",
      "default": "gpt_image_25"
    },
    {
      "key": "mode",
      "label": "режим",
      "kind": "select",
      "options": [
        {
          "value": "авто",
          "label": "авто (по референсам)"
        },
        {
          "value": "text_to_image",
          "label": "из промпта"
        },
        {
          "value": "reference_edit",
          "label": "с референсом"
        },
        {
          "value": "multi_reference",
          "label": "мульти-референс"
        }
      ],
      "dynamic": true,
      "disclosure": "advanced",
      "default": "авто"
    },
    {
      "key": "prompt",
      "label": "промпт (направление)",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "__context_preview",
      "label": "контекст (из провода)",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "element_context",
      "label": "контекст из элементов",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "advanced",
      "default": true
    },
    {
      "key": "aspect",
      "label": "формат",
      "kind": "select",
      "options": [
        {
          "value": "9:16",
          "label": "9:16"
        },
        {
          "value": "3:4",
          "label": "3:4"
        },
        {
          "value": "1:1",
          "label": "1:1"
        },
        {
          "value": "16:9",
          "label": "16:9"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "9:16"
    },
    {
      "key": "quality",
      "label": "качество",
      "kind": "select",
      "options": [
        {
          "value": "",
          "label": "по умолчанию · баланс"
        },
        {
          "value": "draft",
          "label": "черновик"
        },
        {
          "value": "light",
          "label": "лёгкое"
        },
        {
          "value": "balanced",
          "label": "баланс"
        },
        {
          "value": "premium",
          "label": "премиум"
        },
        {
          "value": "max",
          "label": "максимум"
        }
      ],
      "dynamic": true,
      "disclosure": "essential",
      "default": ""
    },
    {
      "key": "resolution",
      "label": "разрешение",
      "kind": "select",
      "options": [],
      "dynamic": true,
      "disclosure": "essential",
      "default": ""
    }
  ]
}

avg/video

Generate a clip using the selected task, model and available media.

When to use it: Animate a first frame, follow references, shoot from frames, edit, extend or transfer motion where supported, or animate a hero picture with a clip or a sound (Avatar).

Example: Connect a finished scene image to первый кадр and describe the subsequent motion.

Check the result: Watch the entire clip with sound; verify the opening, ending, action and reference use.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• A retained wire does not prove its content is sent: read input warnings for the selected mode.
• Storyboard sheets and timed keyframes are different inputs; a character sheet should not be used as the scene’s opening frame by accident.
• New nodes select the least expensive supported quality and resolution with sound enabled. Saved node settings retain their previous effective values. Resolution choices follow the selected quality; choosing Draft after 4K returns to a supported resolution.
• Some models always generate sound and provide no audio-off switch. Inspect the model page for resolution, quality, duration and sound limits.
• Avatar animates one hero picture either by the movement of a source clip or, with the driver voice, by a sound connected to sound 1; the output is as long as the clip or the sound, and the cost shown before the run is counted by that length.
• The FLOW pill shares the panel settings: model, actual read-only aspect, resolution, quality, duration and sound. Unavailable controls stay read-only or absent; the row wraps on a narrow canvas.

Live ports and controls

{
  "label": "Видео",
  "inputs": [
    {
      "name": "сцена",
      "type": "SCENE"
    },
    {
      "name": "первый кадр",
      "type": "IMAGE_FRAME"
    },
    {
      "name": "финальный кадр",
      "type": "IMAGE_FRAME"
    },
    {
      "name": "текст 1",
      "type": "TEXT",
      "aliases": [
        "промпт"
      ],
      "repeatable": true,
      "repeat_max": 10,
      "grows_by_address": true
    },
    {
      "name": "элемент 1",
      "type": "ELEMENT,IMAGE_FRAME,VIDEO",
      "note": "картинка здесь — референс (облик, место, стиль), а не герой. Героя ролика веди через «Элемент» с листом: портрет → лист → «Элемент», и уже его выход сюда",
      "repeatable": true,
      "repeat_max": 30,
      "grows_by_address": true
    }
  ],
  "outputs": [
    {
      "name": "видео",
      "type": "VIDEO"
    },
    {
      "name": "сцена",
      "type": "SCENE,TEXT,IMAGE_FRAME,VIDEO"
    }
  ],
  "params": [
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "kling_v3_standard",
          "label": "Kling",
          "facts": "EN/ZH bronze · ≤15с · реф.3 · герой: картинкой · правка: картинкой"
        },
        {
          "value": "seedance2",
          "label": "Seedance 2.0",
          "facts": "EN/ZH · ≤15с · реф.9 · герой: картинкой · правка: картинкой · люди: риск"
        },
        {
          "value": "seedance25",
          "label": "Seedance 2.5",
          "facts": "EN/ZH · ≤30с · реф.30 · герой: картинкой · правка: картинкой · люди: риск"
        },
        {
          "value": "grok15",
          "label": "Grok Video 1.5",
          "facts": "EN/RU platinum · ≤15с · реф.7 · герой: картинкой · правка: словами · рекомендуем"
        },
        {
          "value": "gemini_omni_flash",
          "label": "Gemini Omni Flash",
          "facts": "EN/RU platinum · ≤10с · реф.10 · герой: картинкой · правка: словами · проверено 23.10 · ролик: проверено 23.10 · правка: проверено 05.10"
        },
        {
          "value": "minimax_h3",
          "label": "MiniMax H3",
          "facts": "EN/RU silver · ≤15с · реф.9 · герой: картинкой · правка: картинкой"
        },
        {
          "value": "happyhorse11",
          "label": "HappyHorse 1.1",
          "facts": "EN gold · ≤15с · реф.9 · герой: картинкой · правка: картинкой · люди: не проверяли"
        },
        {
          "value": "flux3",
          "label": "FLUX 3",
          "facts": "EN/RU platinum · ≤20с · без реф. · герой: — · правка: словами"
        },
        {
          "value": "wan",
          "label": "Wan",
          "facts": "EN platinum · ≤30с · реф.10 · герой: картинкой · правка: картинкой · люди: не проверяли"
        },
        {
          "value": "dreamactor",
          "label": "DreamActor 2",
          "facts": "— · ≤30с · без реф. · герой: — · правка: нет"
        },
        {
          "value": "omnihuman",
          "label": "OmniHuman 2",
          "facts": "— · ≤300с · без реф. · герой: — · правка: нет"
        }
      ],
      "dynamic": false,
      "disclosure": "advanced",
      "default": "seedance2"
    },
    {
      "key": "mode",
      "label": "режим",
      "kind": "select",
      "options": [
        {
          "value": "авто",
          "label": "авто (по проводам и элементам)"
        },
        {
          "value": "from_first_frame",
          "label": "по первому кадру"
        },
        {
          "value": "from_references",
          "label": "новый ролик по референсу"
        },
        {
          "value": "text_to_video",
          "label": "из текста"
        },
        {
          "value": "edit_video",
          "label": "правка видео (как исходное)"
        },
        {
          "value": "extend_video",
          "label": "продолжение видео"
        }
      ],
      "dynamic": true,
      "disclosure": "advanced",
      "default": "авто"
    },
    {
      "key": "element_scope",
      "label": "элементы в генерации",
      "kind": "select",
      "options": [
        {
          "value": "all",
          "label": "все: свои + из сцены"
        },
        {
          "value": "explicit",
          "label": "только подключённые явно"
        },
        {
          "value": "none",
          "label": "без медиа элементов"
        }
      ],
      "dynamic": false,
      "disclosure": "advanced",
      "default": "all"
    },
    {
      "key": "prompt",
      "label": "промпт движения (направление)",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "__context_preview",
      "label": "контекст (из провода)",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "__normalizer_trace",
      "label": "промпт: как написал → как уехал",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "element_context",
      "label": "контекст из элементов",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "advanced",
      "adminOnly": true,
      "default": true
    },
    {
      "key": "ref_preamble",
      "label": "преамбула референсов",
      "kind": "select",
      "options": [
        {
          "value": "full",
          "label": "правила + список"
        },
        {
          "value": "list",
          "label": "только список"
        },
        {
          "value": "off",
          "label": "выкл"
        }
      ],
      "dynamic": false,
      "disclosure": "advanced",
      "adminOnly": true,
      "default": "full"
    },
    {
      "key": "duration",
      "label": "длительность, с",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 4,
      "max": 15,
      "disclosure": "essential",
      "default": 5
    },
    {
      "key": "duration_from_audio",
      "label": "длина — по звуку",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "advanced",
      "default": null
    },
    {
      "key": "audio",
      "label": "звук",
      "kind": "select",
      "options": [
        {
          "value": "origin",
          "label": "звук исходника"
        },
        {
          "value": "model",
          "label": "модель сочиняет"
        },
        {
          "value": "off",
          "label": "без звука"
        }
      ],
      "dynamic": true,
      "disclosure": "advanced",
      "default": null
    },
    {
      "key": "video_options_version",
      "label": "версия параметров видео",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 0,
      "max": 1,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "quality",
      "label": "вариант",
      "kind": "select",
      "options": [
        {
          "value": "draft",
          "label": "черновик"
        },
        {
          "value": "fast",
          "label": "быстрое"
        },
        {
          "value": "full",
          "label": "полное"
        }
      ],
      "dynamic": true,
      "disclosure": "advanced",
      "default": null
    },
    {
      "key": "resolution",
      "label": "разрешение",
      "kind": "select",
      "options": [
        {
          "value": "480p",
          "label": "480p"
        },
        {
          "value": "720p",
          "label": "720p"
        },
        {
          "value": "1080p",
          "label": "1080p"
        },
        {
          "value": "4k",
          "label": "4k"
        }
      ],
      "dynamic": true,
      "disclosure": "advanced",
      "default": null
    },
    {
      "key": "negative",
      "label": "негатив",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "advanced"
    }
  ]
}

avg/element

Carry a reusable character, object, location or style sample with its name and reference material.

When to use it: The same subject or source of context should be recognized across scenes.

Example: Apply an owned library Element here, then connect элемент to Prompt Scriptwriter’s элемент 1.

Check the result: Check its name, role, description and references before inspecting how a scene uses it.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• An Element is not a finished frame or clip.
• Reference count and media support depend on the receiving model and mode; identity consistency is not guaranteed.

Live ports and controls

{
  "label": "Элемент",
  "inputs": [
    {
      "name": "ракурс 1",
      "type": "IMAGE_FRAME",
      "repeatable": true,
      "repeat_max": 4
    },
    {
      "name": "видео 1",
      "type": "VIDEO",
      "repeatable": true,
      "repeat_max": 10
    },
    {
      "name": "звук 1",
      "type": "AUDIO,VIDEO",
      "repeatable": true,
      "repeat_max": 10
    }
  ],
  "outputs": [
    {
      "name": "элемент",
      "type": "ELEMENT"
    }
  ],
  "params": [
    {
      "key": "name",
      "label": "имя (для @имя в промпте)",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "role",
      "label": "роль",
      "kind": "select",
      "options": [
        {
          "value": "персонаж",
          "label": "персонаж"
        },
        {
          "value": "объект",
          "label": "объект"
        },
        {
          "value": "локация",
          "label": "локация"
        },
        {
          "value": "раскадровка",
          "label": "раскадровка"
        },
        {
          "value": "образец стиля",
          "label": "образец стиля"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "персонаж"
    },
    {
      "key": "description",
      "label": "описание",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "character",
      "label": "характер",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "look",
      "label": "что брать с картинки",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    }
  ]
}

avg/prompt_scriptwriter

Turn an idea and connected Elements into scenes with image and video instructions.

When to use it: Plan a multi-scene film before generating media.

Example: Connect the idea to тема, add a character Element and inspect each сцена N plus полный текст.

Check the result: Review all scenes, durations, dialogue and the critic’s reported outcome before expansion.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• A failed or unknown critic check is not a clean review.
• Re-running can replace hand-edited scenes; read the run warnings before agreeing.

Live ports and controls

{
  "label": "Промпт Сценарист",
  "inputs": [
    {
      "name": "тема",
      "type": "TEXT"
    },
    {
      "name": "элемент 1",
      "type": "ELEMENT,IMAGE_FRAME",
      "note": "картинка здесь — референс (облик, место, стиль), а не герой. Героя ролика веди через «Элемент» с листом: портрет → лист → «Элемент», и уже его выход сюда",
      "repeatable": true,
      "repeat_max": 10
    }
  ],
  "outputs": [
    {
      "name": "сцена 1",
      "type": "SCENE,TEXT,IMAGE_FRAME,VIDEO",
      "repeatable": true,
      "repeat_max": 60
    },
    {
      "name": "полный текст",
      "type": "TEXT"
    }
  ],
  "params": [
    {
      "key": "skill_ids",
      "label": "скиллы",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "advanced"
    },
    {
      "key": "prompt",
      "label": "запрос",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "style",
      "label": "визуальный стиль",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "critic",
      "label": "проверить сцены критиком (+$0.10)",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "essential",
      "default": false
    },
    {
      "key": "aspect",
      "label": "формат ролика",
      "kind": "select",
      "options": [
        {
          "value": "9:16",
          "label": "9:16"
        },
        {
          "value": "3:4",
          "label": "3:4"
        },
        {
          "value": "1:1",
          "label": "1:1"
        },
        {
          "value": "16:9",
          "label": "16:9"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "video_priority",
      "label": "модель видео",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "Авто"
        },
        {
          "value": "budget",
          "label": "Экономно"
        },
        {
          "value": "quality",
          "label": "Качество"
        },
        {
          "value": "speech",
          "label": "Язык реплик"
        },
        {
          "value": "kling_v3_standard",
          "label": "Kling",
          "facts": "EN/ZH bronze · ≤15с · реф.3 · герой: картинкой · правка: картинкой"
        },
        {
          "value": "seedance2",
          "label": "Seedance 2.0",
          "facts": "EN/ZH · ≤15с · реф.9 · герой: картинкой · правка: картинкой · люди: риск"
        },
        {
          "value": "seedance25",
          "label": "Seedance 2.5",
          "facts": "EN/ZH · ≤30с · реф.30 · герой: картинкой · правка: картинкой · люди: риск"
        },
        {
          "value": "grok15",
          "label": "Grok Video 1.5",
          "facts": "EN/RU platinum · ≤15с · реф.7 · герой: картинкой · правка: словами · рекомендуем"
        },
        {
          "value": "gemini_omni_flash",
          "label": "Gemini Omni Flash",
          "facts": "EN/RU platinum · ≤10с · реф.10 · герой: картинкой · правка: словами · проверено 23.10 · ролик: проверено 23.10 · правка: проверено 05.10"
        },
        {
          "value": "minimax_h3",
          "label": "MiniMax H3",
          "facts": "EN/RU silver · ≤15с · реф.9 · герой: картинкой · правка: картинкой"
        },
        {
          "value": "happyhorse11",
          "label": "HappyHorse 1.1",
          "facts": "EN gold · ≤15с · реф.9 · герой: картинкой · правка: картинкой · люди: не проверяли"
        },
        {
          "value": "flux3",
          "label": "FLUX 3",
          "facts": "EN/RU platinum · ≤20с · без реф. · герой: — · правка: словами"
        },
        {
          "value": "wan",
          "label": "Wan",
          "facts": "EN platinum · ≤30с · реф.10 · герой: картинкой · правка: картинкой · люди: не проверяли"
        },
        {
          "value": "dreamactor",
          "label": "DreamActor 2",
          "facts": "— · ≤30с · без реф. · герой: — · правка: нет"
        },
        {
          "value": "omnihuman",
          "label": "OmniHuman 2",
          "facts": "— · ≤300с · без реф. · герой: — · правка: нет"
        }
      ],
      "dynamic": true,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "claude-opus-5-5",
          "label": "Claude Opus 5.5"
        },
        {
          "value": "claude-fable-5-1",
          "label": "Claude Fable 5.1"
        },
        {
          "value": "claude-opus-5",
          "label": "Claude Opus 5"
        },
        {
          "value": "claude-sonnet-5",
          "label": "Claude Sonnet 5"
        },
        {
          "value": "claude-opus-4-8",
          "label": "Claude Opus 4.8"
        }
      ],
      "dynamic": false,
      "disclosure": "advanced",
      "default": "claude-opus-5-5"
    },
    {
      "key": "max_tokens",
      "label": "лимит токенов (мышление + ответ)",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 4000,
      "max": 128000,
      "disclosure": "advanced",
      "default": 32000
    }
  ]
}

avg/video_analysis

Describe a reference clip’s idea, camera, action, place, lighting and sound.

When to use it: Study how a sample works before choosing what to reuse.

Example: Connect a clip to видео; choose quick or detailed with sound; send планы to Prompt for the model.

Check the result: Read the notes and compare the breakdown with the clip; listen when using the detailed mode.

• UI: available in the studio. MCP: run_node supports this node; inspect its dependencies and quote before execution.
• The quick/detailed setting controls depth of analysis; the receiving authoring node decides what to take.
• Analysis has an estimated ceiling and charges measured work; a style pack is saved to personal prompts, not a global style preset.

Live ports and controls

{
  "label": "Разбор ролика",
  "inputs": [
    {
      "name": "видео",
      "type": "VIDEO"
    },
    {
      "name": "вопрос",
      "type": "TEXT"
    }
  ],
  "outputs": [
    {
      "name": "разбор",
      "type": "TEXT"
    },
    {
      "name": "стиль кадра",
      "type": "TEXT"
    },
    {
      "name": "стиль движения",
      "type": "TEXT"
    },
    {
      "name": "планы",
      "type": "TEXT"
    }
  ],
  "params": [
    {
      "key": "skill_ids",
      "label": "скиллы",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "mode",
      "label": "подробность",
      "kind": "select",
      "options": [
        {
          "value": "style",
          "label": "быстро"
        },
        {
          "value": "story",
          "label": "подробно, со звуком"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    }
  ]
}

avg/assemble

Join the connected clips in input order and optionally mix music or voice.

When to use it: Approved scene clips need to become one film.

Example: Connect clips in scene order to видео 1, видео 2 and following inputs; run Montage in the studio.

Check the result: Play the whole result; check order, cuts, aspect handling and sound levels.

• UI: available in the studio. MCP: run_node does not support this node; assemble_video joins explicitly supplied clips.
• run_node currently refuses this node; the separate assemble_video tool accepts explicit clips.
• Expanding scene branches out of order can also affect Montage’s input order; inspect warnings.
• Clip audio (звук клипов) set to Turn off removes the clips’ own sound from the film; music and voice remain. Without a voice the music plays at its own level and music volume is hidden, because there is nothing to sit under. Use it when a model always generates sound, for example MiniMax H3, and the film goes under your own song.
• Montage Studio (Open the timeline in the panel, not on the card) reorders clips by rewiring the видео N inputs, trims a clip’s edges in 1/30 s frames, mutes one clip and shifts the song; the server builds exactly what the timeline shows. The edits live in the node and do not travel with a published recipe; run_node, assemble_video’s description and the copilot do not edit them.
• In the add-node menu (✚) Montage is in the Montage section, next to Design system, Subtitles and Decoration.

Live ports and controls

{
  "label": "Монтаж",
  "inputs": [
    {
      "name": "музыка",
      "type": "AUDIO,VIDEO"
    },
    {
      "name": "голос",
      "type": "AUDIO,VIDEO"
    },
    {
      "name": "видео 1",
      "type": "VIDEO",
      "repeatable": true,
      "repeat_max": null
    }
  ],
  "outputs": [
    {
      "name": "видео",
      "type": "VIDEO"
    }
  ],
  "params": [
    {
      "key": "clip_audio",
      "label": "звук клипов",
      "kind": "select",
      "options": [
        {
          "value": "keep",
          "label": "оставить"
        },
        {
          "value": "mute",
          "label": "выключить"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "music_volume",
      "label": "громкость музыки",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 0,
      "max": 1,
      "disclosure": "advanced",
      "default": 0.2
    },
    {
      "key": "plan_trim",
      "label": "обрезка по плану",
      "kind": "select",
      "options": [
        {
          "value": "plan",
          "label": "по плану"
        },
        {
          "value": "off",
          "label": "не резать"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    }
  ]
}

avg/subtitles

Recognize speech in an audio track or a video and provide SRT and VTT subtitles.

When to use it: A finished spoken clip or a voice/audio track needs captions or an editable transcript.

Example: Connect a video or an audio track to звук/видео, recognize its speech, inspect timings, then connect VTT to Decoration.

Check the result: Watch with captions and correct names, wording and timing; verify the chosen language.

• UI: available in the studio. MCP: run_node does not support this node; make_subtitles and normalize_subtitles operate on supplied media/text.
• Silent video or audio cannot yield spoken words.
• Recognition and optional text normalization are separate operations; neither burns captions into the image by itself.

Live ports and controls

{
  "label": "Субтитры",
  "inputs": [
    {
      "name": "звук/видео",
      "type": "AUDIO,VIDEO"
    }
  ],
  "outputs": [
    {
      "name": "SRT",
      "type": "TEXT"
    },
    {
      "name": "VTT",
      "type": "TEXT"
    }
  ],
  "params": [
    {
      "key": "skill_ids",
      "label": "скиллы",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "language",
      "label": "язык речи",
      "kind": "select",
      "options": [
        {
          "value": "ru",
          "label": "русский"
        },
        {
          "value": "en",
          "label": "английский"
        },
        {
          "value": "auto",
          "label": "автоопределение"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "ru"
    },
    {
      "key": "asr_model",
      "label": "модель распознавания",
      "kind": "select",
      "options": [
        {
          "value": "scribe_v2",
          "label": "Scribe v2 — песни и музыка"
        },
        {
          "value": "whisper-large-v3",
          "label": "Whisper large-v3"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "scribe_v2"
    },
    {
      "key": "context",
      "label": "контекст для нормализации",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "normalize_model",
      "label": "модель нормализации",
      "kind": "select",
      "options": [
        {
          "value": "claude-opus-5-5",
          "label": "Claude Opus 5.5"
        },
        {
          "value": "claude-fable-5-1",
          "label": "Claude Fable 5.1"
        },
        {
          "value": "claude-opus-5",
          "label": "Claude Opus 5"
        },
        {
          "value": "claude-sonnet-5",
          "label": "Claude Sonnet 5"
        },
        {
          "value": "claude-opus-4-8",
          "label": "Claude Opus 4.8"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "claude-opus-5-5"
    }
  ]
}

avg/decorate

Design and render captions, titles, plates and an optional end card over a video, then output its reusable series style.

When to use it: Finish an assembled episode with graphics grounded in its actual images, speech and connected project intent.

Example: Connect the finished video and run V2; optionally connect context, reviewed VTT or a saved series style. Review the result in Series design and reuse its style output for the next episode.

Check the result: Watch the entire result for exact words, readable placement and timing; review Why these overlays and the source coverage limits.

• UI: available in the studio. MCP: run_node does not support this node; decorate_video is a separate lower-level route.
• New V2 nodes start with server rendering and need only a video connection. Without a connected style, the agent creates an original CSS/component/motion kit; a connected authored style reuses those exact components.
• V2 analyses sampled source images and source audio, preserving supplied VTT instead of transcribing it again. Sampling can miss events, and long videos have sparse visual coverage; complete long-project production is unverified.
• The series style output, library save, export, graph reload and recipes carry the complete authored kit and pinned skill revisions. Saved text edits and server rebuilds do not call the author again.
• Saved legacy nodes remain in the previous decoration mode and retain their studio-first behavior. Card/TikTok look and browser Build here apply to that previous mode; explicit Editorial/Signal/Kinetic directions use HyperFrames server export.

Live ports and controls

{
  "label": "Оформление",
  "inputs": [
    {
      "name": "видео",
      "type": "VIDEO"
    },
    {
      "name": "контекст",
      "type": "TEXT"
    },
    {
      "name": "VTT",
      "type": "TEXT"
    },
    {
      "name": "дизайн-система",
      "type": "DESIGN"
    }
  ],
  "outputs": [
    {
      "name": "видео",
      "type": "VIDEO"
    },
    {
      "name": "дизайн-система",
      "type": "DESIGN"
    }
  ],
  "params": [
    {
      "key": "design_direction",
      "label": "оформление",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "V2 · по смыслу проекта"
        },
        {
          "value": "legacy",
          "label": "Прежнее оформление"
        },
        {
          "value": "editorial",
          "label": "Editorial"
        },
        {
          "value": "signal",
          "label": "Signal"
        },
        {
          "value": "kinetic",
          "label": "Kinetic"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "auto"
    },
    {
      "key": "series_mode",
      "label": "стиль серии",
      "kind": "select",
      "options": [
        {
          "value": "reuse",
          "label": "сохранить созданный стиль"
        },
        {
          "value": "new",
          "label": "создать новый стиль"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "run_action",
      "label": "что делает ▶",
      "kind": "select",
      "options": [
        {
          "value": "studio",
          "label": "план + студия оформления"
        },
        {
          "value": "render",
          "label": "собрать сразу на сервере"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "studio"
    },
    {
      "key": "model",
      "label": "модель плана",
      "kind": "select",
      "options": [
        {
          "value": "claude-opus-5-5",
          "label": "Claude Opus 5.5"
        },
        {
          "value": "claude-fable-5-1",
          "label": "Claude Fable 5.1"
        },
        {
          "value": "claude-opus-5",
          "label": "Claude Opus 5"
        },
        {
          "value": "claude-sonnet-5",
          "label": "Claude Sonnet 5"
        },
        {
          "value": "claude-opus-4-8",
          "label": "Claude Opus 4.8"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "claude-opus-5-5"
    },
    {
      "key": "burn_subtitles",
      "label": "вжигать субтитры",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "essential",
      "default": true
    },
    {
      "key": "caption_look",
      "label": "вид субтитров",
      "kind": "select",
      "options": [
        {
          "value": "card",
          "label": "Карточка"
        },
        {
          "value": "tiktok",
          "label": "TikTok"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "tiktok"
    },
    {
      "key": "plates_on",
      "label": "плашки",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "essential",
      "default": true
    },
    {
      "key": "titles_on",
      "label": "титры секций",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "essential",
      "default": true
    },
    {
      "key": "overlay_brief",
      "label": "пожелания к оформлению",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "outro",
      "label": "финальная карта (аутро)",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "авто (решает режиссёр)"
        },
        {
          "value": "on",
          "label": "всегда"
        },
        {
          "value": "off",
          "label": "выключить"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "auto"
    },
    {
      "key": "outro_text",
      "label": "текст аутро",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "cta",
      "label": "призыв на аутро (CTA)",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "series_profile",
      "label": "сохранённый профиль серии",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    }
  ]
}

avg/music

Create music from a musical brief, optional lyrics or a source audio track.

When to use it: A scene or film needs an instrumental track or song.

Example: For a new instrumental track, enter a music brief. To transform an existing track, choose Stable Audio 3 or 2.5, connect audio or video to аудио, and describe the desired music.

Check the result: Listen to the complete track, then audition it under the film.

• UI: available in the studio. MCP: run_node does not support this node; generate_music accepts an explicit brief.
• Lyrics request singing only on a model that supports vocals.
• The low-level generate_music tool is queued; the Music node is not currently supported by run_node.
• Audio input is connected for Stable Audio 3 and Stable Audio 2.5. MiniMax Music 3 and Eleven Music 2.5 reject this input; disconnect it or choose Stable Audio.
• A video on the audio input supplies its soundtrack. The source is not silently omitted when a model cannot accept it.

Live ports and controls

{
  "label": "Музыка",
  "inputs": [
    {
      "name": "промпт",
      "type": "TEXT"
    },
    {
      "name": "лирика",
      "type": "TEXT"
    },
    {
      "name": "аудио",
      "type": "AUDIO,VIDEO"
    }
  ],
  "outputs": [
    {
      "name": "аудио",
      "type": "AUDIO"
    }
  ],
  "params": [
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "stable_audio_3",
          "label": "Stable Audio 3"
        },
        {
          "value": "minimax_music_3",
          "label": "MiniMax Music 3"
        },
        {
          "value": "elevenlabs_music_25",
          "label": "ElevenLabs Music v2.5"
        },
        {
          "value": "stable_audio_25",
          "label": "Stable Audio 2.5 (прошлая версия)"
        }
      ],
      "dynamic": true,
      "disclosure": "essential",
      "default": "stable_audio_3"
    },
    {
      "key": "prompt",
      "label": "промпт",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "lyrics",
      "label": "лирика",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "seconds",
      "label": "длительность, с",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 1,
      "max": 380,
      "disclosure": "essential",
      "default": 30
    }
  ]
}

avg/music_caption

Rewrite a music brief with the original MiniMax Music Caption Rewriter skill, and turn supplied lyrics into a clean sheet of tags and sung words.

When to use it: Prepare a structured MiniMax Music 3 caption from a short description or a reviewed Audio Breakdown answer.

Example: Enter “Gentle acoustic folk, instrumental, no drums” as Caption. Add “Leave space for dialogue” as additional instructions. Run, review the caption, and connect промпт to Music → промпт. Connect лирика only for a vocal track on a model that supports it.

Check the result: Read the complete caption and compare it with your requirements. Confirm that instrumental or vocal choices are correct. If the lyrics breakdown is on, open it: check which voices sing which sections and read every removed fragment with its reason.

• Caption is the required musical description. Lyrics may be written in any form, including Markdown, performer names, stage directions and comments. Connected Caption or Lyrics replaces its field. Put the scenario and extra context in additional instructions or up to ten ordered текст N inputs.
• Output промпт contains Global Metadata, Vocal Details and Arrangement. English is the default unless another language is explicitly requested. Output лирика is a clean sheet when the lyrics breakdown is on (see below); otherwise it is the supplied text unchanged. No new lyrics or audio are generated.
• Lyrics breakdown: a separate studio skill splits the lyrics into voices and sections. The node then outputs only section tags and sung or spoken lines. Titles, style paragraphs, separators, performer-name labels, stage directions and comments are removed, and the card shows how many were removed by reason; the panel lists each removed fragment verbatim with its reason and the voice of every section. Code checks that every word of the sheet is in your text and that every word of your text is either kept or listed as removed; a failed check allows one correction, then the run stops with an error. The node does not write or reword lyrics. Nodes run before this change keep their results until you run them again.
• Voices on ElevenLabs Music v2.5: when the breakdown finds two or more voices or a spoken section, and the prompt and lyrics both come from this node with the ElevenLabs skill, Music sends a composition plan, one chunk per section, instead of one prompt. Chunk lengths are computed by code from the lines and scaled to the Music node’s seconds. A chunk is 3 to 120 s, at most 30 chunks, 10 minutes in total; a sheet that does not fit is refused with the length it needs, never cut. The price is unchanged: per started minute of seconds. On MiniMax Music 3 the lyrics go as tags and one voice sings; a duet is not guaranteed, and the Music card says so. A listening test of one lyrics sheet separated voices only with the composition plan; it is not a quality guarantee.
• A separate semantic check compares the caption with the installed skill and complete inputs. One grounded rejection permits one repair and recheck within the same spending ceiling; invalid/unavailable checking or a second rejection fails. Supplied Lyrics is never silently cleaned.
• The selected skill version is pinned per node. Administrator update installs a complete revision and changes only the selected node; existing nodes retain their versions.
• The displayed $1 is the current spending ceiling, not a flat fee; successful work settles from actual token usage. The later Music generation has its own cost.
• Incomplete required reference reads or an incomplete response fail explicitly; there is no ordinary-rewrite fallback and oversized input is rejected without cutting text.
• UI: available in Studio. MCP: run_node does not support this node. Contract tests do not establish caption or music quality.

Live ports and controls

{
  "label": "Music Caption Rewriter",
  "inputs": [
    {
      "name": "Caption",
      "type": "TEXT"
    },
    {
      "name": "Lyrics",
      "type": "TEXT"
    },
    {
      "name": "текст 1",
      "type": "TEXT",
      "repeatable": true,
      "repeat_max": 10
    }
  ],
  "outputs": [
    {
      "name": "промпт",
      "type": "TEXT"
    },
    {
      "name": "лирика",
      "type": "TEXT"
    }
  ],
  "params": [
    {
      "key": "skill",
      "label": "скил",
      "kind": "select",
      "options": [
        {
          "value": "minimax",
          "label": "MiniMax Music 3"
        },
        {
          "value": "elevenlabs_music_25",
          "label": "ElevenLabs Music v2.5"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "minimax"
    },
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "claude-opus-5-5",
          "label": "Claude Opus 5.5"
        },
        {
          "value": "claude-opus-5",
          "label": "Claude Opus 5"
        },
        {
          "value": "claude-sonnet-5",
          "label": "Claude Sonnet 5"
        },
        {
          "value": "claude-opus-4-8",
          "label": "Claude Opus 4.8"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "claude-opus-5-5"
    },
    {
      "key": "genre",
      "label": "жанр",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "авто"
        },
        {
          "value": "general-pop-ballad",
          "label": "поп, поп-баллада"
        },
        {
          "value": "hip-hop-rap",
          "label": "хип-хоп, рэп"
        },
        {
          "value": "pop-alternative-rock",
          "label": "поп-рок, альтернатива"
        },
        {
          "value": "club-edm-house-trance",
          "label": "EDM, хаус, транс"
        },
        {
          "value": "dance-pop-disco-funk",
          "label": "дэнс-поп, диско, фанк"
        },
        {
          "value": "electronic-synth-ambient-pop",
          "label": "синти-поп, электроника, эмбиент"
        },
        {
          "value": "modern-rnb-neo-soul",
          "label": "R&B, нео-соул"
        },
        {
          "value": "cinematic-pop-ballad",
          "label": "кинематографичная баллада"
        },
        {
          "value": "cinematic-orchestral-epic",
          "label": "оркестр, эпик, саундтрек"
        },
        {
          "value": "metal-heavy-rock",
          "label": "метал, тяжёлый рок"
        },
        {
          "value": "contemporary-folk-acoustic",
          "label": "фолк, акустика"
        },
        {
          "value": "country-americana",
          "label": "кантри, американа"
        },
        {
          "value": "jazz-swing-big-band",
          "label": "джаз, свинг, биг-бэнд"
        },
        {
          "value": "soul-blues-gospel",
          "label": "соул, блюз, госпел"
        },
        {
          "value": "roots-traditional-global",
          "label": "народная, этника, регги"
        },
        {
          "value": "traditional-vocal-stage",
          "label": "эстрада, мюзикл, кабаре"
        },
        {
          "value": "east-asian-modern",
          "label": "азиатский поп, современный"
        },
        {
          "value": "east-asian-ballad-heritage",
          "label": "азиатская баллада, гофэн"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "mood",
      "label": "настроение",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "авто"
        },
        {
          "value": "calm",
          "label": "спокойно"
        },
        {
          "value": "drive",
          "label": "драйв"
        },
        {
          "value": "epic",
          "label": "эпично"
        },
        {
          "value": "dark",
          "label": "мрачно"
        },
        {
          "value": "fun",
          "label": "весело"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "vocals",
      "label": "вокал",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "как в лирике"
        },
        {
          "value": "female",
          "label": "женский"
        },
        {
          "value": "male",
          "label": "мужской"
        },
        {
          "value": "duet",
          "label": "дуэт"
        },
        {
          "value": "choir",
          "label": "хор"
        },
        {
          "value": "none",
          "label": "без вокала"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": null
    },
    {
      "key": "caption",
      "label": "Caption",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "lyrics",
      "label": "Lyrics",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "additional_text",
      "label": "дополнительные указания",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "skill_version",
      "label": "версия скилла",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    }
  ]
}

avg/audio_analysis

Describe music, speech and other sounds as text with Audio Understanding.

When to use it: Understand an existing recording before writing a music brief or using it in a film.

Example: Connect an audio file or a video to аудио. Ask “Which instruments are audible, and how does the rhythm change?” in the question field or through the вопрос input; run the node and read разбор.

Check the result: Compare the complete answer with the original recording; an analysis may misidentify sounds or speech.

• UI: available in the studio. MCP: run_node does not support this node, and no separate audio-analysis tool is exposed in this increment.
• Audio Understanding is the single model. Detailed analysis requests a more detailed answer; the output is text, not isolated tracks or a generated audio file.
• Video supplies its extracted soundtrack. The analysis runs as a queued job; completed text is retained in the library and restored after reopening the project.
• The published rate is $0.002 per second of input audio. The server measures duration before reserving credits. These checks used mocked model responses; audio-analysis quality was not benchmarked.

Live ports and controls

{
  "label": "Разбор аудио",
  "inputs": [
    {
      "name": "аудио",
      "type": "AUDIO,VIDEO"
    },
    {
      "name": "вопрос",
      "type": "TEXT"
    }
  ],
  "outputs": [
    {
      "name": "разбор",
      "type": "TEXT"
    }
  ],
  "params": [
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "fal-ai/audio-understanding",
          "label": "Audio Understanding"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "fal-ai/audio-understanding"
    },
    {
      "key": "prompt",
      "label": "вопрос",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "detailed_analysis",
      "label": "подробный разбор",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "essential",
      "default": false
    }
  ]
}

avg/voice

Select a saved voice or create a voice from a recording.

When to use it: Choose the voice identity before generating spoken audio.

Example: Select a ready voice and connect голос to Voiceover’s голос input.

Check the result: Listen to the voice preview and confirm the intended voice before a full script.

• UI: available in the studio. MCP: run_node does not support this node; clone_voice can create an identifier from an owned sample; use the browser voice picker for selection.
• Creating a voice from a sample requires an owned uploaded recording and can charge.
• This node chooses a voice; it does not speak the script by itself.

Live ports and controls

{
  "label": "Голос",
  "inputs": [
    {
      "name": "образец",
      "type": "AUDIO,VIDEO"
    }
  ],
  "outputs": [
    {
      "name": "голос",
      "type": "VOICE"
    }
  ],
  "params": [
    {
      "key": "voice_source",
      "label": "источник голоса",
      "kind": "select",
      "options": [
        {
          "value": "ready",
          "label": "готовый (ElevenLabs)"
        },
        {
          "value": "clone",
          "label": "клон из образца"
        },
        {
          "value": "saved",
          "label": "из библиотеки"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "ready"
    },
    {
      "key": "alias",
      "label": "имя голоса",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "remove_bg_noise",
      "label": "убрать фон в образце",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "essential",
      "default": false
    },
    {
      "key": "voice_id",
      "label": "voice_id (ElevenLabs)",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    }
  ]
}

avg/voiceover

Speak supplied text with a selected voice.

When to use it: Create narration or a replacement speech track.

Example: Connect Voice and a Text (input), then generate and listen to аудио.

Check the result: Verify pronunciation, complete wording, pacing and duration.

• UI: available in the studio. MCP: run_node does not support this node; generate_voice accepts explicit text and a voice ID.
• A voice_id and nonempty text are required by generate_voice.
• The output is audio; attaching it to video or matching lips is separate work.

Live ports and controls

{
  "label": "Озвучка",
  "inputs": [
    {
      "name": "голос",
      "type": "VOICE"
    },
    {
      "name": "текст",
      "type": "TEXT"
    }
  ],
  "outputs": [
    {
      "name": "аудио",
      "type": "AUDIO"
    }
  ],
  "params": [
    {
      "key": "text",
      "label": "текст",
      "kind": "textarea",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "voice_id",
      "label": "voice_id",
      "kind": "text",
      "options": [],
      "dynamic": false,
      "disclosure": "essential"
    },
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "eleven_multilingual_v2",
          "label": "Multilingual v2 (дословно, RU)"
        },
        {
          "value": "eleven_v3",
          "label": "v3 (выразительно)"
        },
        {
          "value": "eleven_flash_v2_5",
          "label": "Flash v2.5 (быстро)"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "eleven_multilingual_v2"
    },
    {
      "key": "stability",
      "label": "стабильность",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 0,
      "max": 1,
      "disclosure": "essential",
      "default": 0.5
    },
    {
      "key": "similarity_boost",
      "label": "похожесть на голос",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 0,
      "max": 1,
      "disclosure": "essential",
      "default": 0.75
    },
    {
      "key": "style",
      "label": "экспрессия (style)",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 0,
      "max": 1,
      "disclosure": "essential",
      "default": 0
    },
    {
      "key": "speed",
      "label": "скорость",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 0.7,
      "max": 1.2,
      "disclosure": "essential",
      "default": 1
    }
  ]
}

avg/voice_isolate

Extract cleaner speech from mixed audio or a video soundtrack.

When to use it: A recording contains speech mixed with background sound.

Example: Connect the recording to звук/видео and listen separately to голос and фон.

Check the result: Compare the isolated speech with the original; inspect whether a background output was produced.

• UI: available in the studio. MCP: run_node does not support this node; isolate_audio accepts an owned recording.
• The background is a best-effort residual, not a guaranteed clean music/SFX stem.
• The source must be an owned studio file; the isolation call can charge.

Live ports and controls

{
  "label": "Изоляция голоса",
  "inputs": [
    {
      "name": "звук/видео",
      "type": "AUDIO,VIDEO"
    }
  ],
  "outputs": [
    {
      "name": "голос",
      "type": "AUDIO"
    },
    {
      "name": "фон",
      "type": "AUDIO"
    }
  ],
  "params": [
    {
      "key": "engine",
      "label": "движок",
      "kind": "select",
      "options": [
        {
          "value": "elevenlabs",
          "label": "ElevenLabs (чистая речь)"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "elevenlabs"
    }
  ]
}

avg/dub

Translate and revoice a recording into another language.

When to use it: A completed clip needs an alternate spoken-language track.

Example: Choose the target language, dub the source and inspect the resulting аудио.

Check the result: Listen for translation accuracy, speaker handling and retained background.

• UI: available in the studio. MCP: run_node does not support this node; dub_video accepts an owned recording and target language.
• The result is an audio track even when the input is video.
• Dubbing does not automatically match the visible lips; use a separate supported lipsync step if needed.

Live ports and controls

{
  "label": "Дубляж",
  "inputs": [
    {
      "name": "звук/видео",
      "type": "AUDIO,VIDEO"
    }
  ],
  "outputs": [
    {
      "name": "аудио",
      "type": "AUDIO"
    }
  ],
  "params": [
    {
      "key": "target_lang",
      "label": "на язык",
      "kind": "select",
      "options": [
        {
          "value": "ru",
          "label": "русский"
        },
        {
          "value": "en",
          "label": "английский"
        },
        {
          "value": "es",
          "label": "испанский"
        },
        {
          "value": "de",
          "label": "немецкий"
        },
        {
          "value": "fr",
          "label": "французский"
        },
        {
          "value": "it",
          "label": "итальянский"
        },
        {
          "value": "pt",
          "label": "португальский"
        },
        {
          "value": "pl",
          "label": "польский"
        },
        {
          "value": "uk",
          "label": "украинский"
        },
        {
          "value": "ja",
          "label": "японский"
        },
        {
          "value": "zh",
          "label": "китайский"
        },
        {
          "value": "ko",
          "label": "корейский"
        },
        {
          "value": "ar",
          "label": "арабский"
        },
        {
          "value": "hi",
          "label": "хинди"
        },
        {
          "value": "tr",
          "label": "турецкий"
        },
        {
          "value": "nl",
          "label": "нидерландский"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "ru"
    },
    {
      "key": "source_lang",
      "label": "исходный язык",
      "kind": "select",
      "options": [
        {
          "value": "auto",
          "label": "авто-детект"
        },
        {
          "value": "en",
          "label": "английский"
        },
        {
          "value": "ru",
          "label": "русский"
        },
        {
          "value": "es",
          "label": "испанский"
        },
        {
          "value": "de",
          "label": "немецкий"
        },
        {
          "value": "fr",
          "label": "французский"
        },
        {
          "value": "zh",
          "label": "китайский"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "auto"
    },
    {
      "key": "num_speakers",
      "label": "спикеров (0=авто)",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 0,
      "max": 10,
      "disclosure": "essential",
      "default": 0
    },
    {
      "key": "drop_background",
      "label": "убрать фон",
      "kind": "boolean",
      "options": [],
      "dynamic": false,
      "disclosure": "essential",
      "default": false
    }
  ]
}

avg/lipsync

Fit supplied speech audio to the visible speaker in an existing video.

When to use it: A clip already exists and its mouth movements need to follow a replacement track.

Example: Connect the clip to видео and Voiceover’s audio to звук; choose the duration-matching mode.

Check the result: Watch close-ups and the end of the clip; verify speech alignment and duration.

• UI: available in the studio. MCP: run_node does not support this node; lipsync_video accepts both a clip and audio.
• Both video and audio are required.
• Model and sync mode determine how mismatched lengths are handled; this does not translate or write the speech.

Live ports and controls

{
  "label": "Липсинк",
  "inputs": [
    {
      "name": "видео",
      "type": "VIDEO"
    },
    {
      "name": "звук",
      "type": "AUDIO,VIDEO"
    }
  ],
  "outputs": [
    {
      "name": "видео",
      "type": "VIDEO"
    }
  ],
  "params": [
    {
      "key": "model",
      "label": "модель",
      "kind": "select",
      "options": [
        {
          "value": "lipsync-2",
          "label": "lipsync-2 (баланс)"
        },
        {
          "value": "lipsync-2-pro",
          "label": "lipsync-2-pro (детали лица/зубы)"
        },
        {
          "value": "sync-3",
          "label": "sync-3 (максимум, 4K, экстрим-ракурсы)"
        },
        {
          "value": "kling-lipsync",
          "label": "Kling LipSync (3D и мультяшки, ролик 2–10 с)"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "lipsync-2"
    },
    {
      "key": "sync_mode",
      "label": "при разной длине",
      "kind": "select",
      "options": [
        {
          "value": "cut_off",
          "label": "обрезать длинное (cut_off)"
        },
        {
          "value": "loop",
          "label": "зациклить видео (loop)"
        },
        {
          "value": "bounce",
          "label": "туда-обратно (bounce)"
        },
        {
          "value": "silence",
          "label": "дотишить звук (silence)"
        },
        {
          "value": "remap",
          "label": "подогнать (remap)"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "cut_off"
    }
  ]
}

avg/upscale

Increase an image’s resolution with configurable creative reconstruction.

When to use it: An approved frame needs a larger deliverable.

Example: Connect an image, choose factor and creativity, then compare the result at full size.

Check the result: Inspect faces, fine edges and lettering for new details or distortions.

• UI: available in the studio. MCP: run_node does not support this node; upscale_image accepts an owned image.
• A higher resolution is not proof that original details were preserved.
• The cost depends on source dimensions; this node is not currently supported by run_node.

Live ports and controls

{
  "label": "Upscale",
  "inputs": [
    {
      "name": "картинка",
      "type": "IMAGE_FRAME"
    }
  ],
  "outputs": [
    {
      "name": "картинка",
      "type": "IMAGE_FRAME"
    }
  ],
  "params": [
    {
      "key": "factor",
      "label": "увеличение",
      "kind": "select",
      "options": [
        {
          "value": "2",
          "label": "×2"
        },
        {
          "value": "4",
          "label": "×4"
        }
      ],
      "dynamic": false,
      "disclosure": "essential",
      "default": "2"
    },
    {
      "key": "creativity",
      "label": "креативность",
      "kind": "number",
      "options": [],
      "dynamic": false,
      "min": 0,
      "max": 1,
      "disclosure": "essential",
      "default": 0.35
    }
  ]
}

Reviewed: 2026-10-14. Source: the current studio node catalog and authored workflow examples. Examples are instructions for a human review, not evidence of a new paid render. Continue with connections and inputs when a wire or input behaves unexpectedly.