gengastudio.com WebMCP Tools, Schemas & Agent Capabilities

gengastudio.com

live Unverified (seed import) AI & Agents 50 tools imperative implementation

Genga Studio is a cutting room for AI-generated animation. The page publishes every one of its actions as a WebMCP tool, so an agent drives the same buttons you would click.

First seen 2026-11-05 · Last seen 2026-11-05 · Source: webmcp.com

Answer 10 · Act 40 · Transact 0 · Score: Not yet scored

Tool inspector

generate_takes

act imperative

Generate new takes for a shot — stills, restyles of an existing take, or animated cel clips. Opens the Generate console, fills it, and submits one job per take with a fresh seed. Count 1–4 (default 3); paid backends need confirm_cost. ANIMATE STARTS FROM THE KEEPER STILL — to animate another, pass source (a path or "selected"). Readable text needs model:"pro"; attach_reference pins a face or place, restyle edits a frame. Motion results carry situational guidance.

Page: /app

Input schema

Property	Type	Required	Description
`shot`	string	yes	The shot: a sid (B10-S2), its number in the cut, a beat, or a description like "the Ross close-up".
`lane`	string	no	still = new image · restyle = image-to-image on an existing take · animate = motion clip. Default still.
`count`	integer	no	How many takes to submit, 1–4. Words like "a few" (3) or "a couple" (2) are accepted too.
`prompt`	string	no	Prompt text. Omit to use the shot's written image or motion prompt exactly as the director wrote it.
`prompt_mode`	string	no	replace = use only your prompt (default) · append = add yours to the shot's own prompt.
`source`	string	no	Which image to start FROM: a take path, "selected", "third still" or "keeper". Animate defaults to the keeper; restyle to the selected take.
`source_take`	string	no	For restyle: the take to restyle. A path, or "latest", "newest still", "keeper". Defaults to the selected take. Same as source.
`denoise`	number	no	Restyle strength, 0.35–0.95. 0.55 keeps the layout, 0.85 restyles. Default 0.85.
`region`	array	no	For animate: the cel region as [left, top, right, bottom]. Omit to animate the full frame.
`seconds`	number	no	For animate: clip length in seconds. Defaults to the backend's own clip length and is clamped to what it supports.
`frames`	integer	no	For animate: exact frame count, when you need one. Normally leave it and pass seconds instead.
`live_seconds`	number	no	For animate: freeze after this many seconds. Only for a model that drifts after N seconds — clips play in full otherwise.
`seeds`	array	no	Exact seeds to use, one per take. Omit for fresh random seeds (the usual case).
`backend`	string	no	Force a specific backend id instead of the project's lane default (e.g. "mock", "comfyui", "fal").
`model`	string	no	Force a model. Stills: "pro" for readable text, "flash" (cheap default). Motion: "seedance", "wan". A full endpoint id works too.
`confirm_cost`	boolean	no	Set true to approve a paid backend. Required whenever the lane resolves to a backend that bills money.
`references`	array	no	One-off reference images for this call only: [{image, role}], role character/prop/setting/style. attach_reference makes one stick.

Raw JSON schema

{
  "type": "object",
  "properties": {
    "shot": {
      "type": "string",
      "description": "The shot: a sid (B10-S2), its number in the cut, a beat, or a description like \"the Ross close-up\"."
    },
    "lane": {
      "type": "string",
      "enum": [
        "still",
        "restyle",
        "animate"
      ],
      "description": "still = new image · restyle = image-to-image on an existing take · animate = motion clip. Default still."
    },
    "count": {
      "type": "integer",
      "minimum": 1,
      "maximum": 4,
      "default": 3,
      "description": "How many takes to submit, 1–4. Words like \"a few\" (3) or \"a couple\" (2) are accepted too."
    },
    "prompt": {
      "type": "string",
      "description": "Prompt text. Omit to use the shot's written image or motion prompt exactly as the director wrote it."
    },
    "prompt_mode": {
      "type": "string",
      "enum": [
        "replace",
        "append"
      ],
      "description": "replace = use only your prompt (default) · append = add yours to the shot's own prompt."
    },
    "source": {
      "type": "string",
      "description": "Which image to start FROM: a take path, \"selected\", \"third still\" or \"keeper\". Animate defaults to the keeper; restyle to the selected take."
    },
    "source_take": {
      "type": "string",
      "description": "For restyle: the take to restyle. A path, or \"latest\", \"newest still\", \"keeper\". Defaults to the selected take. Same as source."
    },
    "denoise": {
      "type": "number",
      "description": "Restyle strength, 0.35–0.95. 0.55 keeps the layout, 0.85 restyles. Default 0.85."
    },
    "region": {
      "type": "array",
      "items": {
        "type": "number"
      },
      "description": "For animate: the cel region as [left, top, right, bottom]. Omit to animate the full frame."
    },
    "seconds": {
      "type": "number",
      "description": "For animate: clip length in seconds. Defaults to the backend's own clip length and is clamped to what it supports."
    },
    "frames": {
      "type": "integer",
      "description": "For animate: exact frame count, when you need one. Normally leave it and pass seconds instead."
    },
    "live_seconds": {
      "type": "number",
      "description": "For animate: freeze after this many seconds. Only for a model that drifts after N seconds — clips play in full otherwise."
    },
    "seeds": {
      "type": "array",
      "items": {
        "type": "integer"
      },
      "description": "Exact seeds to use, one per take. Omit for fresh random seeds (the usual case)."
    },
    "backend": {
      "type": "string",
      "description": "Force a specific backend id instead of the project's lane default (e.g. \"mock\", \"comfyui\", \"fal\")."
    },
    "model": {
      "type": "string",
      "description": "Force a model. Stills: \"pro\" for readable text, \"flash\" (cheap default). Motion: \"seedance\", \"wan\". A full endpoint id works too."
    },
    "confirm_cost": {
      "type": "boolean",
      "description": "Set true to approve a paid backend. Required whenever the lane resolves to a backend that bills money."
    },
    "references": {
      "type": "array",
      "items": {
        "type": "object"
      },
      "description": "One-off reference images for this call only: [{image, role}], role character/prop/setting/style. attach_reference makes one stick."
    }
  },
  "required": [
    "shot"
  ],
  "additionalProperties": false
}