| name | create |
| description | Real media generation tools (image, video, TTS, SFX, music) exposed to this agent via an amiko MCP server, plus the rule against fabricating capabilities |
| metadata | {"openclaw":{"emoji":"🎨"}} |
Create Skill
You have five real media generation tools: create_image, create_video, create_tts, create_sfx, create_music. They are MCP tools, served by an amiko MCP server your agent connects to, not an external CLI and not something you shell out to.
In your tool list they appear under a per-agent namespaced prefix (the server id includes your agent id, since the auth header carries your own token), something like mcp__amiko-create-<agentId>__create_image. This doc refers to them by their base function name, use whatever exact namespaced name shows up in your own tool list.
THE RULE: never fabricate a capability
If you do not have a tool for what the user asks, say so plainly. NEVER invent a command, tool, job id, cost, or status. NEVER claim media is "generating" or "processing" unless a real create_* tool actually returned a job id. If a create_* tool returns an error, relay the real reason to the user, do not blame your own syntax or silently retry.
The tools
- create_image: generates an image. Args:
prompt, plus optional model, size.
- create_video: generates a video. Args: optional motion
prompt, optional firstFrameImage (image URL), plus optional model, resolution, seconds, aspectRatio. Default model is I2V (MiniMax-Hailuo-2.3-Fast): you must pass firstFrameImage (and may add a motion prompt). Hailuo 2.3 and 2.3-Fast both need a first frame. For prompt-only (T2V) video, pass a T2V-capable model such as MiniMax-Hailuo-02 (do not call with prompt alone under the default model; the server rejects it).
- create_tts: generates spoken audio from text. Args:
prompt, plus optional model, voiceId.
- create_sfx: generates a sound effect. Args:
prompt, plus optional durationSeconds.
- create_music: generates a music track. Args:
prompt and/or lyrics, plus optional model, durationMs, isInstrumental.
Defaults for optional args are applied server-side. Image, TTS, SFX, and music usually need only prompt. Video is different: under the default I2V model, pass firstFrameImage; for prompt-only video, also pass a T2V-capable model (e.g. MiniMax-Hailuo-02).
How generation works
Calling a create_* tool is asynchronous and charged only on success. The tool call returns a job id, and a placeholder appears in the chat right away; it resolves into playable media once generation finishes, or into an error if it fails. A generation that fails is never billed.
When the tools can't do it
If the user wants media beyond what these five tools support, tell them to use the web Create Studio at platform.heyamiko.com/create.
