AI Generation Endpoints

Direct generation API for images, video, transcription, and prompt enhancement.

List models

GET /api/ai/models

Returns:

{
  "image": [
    { "id": "openai/gpt-image-1", "label": "GPT Image 1", "provider": "openai" },
    { "id": "wavespeed/flux-2", "label": "FLUX.2 Dev [Wavespeed]", "provider": "wavespeed" },
    …
  ],
  "video": [ … ],
  "chat":  [ … ]
}

Only models for providers with a configured key are returned.

Generate image

POST /api/ai/generate
{
  "board_id": "…",
  "model": "wavespeed/flux-2",
  "prompt": "…",
  "aspect_ratio": "16:9",
  "input_image_url": "/media/…",   // optional
  "negative_prompt": "…",          // optional
  "params": { … }                  // model-specific
}

Returns:

{
  "element": { "id": "…", "type": "image", "url": "/media/…" },
  "history_entry_id": "…",
  "cost_estimate": 0.012
}

The element is already added to the board.

For video models, use the same endpoint with a video model ID — the response includes a status: "queued" payload for providers (like Kling) that queue jobs. Poll with GET /api/ai/jobs/{job_id}.

Prompt enhancer

POST /api/ai/prompt-enhancer/analyze
{
  "prompt": "…",
  "reference_image_url": "/media/…",
  "model": "anthropic/claude-sonnet"
}

Returns:

{ "enhanced_prompt": "…" }

Transcribe audio

POST /api/ai/transcribe
{
  "board_id": "…",
  "element_id": "…",
  "language": null  // auto-detect
}

Returns:

{
  "text": "…",
  "language": "en"
}

Where to go next

• HTML Render Endpoints
• Agent Endpoints
• API Overview