Skip to main content
POST
cURL
Enhance a raw user prompt into a detailed, model-friendly prompt using the AI guide that matches the target model and inference type.
  • Set type (v2 dot notation) and model_slug to pick the right guide.
  • For images.edits and videos.animations, also send the source image (URL, data-URI, base64, or file).
  • Synchronous — the enhanced prompt is returned directly in the response body ({ prompt, negative_prompt }).
Required fields: prompt, type, model_slug (plus image for images.edits / videos.animations). The returned negative_prompt is provided for compatibility and is ignored by models that do not support negative prompts (e.g. FLUX.2 Klein).

Accepted type values

images.generations, images.edits, videos.generations, videos.animations, audio.speech, audio.music Only prompt-driven inference types have a booster. Upscaling, background removal, OCR, transcription and embeddings are not supported and are rejected with 422.

Example (image-to-image)

Then send the returned prompt to the matching generation endpoint (e.g. POST /api/v2/images/edits).

Authorizations

Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

multipart/form-data
prompt
string
required
Minimum string length: 3
Example:

"A beautiful landscape with mountains"

type
string
required

Inference type in v2 dot notation (e.g. images.generations, videos.animations, audio.speech)

Example:

"images.generations"

model_slug
string
required
Example:

"Flux1schnell"

negative_prompt
string | null
Minimum string length: 3
Example:

"blurry, low quality"

Response

Enhanced prompts returned successfully

prompt
string
negative_prompt
string | null