Creative vs Verbatim Mode

Understand how Wonda handles prompts on the current public API surface.

The two modes

Mode	Prompt handling	Best for
Creative	Wonda may enhance, rewrite, or expand a prompt before it reaches a model.	Exploration, brainstorming, and end-user prompts that need improvement.
Verbatim	Your prompt is sent to the model exactly as provided.	Deterministic pipelines, tests, and prompts produced by your own system.

Current public API behavior

Direct generation endpoints are verbatim. When you call endpoints such as POST /image/generate, POST /video/generate, POST /text/generate, POST /music/generate, or POST /audio/speech, the prompt field is used as provided.

No request flag is required to enable verbatim mode on direct endpoints.

Surface	Default mode	How to switch
Direct generation endpoints	Verbatim	No switch needed.
CLI generation commands	Verbatim for the prompt passed to the direct endpoint.	No switch needed.

Recommendation for API consumers

If you are building an automated pipeline where prompts are pre-crafted or generated by your own system, use direct generation endpoints. This keeps outputs predictable and makes request logs easier to audit.

Use creative prompt rewriting only in product surfaces where you explicitly control that behavior before calling the API.

Example: direct image generation

curl -X POST https://api.wondercat.ai/api/v1/image/generate \
  -H "Authorization: Bearer sk_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-2",
    "prompt": "16:9 image of a golden retriever on a beach, 35mm photo, late afternoon light",
    "params": {
      "aspectRatio": "16:9"
    }
  }'

The model receives the prompt string in the request body.