Images

POST /v1/images/generations
GET  /v1/images/models

Body: { "model", "prompt", "n", "size", "aspect_ratio", "quality", "response_format", "image?", "mask?", "mode?" }.

Response: { "created", "data": [{ "b64_json" }] } or { "url" }.

Order

If PRIVATE_IMAGE_GEN_URL is set (or a dedicated image deployment FQDN exists), POST {base}/v1/images/generations on that OpenAI-compatible worker.
Else OpenAI (platform key or org BYOK) or Azure Foundry when those keys exist.
Else Vertex image models when ADC / project env is already configured (not branded in Studio).
Else 503 { "error": "Studio GPU offline" }.

Studio UI: Dashboard → Studio (/dashboard/studio). The browser only calls /api/studio/*, which proxies to this endpoint.

bash
curl http://localhost:3001/v1/images/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A red door on a white wall",
    "n": 1,
    "size": "1024x1024",
    "response_format": "b64_json"
  }'

PRIVATE_IMAGE_GEN_URL is optional and is not http://127.0.0.1:8188. Default kind is openai.

Batches

Videos