Quickstart

Two calls, no key. Pick any string as the client id; a new one per image is what the clients do.

1. Submit

curl -s https://veida.ai/api/ai/generate \
  -H 'content-type: application/json' \
  -H 'x-anon-id: my-test-123' \
  -d '{"provider":"kie","mediaType":"image","model":"veida-image-v1","scene":"text-to-image",
       "prompt":"a paper boat on a calm lake at sunrise, soft light",
       "options":{"aspect_ratio":"16:9"}}'

Success is {"code":0,"data":{"id":"<task id>", ...}}. The quota wall is also HTTP 200 with code 0: {"code":0,"data":{"wall":true,"reason":"anon_ip_daily"}}. Check data.wall before reading data.id, or the failure surfaces as a missing task id.

2. Poll

curl -s 'https://veida.ai/api/ai/anon-query?taskId=<task id>&provider=kie&mediaType=image' \
  -H 'x-anon-id: my-test-123'

Poll every few seconds with the same client id. When data.images[0] is present it is the permanent CDN URL, and data.watermarked[0] says whether a mark was drawn. data.status moves between pending and processing and is not monotonic; only failed or error is terminal, and that means the content filter refused the prompt. A run usually takes 40 to 120 seconds.

Get started

Clients

1. Submit

2. Poll

​1. Submit

​2. Poll

1. Submit

2. Poll