Every endpoint runs on the same optimized runtime that powers Image Studio — which is why the prices look like typos. Credits are shared across the API and the Studio.
| Endpoint | What it does | Price |
|---|---|---|
POST /v1/edit |
Instruction-led image editing (Qwen Image Edit 2511). Prompt + image in, edited image out. Supports multi-image reference, negative prompt, seed, size. | $0.002 / edit |
POST /v1/edit/batch |
Batch editing with a shared context image. One instruction + one reference applied across up to 200 images per job. | $0.002 / image |
POST /v1/camera |
Camera-control reshoot. Subject image + azimuth/elevation (or presets) in, up to 12 consistent angles out. | $0.002 / angle |
POST /v1/upsample |
Upsample any result to 4K. | $0.004 / image |
POST /v1/generate Coming soon |
Text-to-image on the Z-Image runtime. | — |
POST /v1/video Coming soon |
Wan2.2 text-to-video and image-to-video on GGUF-quantized runtimes. | — |
POST /v1/3d Coming soon |
TRELLIS.2 image-to-3D with GLB export. | — |
Standard edit = 1120×736 output, ~13s latency. Larger outputs bill proportionally. Fair-use rate limits apply on the free tier; paid tokens get priority queueing. Placeholder limits — set these to match your capacity before launch.
Sign in with Google and your personal token is issued instantly — the same token system as our Colab wheels.
One HTTPS request with your token in the Authorization header. Send an image URL or base64, plus a plain-English instruction.
The response includes a hosted URL for the edited image, the seed used, and your remaining credit balance.
Endpoint paths and payload fields shown here are the proposed v1 contract — adjust to match your backend before publishing.
Buy credits once, spend them anywhere — in the browser Studio or through the API. 1 credit = 1 standard edit. No monthly commitment, credits never expire.
Need volume beyond this — millions of edits, dedicated capacity, or the Triton kernels for your own serving stack? Talk to us.