AI image generation API
Call image models from several providers with one API key. POST /v1/image/generations returns a task id; fetch the result from GET /v1/image/generations/{id}. Every model page lists its reference limits, resolutions and output formats.
How it works
- Sign up and create an API key in the console.
- Submit the model ID, prompt and optional images (public URLs), plus resolution and output format.
- Poll the task and download the image from data.result_url once data.status is SUCCESS.
Available models
- Dola Seedream 5.0 Pro — Dola Seedream 5.0 Pro is the international Seedream 5.0 Pro image model: t2i creates images from text and i2i edits or restyles 1–10 reference images, producing one 1k or 2k image per request.
- Seedream 5.5 — Seedream 5.5 is the newest Seedream image model: t2i creates images from text and i2i edits or restyles 1–10 reference images, with transparent backgrounds, one image per request at 1k, 1.5k or 2k. seedream-v5.5 runs in China and dola-seedream-5.5 is the international version.
- Gemini 4 Argon — Gemini 4 Argon supports text and image input, tool calling, and streaming, with a 1,048,576-token context window.
- Google Nano Banana Pro — NB Pro is the Nano Banana Pro image model: text-to-image and editing with up to 14 reference images, one image per request at 1k, 2k or 4k.
- GPT Image 2.5 — Image G v2.5's low-price extension supports text-to-image and reference-image editing at 1k, 2k or 4k, with one output per request. The upstream upgrade is rolling out; the actual version depends on the account rollout.
- GPT Image 2.5 Sunburst — GPT Image Sunburst (v2.5 and v2.6) is the GPT Image line tuned for precise editing: text-to-image or editing with up to 16 reference images, 1–4 outputs per request at 1k, 2k or 4k.
- GPT Image 2.6 — GPT Image 2.6 low-price creates and edits images with up to 16 references, one image per request at 1k, 2k or 4k, with a choice of the Flare or Sunburst version.
- Seedream 5.0 Pro — Seedream 5.0 Pro is the China version of the Seedream 5.0 Pro image model: t2i creates images from text and i2i edits or restyles 1–10 reference images, producing one 1k or 2k image per request, billed in CNY.
FAQ
Is it the same as OpenAI's /v1/images/generations?
No. It is an async task API: submit to /v1/image/generations and poll for the result. Each model page lists its exact fields.
How do I send reference images?
Put publicly reachable image URLs in the images array. Count and size limits depend on the model.
How is it billed?
Per generation. The task result reports the actual charge, and each model page shows the current price.