Which Sume image route to call in October 2026
Use POST /v1/images for new work. Image 1.0 is retiring soon and image-router is deprecated. Here is the route table, with what each still does.
For new image work, call POST /v1/images. It takes a catalog model id or model: "sume/auto". The older routes still answer, but Image 1.0 is marked retiring soon and the legacy /v1/image-router/generate is deprecated and will not gain new parameters.
That is the guidance in Sume's Image API and Image 1.0 docs as read on 2026-10-02. New models, such as ChatGPT Image 2.5, appear on the new surface.
What does each route do today?
Image 1.0 keeps its legacy request shape, including avatar references and transparency, and uses the same Auto model selection: it returns job.model: "sume/auto".
| Route | Status | Notes |
|---|---|---|
POST /v1/images | Current | Catalog id or sume/auto; data[].url; 202 job envelope past 30 s |
POST /v1/image-1.0/generate | Retiring soon | Compatibility alias for Auto; image_urls 1-10; legacy transparency |
POST /v1/models/sume/image-1.0/runs | Retiring soon | Same body as Image 1.0 |
POST /v1/image-router/generate | Deprecated | Works, unchanged, no new parameters |
What changes when I migrate?
Reference images move from image_urls to input_references, with up to 16 on ChatGPT Image 2.5; n and format become the documented names; mode defaults to sync on /v1/images with a 30-second wait. Image 1.0 quality defaults to low, while ChatGPT Image 2.5 on /v1/images defaults to high, so set it explicitly when you compare cost.
Should I use Auto or pin a model?
Use sume/auto when you want Sume to choose; it never discloses which family ran, so you cannot rely on a specific look from run to run. Pin a catalog id from GET /v1/images/models for a series that must match. Both paths use the same job envelope, so polling code does not change. Details are in the Image API docs and Image 1.0 docs.
Sources
Related posts
More in Developers
- Which Sume video models accept 30 seconds? A Python filter
Ask GET /v1/videos/models which models list a 30-second duration, instead of trusting a launch post. A short Python script, plus what the docs name.
- Sume video tools: public URL or media import first? Per tool
Video captions takes a public HTTPS URL; trim, filter, inspect, frames, compose and detach need a workspace media.sume.com clip. A tool-by-tool input guide.
- Which voice does my avatar speak with? Check voice.status is ready
Sume TTS speaks in an avatar's voice when voice.status is ready. List avatars, check voice.status, then send avatar_id or avatar_handle on the TTS request.
- Will my clip fit Runway Enhance Frame Rate? Probe it first
Runway lists 4K and 5 minutes as Enhance Frame Rate limits. Sume's Video inspect probe returns duration, size and fps for free, so you can check before you.
Written by Sume