13 days after the Sora Videos API removal: the Sume /v1/videos route
OpenAI removed the Sora Videos API on September 24 with no replacement. What a team that still needs API video gets from Sume: submit, poll, download.
OpenAI lists the Sora Videos API removal for September 24, 2026 and names no replacement. If you still need video from an API, Sume offers an asynchronous /v1/videos route: submit, poll, download. It serves other vendors' models, not Sora, so prompts and durations must be re-tested rather than ported blindly.
What the OpenAI page says
The OpenAI deprecations page, read on October 7, shows the Sora Videos API with a removal date of September 24, 2026 and no recommended replacement. That makes today day 13 after the removal.
The Sume shape
The Videos docs describe four steps, and say the surface agrees field for field with the OpenRouter Video Generation API apart from a short list of differences.
| Step | Call | You get |
|---|---|---|
| Submit | POST /v1/videos | Job id and polling_url immediately |
| Poll | GET /v1/videos/{jobId} | status: pending, in_progress, completed, failed or cancelled |
| Download | GET /v1/videos/{jobId}/content | The mp4 once completed |
| Discover | GET /v1/videos/models | Resolutions, ratios, durations, audio and price per model |
What to re-test
Every catalog model has its own limits, so read them before you resubmit old prompts.
- Durations differ by model: seedance-2.5 takes 4 to 30 seconds, wan-3.0 takes 2 to 30, minimax-h3 takes 5 to 15, and gemini-omni-flash-1.1 takes 3 to 10.
- Sume uses bare model ids, with no provider prefix.
- Sume does not accept a size field on video; use resolution and aspect_ratio.
- A seed field is not on every model; check the seed flag in the model row.
Sources
Related posts
More in Developers
- 19 Omni jobs or 7 Wan jobs: Timeline render strategy for 181 s
A 181-second Reel is 7 Wan clips or 19 Omni clips. Timeline render.strategy auto chunks past 12 segments and single is refused above 12.
- 1920x1088 works on GPT Image 2.5 where 1920x1080 fails: the 16 rule
1080 is not a multiple of 16, so a custom 1920x1080 size is rejected on ChatGPT Image 2.5. Use 1920x1088 or 1536x864, then trim. Prices included.
- 21:9 banner still to video: only Seedance and MiniMax H3 list 21:9
A 21:9 still can only become a 21:9 clip on Seedance or MiniMax H3 in Sume's catalog. GPT Image 2.5 reaches 21:9 with a 2352x1008 custom size.
- A 24-second Reel as three 8-second sume/auto jobs: $3.00
sume/auto is Gemini Omni Flash 1.1 at 3 to 10 s and 9:16. Three 8-second jobs cost $3.00 at 720p, chained by passing each last frame as the next first_frame.
Written by Sume