Sora shutdown audit: which feature did you use, what replaces it
OpenAI shut down Sora 2 and the Videos API on September 24 with no replacement. A feature-by-feature checklist of what Sume covers and what it does not.
Start by listing which Sora features your code called, because OpenAI's own page says there is no one-to-one replacement and the shutdown on September 24, 2026 has passed. Text-to-video, image-guided clips and long-form assembly have a path on Sume; character uploads and the Sora editing endpoint do not map one for one.
What did OpenAI list as gone?
The video generation guide I read on 2026-10-02 names these Sora capabilities as no longer available: text-to-video, image references, character uploads for consistency, video extensions, editing (which replaced the older remix endpoint) and Batch API support. The deprecations page lists sora-2, sora-2-pro and the dated snapshots, with the Videos API, and states the replacement as none.
Which Sume feature covers each one?
Read the table as a starting point, then check the model's catalog entry, because Sume documents limits per model.
| Sora feature | Sume path | Gap to check |
|---|---|---|
| Text-to-video | POST /v1/videos with a catalog model or sume/auto | Pick durations from supported_durations |
| Image-guided clip | Image-to-video through the video catalog | Model-specific frame image inputs |
| Offline batch queue | Async jobs, queue, signed webhooks | No batch discount documented |
| Character uploads | Reference images on models that accept them | Check which reference inputs the chosen model accepts |
| Extensions and editing | Longer clips and Timeline assembly | Not an in-place extend call |
What are Sume's real limits?
The video docs say limits are per model: for example seedance-2.5 accepts 4 to 30 seconds at 480p, 720p and 1080p, and Auto defaults to 720p and 8 seconds with 3 to 10 second clips at 16:9 or 9:16. size returns 400 unsupported_parameter on v1 models that report supported_sizes: null, so use resolution plus aspect_ratio.
Query GET /v1/video-router/models for the live catalog instead of copying numbers from a blog.
What should I do this week?
Grep your code for the retired model ids, the /v1/videos OpenAI path, and any code that stored Sora job ids. Decide per feature whether you rebuild, drop it or pin a Sume catalog id. Test one real prompt per feature, and check the job result rather than assuming parity with Sora output.
For a deeper per-feature comparison, read Sora API shutdown and video migration.
Sume does not promise Sora-looking output, and no model here replicates a Sora clip. Judge the results on your own prompts.
Sources
Related posts
More in Use cases
- Wix Stores wants 3000x3000 images: check your model's cap
Wix Stores lists 3000 x 3000 px as its minimum product image. Check GET /v1/images/models for each model's largest size before you generate for it.
- X Ads card video: 10 minutes, 500 MB, MP4 or MOV, 16:9 or 1:1
X's Ads API docs list card video up to 10 minutes and 500 MB, MP4 or MOV, 16:9 or 1:1, and images up to 3 MB. Here is how to fit a Sume clip to that.
- YouTube ads need a business name and logo from Oct 12, 2026
From October 12, 2026 YouTube and partners line items in DV360 need a business name (25 characters) and a square logo. Specs, and why not to generate the logo.
- What happens if I don't label AI video on YouTube? Penalties
YouTube says missing an AI disclosure can lead to a manual label, removal or Partner Program suspension. What triggers it and how Sume users avoid it.
Written by Sume