Sora API shutdown Sept 24, 2026: what Omni 1.1 Flash covers on Sume
OpenAI lists the Videos API and sora-2 models for removal on Sept 24, 2026, with no replacement named. Omni 1.1 Flash on Sume: 3-10 s, 4K, edit, end frame.
OpenAI's deprecations page lists the Videos API and the sora-2 and sora-2-pro models, including dated snapshots, for removal from the API on September 24, 2026, and names no replacement (OpenAI, read 2026-10-05). If you are moving calls off it, Gemini Omni Flash 1.1 on Sume covers text-to-video, image-to-video with an end frame, references and a video edit mode, at 3 to 10 seconds.
This post uses only the vendor's deprecation entry for Sora, and it does not compare quality. It is a checklist of what you need to map, with what Sume's Omni row offers on each line.
What OpenAI's page says
The entry says that on March 24, 2026 OpenAI notified developers using the Videos API and the Sora 2 model aliases and snapshots of their deprecation and removal on September 24, 2026. The replacement column is empty. That means there is no one-line swap, and the migration is yours to design: pick a model per use, check limits, and re-test the prompts.
The mapping checklist
Each row is a question about your current calls and the Sume field that answers it. Facts are from the Video Router page and Google's launch post (read 2026-10-05).
| You need | Omni on Sume | Field |
|---|---|---|
| Text to video | 3-10 s, 16:9 or 9:16 | prompt, duration, aspect_ratio |
| Start from an image | Yes | image_url |
| End on an image | Yes | end_image_url |
| Character or product consistency | Up to 10 images, 3 clips of 3 s | reference_image_urls, reference_video_urls |
| Edit an existing clip | Yes, prompt-driven | video_url |
| Resolution | 360p, 720p, 1080p, 4K | resolution |
| Sound | Native, always on | No toggle |
| Clips longer than 10 s | Chain clips | New job per clip |
Prices to plan with
Sume bills the provider list times 1.25 per output second, rounded up to the cent per clip.
| Resolution | 10 s clip |
|---|---|
| 360p | $0.38 |
| 720p | $1.25 |
| 1080p | $1.88 |
| 4K | $3.75 |
What this does not tell you
A removal date is a fact. Quality, prompt style and clip look are not, and this post makes no claim about how Omni output compares with Sora output. Test with your own prompts. The 360p draft tier makes that cheap: twenty prompts at 3 seconds are $2.40 at 360p, and the same twenty at 1080p are $11.40.
If your product shipped with Sora's native clip lengths, the mapping can be the harder part than the model. Omni's range is 3 to 10 seconds, so a longer shot needs to be split into clips or joined on a Timeline.
Where each OpenAI-era habit goes
Sume's flow is submit, receive a job id and polling URL, poll until completed, then download. For editing an existing video, the Sume feature is Omni's video_url edit mode, which rewrites a clip from a prompt. To start a shot from one image, use image_url. For several images that must stay consistent, use references with <IMAGE_REF_0> style addresses.
The differences that bite are limits: 16:9 and 9:16 only, 3 to 10 seconds, native audio you cannot turn off, and an edit mode that takes no references. Write these down as test cases before you switch traffic.
A migration plan
- List every prompt your app sends with its duration and aspect ratio. Anything over 10 seconds becomes two clips.
- Run all prompts at 360p first. A 3-second draft is $0.12.
- Rewrite prompts that depended on features Omni does not list, then re-run at the target resolution.
- Add an
Idempotency-Keyto each call and log job ids, so the new pipeline is traceable. - Switch traffic only after a side-by-side review of your own clips.
Sources
Related posts
More in Comparisons
- Spatius client-side Gaussian splatting avatar vs a finished MP4
Spatius renders avatars on the client with 3D Gaussian splatting and needs its SDK to show them. A finished Sume MP4 plays in any video element, no SDK needed.
- Speech-to-text price per hour, ranked: nine rates and Sume STT
Nine STT rates in dollars per audio hour, from Cloudflare's $0.03 to Sume STT 1.0 at $0.60, each read from the vendor's own page with the conversion shown.
- Strands Decider 2B vs Cloudflare Clef for gating agent runs
Both return typed answers (yes/no, choice, score) with confidence. Compare what each vendor page says, and what stays your job when a Sume run costs money.
- 3 days of Stripe retries vs 10 Sume attempts 30 seconds apart
Stripe retries live webhooks for up to 3 days; Sume tries 10 times, 30 seconds apart, about 4.5 minutes. After that, poll the job or call redeliver.
Written by Sume