Sora API shut down: pick a replacement video model by what you used
The Sora API closed on 2026-09-24. Map your use (length, audio, image start, references) to Seedance, Wan, Kling, MiniMax or Gemini Omni Flash on Sume.

The Sora API closed on 2026-09-24, so a pipeline that still calls it needs a new model id. There is no one-to-one replacement. The practical move is to list what your Sora clips needed (length, audio, a start image, resolution, a reference) and pick the Sume catalog row that covers those, starting with Seedance 2.5, Wan 3.0, Kling 3, MiniMax H3 or Gemini Omni Flash 1.1. News coverage describes the API closure as the second stage of a shutdown that began with the app in April 2026.
Map your old requirements to a row
Do not copy parameters across. Each Sume model validates its own envelope, so a 12-second request that worked on one row can fail on another.
| What your Sora calls needed | Start with | Why |
|---|---|---|
| Prompt to clip, up to 30 s | seedance-2.5 or wan-3.0 | 30 s ceiling, 480p to 1080p |
| Photo to clip | kling-3, seedance-2.5, wan-3.0 | first-frame and end-frame image-to-video |
| Spoken audio from the model | minimax-h3 or gemini-omni-flash-1.1 | audio always on |
| Reference images or clips | wan-3.0, minimax-h3, gemini-omni-flash-1.1 | reference-to-video |
| Editing an existing video | gemini-omni-flash-1.1 | video_url edit mode |
| Highest resolution | gemini-omni-flash-1.1 | up to 4K, 3 to 10 s |
Port the integration
Submit asynchronously, poll the job or take a webhook, then download the result. On Sume that is POST /v1/videos with a bare model id such as seedance-2.5, then GET /v1/videos/{jobId} until completed. Send an Idempotency-Key so retries reuse the job.
Re-test your prompts. Models differ in how they follow camera direction and text, and no seed means you cannot reproduce an old clip, only make a new one.
What not to assume
Do not assume your old clip lengths or aspect ratios exist on the new row, and do not assume audio. Read the model list and handle the 400s explicitly. A first run at 480p on two candidate models costs little and answers most of the question.
Sources
Related posts
More in Comparisons
- Tagalog text to speech API: Sume tl voices vs Eleven v4 Filipino
Sume's Voices library tags voices with tl for Tagalog; Eleven v4 lists Filipino among 90+ languages. What to send, and how to check the voice speaks it.
- Veo 3.1 Lite vs Wan 3.0: price per second at 480p, 720p and 1080p
Google lists Veo 3.1 Lite at $0.05 (720p), $0.08 (1080p) per second; fal lists Wan 3.0 at $0.05 (480p), $0.10 (720p), $0.20 (1080p).
- Veo 3.1 tiers vs Seedance 2.5: price per second of video
Google lists Veo 3.1 at $0.40 (Standard), $0.10-$0.12 (Fast) and $0.05-$0.08 (Lite) per second; Seedance 2.5 on fal is about $0.46 at 720p. Sume x 1.25: $0.58.
- Wan 3.0 or MiniMax H3 Max: which to pin for reference-to-video
Both take image, video and audio references on Sume. Wan runs 2 to 30 seconds with 5 reference videos; H3 Max runs 5 to 15 seconds with always-on stereo audio.
Written by Sume