Luma Ray 3.2 video-to-video: 20 s, 64 keyframes, vs Omni edit on Sume
Luma's Ray 3.2 video-to-video keeps the source length up to 20 s and takes up to 64 keyframes. Sume does not list Luma; its edit route is Gemini Omni Flash 1.1.

Luma's Ray 3.2 video-to-video page says the output keeps the source clip's duration, up to 20 seconds, at 360p, 540p, 720p or 1080p, and accepts up to 64 keyframes. Sume does not list a Luma model, so the nearest Sume route for editing an existing clip is gemini-omni-flash-1.1 through the Video Router video_url field, priced at $0.125 per second at 720p.
Luma's published limits
These come from Luma's own pages. Ray 3.2 was introduced on 2026-06-09 and lists clips up to 20 seconds at 1080p and up to 16 keyframes for generation.
| Feature | Limit |
|---|---|
| Video to video: output duration | Same as the source, max 20 seconds |
| Video to video: resolutions | 360p, 540p, 720p, 1080p |
| Video to video: keyframes | Up to 64 |
| Generation: clip length at 1080p | Up to 20 seconds |
| Generation: keyframes | Up to 16 |
What Sume's edit route does
Sume's Video Router documents one edit model. You send video_url and a prompt; resolution is optional and defaults to 720p; aspect_ratio and duration are not accepted for the edit. The source is the edit source, not a reference, so it cannot be combined with image_url, end_image_url or reference_*_urls. Native audio is always on.
| Item | Value |
|---|---|
| Input | video_url plus a prompt |
| Resolution | 360p, 720p (default), 1080p or 4K, set with resolution |
| Output length | Follows the source clip |
| Sume per second | $0.0375 at 360p, $0.125 at 720p, $0.1875 at 1080p, $0.375 at 4K |
| 10 s source at 720p | 10 x $0.125 = $1.25 |
Where they differ
Luma accepts a source up to 20 seconds and up to 64 keyframes to steer the result. The Sume page does not state a maximum source length for the edit, and it has no keyframe input: you steer with the prompt alone. If your job needs per-keyframe control over a long clip, Sume's edit route does not replace Luma's.
If your job is a plain instruction edit, such as swapping a product or changing a background, the Omni route fits and the price is simple to compute: seconds of source times the per-second rate. The Omni edit price page shows where that model ranks on the editing board.
A cost example
A 12 second source clip edited at 720p through the Omni route costs 12 x $0.125 = $1.50. At 1080p it is 12 x $0.1875 = $2.25. Sume reserves the amount at submit using the length you give, and the final charge follows the source length.
If the clip is longer than the ten seconds Omni generates elsewhere, check the edit limits in the API reference before you send it, since this page does not state a source-length maximum for the edit. Where Luma's 64 keyframes are the point, buy that job from Luma.
What to send
Keep the source clip under the limit you have checked, host it at a public https URL, and write the instruction as a single change. If you need many changes, run them as separate edits so you can see which one broke the result.
Sources
Related posts
More in Comparisons
- Luma Ray3.2 5 s at 1080p is $1.20: the same clip priced on Sume models
Luma lists Ray3.2 at $1.20 for 5 s at 1080p and $0.30 at 720p. Sume lists no Luma model; a 5 s 1080p clip is $0.70 to $7.11 on Sume models.
- MAI-Transcribe-2-Streaming is a preview with no SLA: add a fallback
Microsoft marks MAI-Transcribe-2-Streaming public preview with no SLA. Record call audio and keep Sume STT as a recorded fallback at about one cent per minute.
- MAI-Transcribe-2-Streaming vs Sume STT: 100 hours, $54 vs $60
A tracker lists MAI-Transcribe-2-Streaming at $0.54 per hour (introductory through 2026) and Sume STT is $0.01 per audio minute, or $0.60 per hour.
- MAI-Voice-2.1 and MAI-Transcribe-2: what Sume lists for voice and STT
Microsoft AI's news page names MAI-Voice-2.1 and MAI-Transcribe-2. Sume lists neither; it lists Sonic TTS (38 micros a character) and STT at $0.01 a minute.
Written by Sume