Luma Ray3.2 release date and specs: June 9, 2026, 20 s at 1080p
Luma released Ray3.2 on 2026-06-09: up to 20 s at 1080p, 16 keyframes, 8 tracked faces, HDR with EXR export, an API. What Sume offers instead.

Luma released Ray3.2 on 2026-06-09. Its announcement lists up to 20 seconds at 1080p, up to 16 keyframes per clip, tracking of 8 faces, native HDR with 16-bit EXR export, reframe, and an available API. Sume has no Ray3.2 id; if you need a long clip at 1080p on Sume, seedance-2.5 and wan-3.0 accept up to 30 seconds.
Ray3.2 facts are from Luma's announcement, read 2026-10-02. Sume's side is from the Video Router docs.
What does Luma state about Ray3.2?
The table restates only what Luma's page says, with the nearest Sume capability from Sume's docs beside it. A matching row means a similar number, not a similar result.
| Item | Luma Ray3.2 | On Sume |
|---|---|---|
| Release | 2026-06-09 | Not applicable |
| Clip length at 1080p | Up to 20 s | seedance-2.5 and wan-3.0: up to 30 s at 1080p |
| Keyframes | Up to 16 per clip | First and last frame only |
| Faces tracked | 8 | Not listed in the docs |
| HDR and EXR | Native HDR, 16-bit EXR export | MP4 results; no EXR documented |
| API | Available | Sume's own API, no Ray3.2 id |
Which Sume model is closest for a 20-second 1080p clip?
For one clip of 20 seconds, seedance-2.5 (4-30 s) and wan-3.0 (2-30 s) both fit at 1080p. Run the same prompt on both at 480p before paying for 1080p; see comparing models on one prompt.
These are different models, so the look will differ. Treat the pairing as a length match only.
What can I not get on Sume?
Multi-keyframe control in one clip, HDR-to-EXR delivery and Luma's face tracking are not documented on Sume. The first has a workaround in chaining first and last frames; the second does not.
If your pipeline needs EXR, generate in Luma. If you need an API-first catalog with Sume's job model, webhooks and Timeline, use the ids above.
Pick on the spec that matters to your job.
Sources
Related posts
More in Models
- Lyria 3.5 is single-turn and varies per call: keep the artifact
Google says Lyria generation is single-turn, not iteratively editable and varies between calls. Why to save every Sume Music artifact you like.
- Lyria 3.5 in another language: prompt in it, then check the song
Google says Lyria 3.5 makes music in other languages when you prompt in that language. On Sume, write the brief and lyrics in it, then check the result.
- MAI-Image-2.6 output cap is 2,359,296 pixels; Sume sizes differ
MAI-Image-2.6 sets a 2,359,296-pixel ceiling and a 768-pixel minimum edge. Sume sets size per model with tiers, ratios and, for GPT models, custom pixels.
- MAI-Image-2.6 edits take 5 references; Sume takes 10 or 16
MAI-Image-2.6 in Foundry accepts up to five JPEG or PNG reference images per edit. On Sume, input_references tops out at 10, or 16 on GPT Image 2.5.
Written by Sume