Luma Ray3.2 makes 5 or 10 second clips; which Sume ids accept both
Luma's page lists 5 or 10 seconds native for text and image to video. On Sume, seedance-2.5, wan-3.0, kling-3, minimax-h3 and Omni all accept both 5 and 10.

Luma's Ray3.2 page lists text-to-video and image-to-video at 5 or 10 seconds native. Both durations are valid on every pinned Sume video id that produces text or image clips: seedance-2.5 (4-30 s), wan-3.0 (2-30 s), kling-3 (4-15 s), minimax-h3 and minimax-h3-max (5-15 s) and gemini-omni-flash-1.1 (3-10 s). Luma itself is not in the Sume catalog.
The Luma figures are from Luma's Ray page, read on 2026-10-03, and the Sume ranges from the Video Router docs and Video generation docs.
A fair comparison clip
If you are comparing a Luma clip with a Sume model, matching the duration first removes one variable. Use 5 seconds for a fast, cheap read on motion and 10 seconds to see whether the model holds identity and lighting over time.
Keep the other settings equal too: same prompt, same first frame, same aspect ratio. The side-by-side workflow is covered in same prompt, three models.
| Model | 5 s | 10 s | Range |
|---|---|---|---|
| Luma Ray3.2 (text, image) | yes | yes | 5 or 10 native |
| seedance-2.5 | yes | yes | 4-30 s |
| wan-3.0 | yes | yes | 2-30 s |
| kling-3 | yes | yes | 4-15 s |
| minimax-h3 / minimax-h3-max | yes | yes | 5-15 s |
| gemini-omni-flash-1.1 | yes | yes | 3-10 s |
Cost grows with seconds
Every Video Router model bills list times 1.25 and most bill per output second, so a 10 second clip costs about twice a 5 second one on the same model and resolution. Generate the 5 second version first; commit to 10 only when the shorter clip is right.
Seedance is billed on video tokens rather than seconds, so use the pricing block from the models list for it and compute a total from that.
Edge cases
Whole seconds only: the docs define duration as an integer and list supported_durations in whole seconds, so ask for 5 or 10, not 7.5.
If a model changes its range, supported_durations changes with it. Read the list when your service starts instead of copying numbers from a blog.
Sources
Related posts
More in Models
- Lyria 3.5 takes ten images; the Sume music router takes one
Google documents up to 10 images with text for Lyria 3.5. The Sume music router accepts a single image_url, so combine your mood board into one still first.
- MAI-Voice-2.1-Flash makes 45 s of audio: narrating a 5-minute script
Microsoft lists 45 seconds of audio per Flash generation. A 5-minute script needs about seven chunks. A Python splitter, and how Sume handles long text.
- MAI-Voice-2.1 vs Flash: Microsoft's best-for column, read closely
Microsoft lists MAI-Voice-2.1 for long-form narration and Flash for real-time agents, and notes the full model favors naturalness over latency.
- MAI-Voice-2.1: 14 serving regions, public preview, no SLA
MAI-Voice-2.1 is in public preview with no SLA and 14 serving regions. A pre-launch checklist, plus how an async TTS job changes the risk.
Written by Sume