Which 8s 9:16 720p model fits a $1.00, $1.50 or $2.50 per-clip ceiling
Under a $1.00 per-clip ceiling, only minimax-h3 (768p), wan-3.0 and Omni Flash fit an 8-second 720p 9:16 clip on Sume. At $1.50 Kling joins; at $2.50 all do.

Under a $1.00 ceiling per clip, three models fit an 8-second 9:16 clip on Sume: minimax-h3 at 60 cents (768p, not 720p), wan-3.0 at 100 cents and gemini-omni-flash-1.1 at 100 cents. At $1.50 kling-3 with audio off joins at 112 cents. At $2.50 every model in the set fits, including seedance-2-fast at 242 cents.
OpenAI removed the Videos API on 2026-09-24 and named no replacement, and most teams that lost sora-2 have a per-clip price in mind already. Start from the ceiling and see which models are left.
The ceiling table
A model fits when its billable cents per clip are at or below the ceiling. The ceilings are 100, 150 and 250 cents.
| Model | Cents per clip | Fits $1.00 | Fits $1.50 | Fits $2.50 |
|---|---|---|---|---|
| minimax-h3 (768p) | 60 | Yes | Yes | Yes |
| wan-3.0 | 100 | Yes | Yes | Yes |
| gemini-omni-flash-1.1 | 100 | Yes | Yes | Yes |
| kling-3, audio off | 112 | No | Yes | Yes |
| seedance-2-mini | 152 | No | No | Yes |
| kling-3, audio on | 168 | No | No | Yes |
| seedance-2-fast | 242 | No | No | Yes |
Edge cases that change the answer
Seedance-2-mini at 152 cents misses the $1.50 ceiling by 2 cents. If the prompt can run at 6 seconds it costs 114 cents and fits. Kling with audio on is 168 cents, which is why a silent track matters in the table. Seedance-2-fast at 242 cents fits only at the top ceiling, and at 6 seconds (182 cents) it fits $2.00 but not $1.50.
MiniMax is the lowest row, but it counts only if 768p is acceptable. If a spec says 1280x720 exactly, treat it as not fitting at any ceiling, because the catalog says 768p is a native tier and not 720p.
What a ceiling means for a batch
A ceiling also caps the batch. At 100 cents per clip, 250 clips cost at most $250.00; at 150 cents, $375.00; at 250 cents, $625.00. Models under the ceiling cost less than that, so the real bill is lower, and the ceiling is the number to put in a budget approval.
Set the ceiling in your submit script too. Before it sends a job, the script looks up the model's cents for the requested length and resolution, and skips the row with a message if the price is above the ceiling. A typo in a duration then costs nothing.
Using the table well
A ceiling is a filter, not a recommendation. Among the models that fit, differences in motion, audio and how closely they follow a prompt matter more than a few cents. Run your five hardest prompts on each candidate, and take the cheapest one that you would approve.
The total for a library follows from the per-clip figure: 1,000 clips at wan-3.0 or Omni is $1,000.00, at Kling silent $1,120.00 and at seedance-2-fast $2,420.00. Pick the ceiling, then multiply.
- Ceilings are for 8 seconds; recompute for other lengths from the grid.
- Audio is a price choice only on Kling.
- The catalog lists limits per model, so check
supported_durationsbefore you assume a length.
Sources
Related posts
More in Pricing
- Which Sume submit errors cost money: 402, queue_full, 429, 503
None of them charges you. 402 fails before the reserve, queue_full releases the hold, and rate_limited and 503 mean the job was never accepted.
- A 4-second avatar clip on Sume: $0.736, $0.98 or $2.20 by tier
The shortest accepted Sume avatar clip is 4 seconds: 4 x $0.184 is $0.736, shown as $0.74; Plus is $0.98 and Max is $2.20. Sub-cent rates round up.
- An 80-cent cap per clip: longest Sume video each model and tier allows
With a hard $0.80 ceiling per clip, which Sume video models can run, and for how long? Wan 3.0 at 480p reaches 12 s; Seedance 2.5 at 1080p cannot start.
- 10 music tracks of 1 min: Sume $1.25 vs ElevenLabs $1.50
Sume Music charges a flat $0.125 per generation; ElevenLabs lists music at $0.15 per minute. 10 tracks of 1 min: $1.25 vs $1.50. Read 2026-10-09.
Written by Sume