Video API price per second: Veo, Runway, Luma, MiniMax, Seedance
A vendor-authored Higgsfield table lists Veo 3.1 at $0.40 per second down to MiniMax H3 at $0.08. What 10 seconds costs and why to verify each rate.

A vendor-authored comparison read on 2026-10-03 puts Veo 3.1 Standard at $0.40 per second, Runway Gen-4.5 at $0.12, Luma Ray 3.2 at $0.09, MiniMax H3 at $0.08 and Seedance 2.5 at $0.231 at 720p. A 10 second clip is $4.00, $1.20, $0.90, $0.80 and $2.31 respectively. Treat every rival price as unverified: the table was written by Higgsfield, a competitor of each vendor listed.
The table and the 10-second totals
Totals are the listed per-second rate times 10, rounded to the cent.
| Model | Per second | 10 seconds |
|---|---|---|
| Veo 3.1 Standard | $0.40 | $4.00 |
| Runway Gen-4.5 | $0.12 | $1.20 |
| Luma Ray 3.2 | $0.09 | $0.90 |
| MiniMax H3 | $0.08 | $0.80 |
| Seedance 2.5 at 720p | $0.231 | $2.31 |
Why not to copy the numbers into a budget
Two reasons. The table is vendor-authored, so each row can lag the owner's own page. Rates vary by resolution and audio, and the table gives one number per model.
The safe use is to find each model's own pricing page, read it the same day, and treat this table as the list of models worth checking.
The same arithmetic on Sume
Sume bills every video model at provider list times 1.25, reserved on the workspace balance at submit. If a provider list were $0.08 per second, a 10 second job would reserve $1.00. That is an illustration of the multiplier, not a quote for any model, because the live rate per model is in pricing_skus from GET /v1/videos/models.
Because the rate depends on model, resolution and duration, ask the catalog for the exact combination you plan to run. The poll response usage.cost then tells you what a finished job actually billed.
A budgeting method
Pick the resolution you will actually ship, since most ads end up at 720p or 1080p. Pick the clip length you will actually use. Multiply, add a retry allowance of your own, and add the cost of audio if the model does not include it. Do this for two or three candidates and compare finished-clip cost, not list rate.
Then run a small paid pilot on each. Real invoices beat any comparison table, including this one.
A short rule for comparing
Compare at the same duration, resolution and audio setting, and note which models include native audio. A cheaper per-second rate without sound can cost more once you add a separate speech or music job.
Sources
Related posts
More in Comparisons
- Vidnoz 450 credits: how many avatar seconds is that vs Sume?
Vidnoz bills 0.5 credits a second, or 2 for photo avatars. See what 450 credits buys, and what the same seconds cost on Sume Avatar 1.0 per tier.
- Vidyard 3 custom AI avatars and 15 AI videos vs Sume handles
Vidyard lists 3 custom AI avatars on paid plans and 15 AI videos on Free. Price the same on Sume: three avatars at $0.95 and 15 clips by tier.
- Vmake AI Video Translator vs a Sume subtitle pipeline
Vmake's June 2026 video translator lists 14 languages, subtitle replacement, voice matching and lip sync. Which of those steps map to Sume endpoints today.
- Voice from a short clip: Voxtral 5-25 s, MAI, Sume avatar voice
How much reference audio Mistral and Microsoft ask for to match a voice, and how Sume selects a voice instead: avatar_id, avatar_handle or a voice id.
Written by Sume