Cheapest AI video API per second: eight listed prices, sorted
The lowest per-second prices on fal, Runway, Google, Replicate and Sume, read 2026-10-09, from $0.025 to $0.09, with which ones Sume actually lists.
The lowest listed price per second of AI video I found across these pages is $0.025, for fal's H3 Max Turbo, and the lowest on Sume is $0.0375, for Gemini Omni Flash 1.1 at 360p. Sume does not list fal's cheaper Turbo row, and I did not confirm a Sume row for Runway's Grok lite. 'Cheapest' also depends on resolution, audio and model, so use this as a shortlist, not a ranking of quality.
Eight rows, sorted by price
Each row is from the vendor's own pricing page or Sume's catalog, read 2026-10-09. Runway's prices use $0.01 per credit. Where a page names no resolution I say so.
| Model | Vendor | Resolution | $ per second | On Sume? |
|---|---|---|---|---|
| fal H3 Max Turbo | fal | not stated | $0.025 | no |
| grok_imagine_1_5_lite | Runway | 720p | $0.03 | not checked; see GET /v1/videos/models |
| gemini-omni-flash-1.1 | Sume | 360p | $0.0375 | yes |
| Veo 3.1 Lite | 720p | $0.05 | no | |
| H3 Max | fal | not stated | $0.05 | as minimax-h3-max, from $0.0625 at 480p |
| gen4_turbo | Runway | not stated | $0.05 | no |
| wan-3.0 and minimax-h3 | Sume | 480p | $0.0625 | yes |
| Wan 2.1 | Replicate | 480p | $0.09 | no |
What changes at higher quality
The ordering changes as resolution rises. At 720p the cheap Sume rows are wan-3.0 and gemini-omni-flash-1.1 at $0.125 per second, against Veo 3.1 Lite at $0.05 and Runway's wan3 at $0.10. At 1080p, Sume's wan-3.0 is $0.25 and Omni Flash is $0.1875, against Veo 3.1 Lite at $0.08.
Sume's prices are provider list times 1.25, so a cheaper route elsewhere stays cheaper unless the extras matter to you.
- Per-second pricing is linear in duration inside each model's valid range.
- Duration limits: Omni Flash 3 to 10 s, minimax-h3 5 to 15 s, wan-3.0 2 to 30 s.
When to pay Sume's multiple
Pick Sume when you want one job format, one USD balance and one set of docs across seedance, wan, minimax and Omni models, or when you need the OpenRouter-compatible /v1/videos route. Pick the vendor directly when one model at scale is your workload. Related: Veo 3.1 Lite at 5 cents vs the cheapest Sume draft.
How to keep this current
Prices move weekly. The reliable method is to pull Sume's GET /v1/videos/models and the vendors' pricing pages into one spreadsheet, convert everything to dollars per output second at one resolution, and sort. Record the read date next to each row, as this post does, because a price without a date is a rumor.
Also record duration limits, audio behavior and whether the price is for a standard or fast tier.
What I could not confirm
Higgsfield's pricing page returned no plan data when fetched, so it is left out, and Luma's quickstart lists no prices, so Luma is absent too. Kling's newest tier has no public API price on the page checked. All vendor figures were read on 2026-10-09 and are listed in the sources; Sume figures come from its public catalog and docs the same day.
Sources
Related posts
More in Comparisons
- Cheapest possible clip per Sume video model: the minimum-length floor
The cheapest legal clip on Sume ranges from $0.11 (Omni, 3 s at 360p) to $1.07 (Seedance 2.5, 4 s at 480p). Per-second rates hide the minimum-duration floor.
- Codex custom Responses provider: can it point at Sume?
Codex CLI 0.162.0 adds custom Responses-compatible provider config. Sume's Agent Completions is not a model endpoint; use Sume's hosted MCP from Codex instead.
- Context stitching and streaming TTS: Eleven v4 vs Sume async jobs
Eleven v4 lists context stitching and bidirectional streaming. Sume TTS is async and non-streaming, so keep a section in one job and join sections.
- Deepgram Flux Multilingual $0.0078/min: 7,500 minutes vs Sume STT $75
Deepgram lists Flux Multilingual streaming at $0.0078 a minute. For 7,500 minutes that is $58.50, against $75.00 on Sume STT. The full Deepgram ladder is below.
Written by Sume