AI video cost per minute: Veo 3.1 vs the Sume catalog
One minute of generated footage costs $4.80 to $24 on Veo 3.1 at list prices and $0.75 to $15 on the Sume video catalog. Per-second rates, dated 2026-10-01.
One minute of finished AI footage costs between $4.80 (Veo 3.1 Lite, 1080p) and $24.00 (Veo 3.1 Standard, 1080p) at Google's list prices. On Sume the same 60 seconds costs $0.75 on Grok Imagine Video 1.5 up to $15.00 on Wan 3.0 at 1080p.
Cost of 60 generated seconds
Rates are per second of output, multiplied by 60. Sume rates are provider list times 1.25. Generated footage is rarely the final minute, because usable footage needs retakes; the keeper-rate post linked below covers that.
| Option | USD per second | USD per minute |
|---|---|---|
| Veo 3.1 Standard (list) | $0.40 | $24.00 |
| Veo 3.1 Fast (list) | $0.12 | $7.20 |
| Veo 3.1 Lite (list) | $0.08 | $4.80 |
| Sume Grok Imagine Video 1.5 (720p) | $0.0125 | $0.75 |
| Sume MiniMax H3 (768p) | $0.075 | $4.50 |
| Sume Wan 3.0 (1080p) | $0.25 | $15.00 |
| Sume Kling 3 audio off | $0.14 | $8.40 |
| Sume Kling 3 audio on | $0.21 | $12.60 |
| Sume Gemini Omni Flash 1.1 (1080p) | $0.1875 | $11.25 |
| Sume MiniMax H3 Max (1080p) | $0.20 | $12.00 |
Reading the table
The spread inside Sume is 20 times, from Grok at $0.0125 to Wan 1080p at $0.25 per second. The spread inside Veo is 5 times. Picking the row matters more than picking the vendor.
For a rough plan, multiply the per-minute figure by the number of takes you expect per kept shot.
- Cheapest minute: Grok Imagine Video 1.5, 4 to 15 seconds per request, 480p or 720p, and image-to-video only (it needs a first-frame image).
- Mid-range: Gemini Omni Flash 1.1 at 1080p and Kling 3 audio off.
- Highest in this table: Wan 3.0 at 1080p, which also allows 2 to 30 seconds per request.
Turning a minute into a shot list
A 60-second video made of twelve 5-second shots is twelve requests. Twelve Grok requests cost $0.75, twelve Kling 3 audio-on requests cost $12.60 and twelve Wan 1080p requests cost $15.00. Add $0.10 for a one-minute timeline render and $0.20 for captions.
If you need audio, the cheapest path is a row with native audio such as Kling 3 audio on, or a music generation at $0.125. Compare that with the price of a separate voice step before choosing.
- Count shots, not minutes, when you set a per-run cap.
- Price the finishing steps once per video, not once per shot.
Limits
Per-minute totals assume every second is billed once. Request windows are not the same across models, so a minute is several jobs. The tables show the model amount. A workspace Agent Fee line can be added on top, so treat the job's own usage cost as the truth and check the live catalog before you budget; limits and rows change.
Sources
Related posts
More in Comparisons
- AI/ML API vs Sume: OpenAI-style base URL or media jobs
AI/ML API gives an OpenAI-compatible base URL across text, image, video and music. Sume is media jobs, and its agent endpoint is async, not choices[].
- Alibaba Wan 3.0 on OpenRouter vs Sume: id, seed and aspect ratios
OpenRouter lists alibaba/wan-3.0 with seed and 2 to 30 seconds. Sume lists wan-3.0 with the same range, no seed, and its own billing. What to change.
- Amazon Ads Agent creative tools vs Sume for off-Amazon video
Amazon's unBoxed 2026 Ads Agent now makes TV-quality video from product pages. What it covers, and how Sume fills the TikTok, Meta and Shopify side.
- Asset Studio 1-Click A/B Testing vs a Sume bulk run of hook variants
Google's Asset Studio adds Gemini Omni video and 1-Click A/B Testing. If your Q4 test spans channels, queue the hook variants as a Sume bulk run instead.
Written by Sume