Veo 3.1 on Luma with audio: 280 credits per second, double 140
Luma lists Veo 3.1 at 140 credits per second without audio and 280 with, at 720p and 1080p. Google lists dollars per second. Sume has a generate_audio field.

On Luma, turning audio on doubles the Veo 3.1 price: 140 credits per second without audio, 280 with, at both 720p and 1080p. For a 5-second clip that is 700 versus 1400 credits, or $2.10 versus $4.20 at the Plus rate of $0.003 per credit.
From Luma's pricing page, read 2026-10-01; the dollar figures are our arithmetic.
What does Luma list for Veo 3.1?
The Veo 3.1 rows cover text-to-video and image-to-video at 720p and 1080p only, each with an Audio column. Resolution does not change the price; the audio switch does.
| Resolution | Audio | Credits per sec | 5 sec clip on Plus |
|---|---|---|---|
| 720p | No | 140 | $2.10 |
| 720p | Yes | 280 | $4.20 |
| 1080p | No | 140 | $2.10 |
| 1080p | Yes | 280 | $4.20 |
What does Google list for Veo 3.1?
Google's Gemini API pricing page, read the same day, prices Veo 3.1 per second in USD and labels each line "video with audio price (default)": Standard $0.40 at 720p and 1080p, Fast $0.10 at 720p and $0.12 at 1080p, Lite $0.05 at 720p and $0.08 at 1080p. The saved text gives no without-audio line, so there is no Google audio-off figure to compare with Luma's 140.
How do I turn audio on or off on Sume?
The video submit body has a boolean generate_audio: "Whether to generate audio alongside the video. Defaults to the model's audio capability." Each model entry in GET /v1/videos/models reports whether it can generate audio, along with its supported resolutions and durations (see Video generation).
Sume reserves video jobs on submit at provider list price multiplied by 1.25, so the dollar amount on the job is the figure to budget with. Whether a given Veo-family model is in your catalog is something to read from that endpoint, not assume; Veo 3.1 model ids on Sume explains the naming.
What should I do to keep the audio cost down?
If you will replace the soundtrack anyway, render without audio and add sound afterwards. Where the model supports it, setting generate_audio to false is how Sume expresses that choice; the catalog tells you whether the field applies.
Sources
Related posts
More in Pricing
- MiniMax H3 license: written authorization above 20M USD revenue
MiniMax H3 license IV.1 requires separate prior written authorization above 20 million USD in yearly revenue. The text and Sume billing.
- OpenRouter image pricing: billable and unit vs Sume pricing
OpenRouter image pricing lines carry billable, unit and cost_usd. Sume's endpoints response shows lines with the same fields in its docs example.
- Pika parallel generations: 2 on Free, 8 on Creator; Sume by plan
Pika lists 2 parallel video generations on Free and Starter and 8 on Creator. Sume sets concurrency by plan and returns 429 queue_full when the queue is full.
- Replicate cost estimate before running vs Sume dry_run preview
Replicate lists cost estimates on each model's page. On Sume, send dry_run=true on a paid MCP tool to preview admission and cost without submitting the job.
Written by Sume