What Veo 3.1 is
Veo is Google’s line of video models. Version 3.1 bets on three things: detail, realistic lighting and native sound. If Seedance is about dynamics and cinema, Veo is about photorealistic credibility of the frame and about video that comes with sound out of the box.
It works from text and from images. It can do transitions between two given frames (a start frame and an end frame), which is handy for precise control.
Feature one: 1080p only
This needs to be understood right away, because it breaks the usual logic of saving money.
Most models have a low resolution (480p) where generation is several times cheaper, and that is where drafts are run. The full Veo 3.1 has no low resolution at all. The model works only at 1080p.
Practical takeaway: you cannot make Veo cheaper with 480p drafts. If you need cheap tests, either take the lightest version (Veo 3.1 Lite, which also has 720p) or debug the prompt on another model that supports 480p (such as Seedance) and move the finished prompt to Veo.
Feature two: sound doubles the price
Veo can generate video together with sound, in sync. That is great for talking scenes and wherever in-frame sounds matter.
But you pay for it literally: turning sound on roughly doubles the price. A 6-second clip at 1080p costs about 132 tokens without sound and 264 with it.
And here is the honest question to ask yourself: do you need sound from the model at all? If you are making an ad or a reel and will lay music and voice-over on top in editing, you will throw Veo’s sound away after paying double for it.
Sound is justified when someone speaks in frame and lip sync matters, or when you need synced sound effects. In every other case switch it off.
Three versions in the line
| Version | Price for 6 s / 1080p / sound | Best for |
|---|---|---|
| Veo 3.1 Lite | 53 tokens (~84 rub.) | the cheapest, drafts, simple scenes |
| Veo 3.1 Fast | 80 tokens (~126 rub.) | balance of price and quality |
| Veo 3.1 (full) | 264 tokens (~417 rub.) | maximum detail and lighting, final takes |
This is not a public offer. KADR prices, October 2026, with sound.
A fivefold spread within one line. The logic is simple: find the idea on Lite, render the final shot on the full version.
What it costs without sound
Since sound doubles the price, here are the same versions with sound off (KADR data, 6 seconds, 1080p):
| Version | With sound | Without sound |
|---|---|---|
| Veo 3.1 Lite | 48 | ~29 |
| Veo 3.1 Fast | 72 | ~58 |
| Veo 3.1 (full) | 240 | 120 |
On the full version that is 132 tokens saved on every clip. At volume that is very noticeable money.
Veo versus Seedance: briefly
Since these are the two main video models, here is a quick orientation without tests, based on their nature.
Veo is strong in detail, realistic lighting and native sound, but 1080p only (for the full version) and short segments.
Seedance is strong in dynamics, holds the character and gives segments up to 15 seconds, but does not work with real faces.
Roughly: Veo for realistic talking scenes, Seedance for long dynamic clips and animation. The full per-minute cost for both is in the article on how much a minute of AI video costs.
Who Veo 3.1 is for
Good for: realistic shots where lighting matters, scenes with sound and speech, short atmospheric clips, precise transitions between two frames.
Keep in mind: no cheap 480p, so drafts go either on Lite or on another model. And watch the sound toggle, it quietly doubles the bill.
On the Google AI subscription and Flow from Russia: how to pay for Google Veo from Russia.
Where to try Veo without paying: Veo 3 for free.
In short
Veo 3.1 is Google’s model for detail, lighting and native sound. Three versions from 53 to 264 tokens per clip. Two things to keep in mind: there is no low resolution, so no cheap 480p drafts, and sound doubles the price, so switch it off when you are laying music on top anyway.
