Veo 3.1 is Google's October 2025 video model. It still makes short clips, but native audio, image-to-video, and reference-image control are the real upgrades.
What Veo 3.1 actually changed
Google released Veo 3.1 and Veo 3.1 Fast on October 15, 2025, in paid preview on the Gemini API, Google AI Studio, and Vertex AI. The same model is in the Gemini app and in Flow. At launch, Google said it cost the same as Veo 3.
The clip length did not jump. A single take is still a few seconds. What changed is sound and control.
Richer native audio. Dialogue, room tone, and synced effects come with the file. You write the spoken line into the prompt. You do not start from a silent MP4.
Better image-to-video. Google says prompt adherence, picture quality, and character consistency improved when you start from a still. Better, not perfect. Faces still drift. Small logos still smear.
Ingredients to video. Up to three reference images of a character, object, or scene. That is the practical way to keep a person or a product recognizable across shots.
First and last frame. Supply a start still and an end still. Veo 3.1 generates the transition, with audio.
Scene extension. Each new clip is generated from the final second of the previous one. Google's Veo 3.1 announcement says you can go a minute or more this way. That is how you get past an 8-second take. Lite cannot extend.
A January 2026 update added native 9:16 for ingredients-to-video, so vertical social is not a crop of a 16:9 render. The same update improved 1080p and 4K upscaling on Flow, the Gemini API, and Vertex AI. Google's current Veo docs list 720p, 1080p, and 4K on Veo 3.1; Lite still has no 4K.
Length, resolution, and audio (the constraints)
These numbers decide whether the model fits the job.
| Lite | Fast | Standard | |
|---|---|---|---|
| Clip length | 4 / 6 / 8 s | 4 / 6 / 8 s | 4 / 6 / 8 s |
| Aspect ratio | 16:9, 9:16 | 16:9, 9:16 | 16:9, 9:16 |
| Resolution | 720p, 1080p | 720p, 1080p, 4K | 720p, 1080p, 4K |
| Native audio | Yes | Yes | Yes |
| Scene extension | No | Yes | Yes |
| 4K | No | Yes | Yes |
A single generation is 4, 6, or 8 seconds. Eight seconds is the ceiling for one take. Ingredients (up to three refs) and first/last-frame control typically need that 8-second setting. On the API, video extension is also limited to 720p.
Audio is native. Put the spoken line and the sound in the prompt: footsteps, traffic, a kettle, a short piece of dialogue. Tiny on-screen type is still a weak spot. Do not ask Veo 3.1 to render a readable price, URL, or logo lockup.
Retries cost money. Google only bills a clip that actually generates. A clip that generates and is unusable still bills. If you need three takes to get one you can publish, you pay for three.
Lite is the cheap iteration tier: 720p or 1080p, no 4K, no extension. Fast and Standard reach 4K. Standard is the quality tier, and the one Google demonstrates for ingredients-to-video.
How to get access
Official Google paths first.
Gemini app. Consumer surface. Prompt, generate, download. Fine for a one-off. You are inside Google's consumer product, not an API.
Flow. Google's filmmaking tool. Ingredients, frames-to-video, and extend live here, with audio on those features. Use this if you are already cutting inside Google's editor.
Gemini API / Google AI Studio. Paid preview. Model IDs include veo-3.1-generate-preview, veo-3.1-fast-generate-preview, and veo-3.1-lite-generate-preview. You need a paid Gemini API key. Veo 3.1 is not on the API free tier.
Vertex AI. Same family, enterprise path. Same model, different billing account.
Those four are Google. Movby.ai is not. It is an independent aggregator. It does not own Veo.
If you just want to run a prompt in a browser without an API key, Movby.ai's Veo page runs Veo 3.1 Lite, Veo 3.1, and Veo 3.1 Quality. Modes are text-to-video, image-to-video, and ingredients-to-video on standard Veo 3.1 (up to three refs). New users get 30 free credits, no card. Typical wait on that page is about two minutes. Site-wide, videos often take 1–5 minutes.
What it costs
On Google's API, you pay per second of video with audio. Figures below are from the live Gemini API Veo 3.1 pricing table. If that page has moved, use the page, not this article.
| Per second, video with audio | 720p | 1080p | 4K |
|---|---|---|---|
| Standard | $0.40 | $0.40 | $0.60 |
| Fast | $0.10 | $0.12 | $0.30 |
| Lite | $0.05 | $0.08 | — (not supported) |
Worked example: one 8-second Standard clip at 1080p is 8 × $0.40 = $3.20. The same length at Fast 1080p is $0.96. Lite 720p is $0.40. Standard 4K is $4.80. You are only charged if generation succeeds. Three successful retries of a Standard 1080p shot are $9.60 before you have a take you keep.
The Gemini API free tier does not include Veo 3.1.
Movby bills in credits, not dollars per second. Do not convert that to a per-clip USD unless you are looking at the current credit cost on the generator. Live plan prices, watermarks, and commercial terms are on Movby.ai pricing. Free-plan outputs are personal and non-commercial, and they may have watermarks. Paid plans are watermark-free and allow commercial use.
What it is good at, and what still breaks
Good. Dialogue-led shorts: a person says one or two lines, the room sounds like a room. Product-in-scene, if you feed a still of the product. Vertical social, now that 9:16 is native. Character lock across a couple of shots, if you use up to three reference images on standard Veo 3.1. First-to-last-frame for a controlled camera move.
Weak. Long-form in one shot. You will not get a 30-second hero film from a single generate. Scene extension is how you go longer, and Lite cannot do it. Tiny on-screen text. Heavy iteration: every miss that still renders is a paid miss. Prompt adherence is better than Veo 3. It is not a storyboard artist. If the prompt is vague, the model will invent blocking you did not ask for.
Treat it as an 8-second unit. Cut, extend, or regenerate. Do not plan a one-take scene that needs 20 seconds of continuous action unless you are ready to extend and accept the join.
A 4-step try
Keep this boring.
- Create a Movby.ai account. New users get 30 free credits. No card.
- Open the Veo 3.1 generator. Pick Veo 3.1 Lite, Veo 3.1, or Veo 3.1 Quality. Use standard Veo 3.1 if you need ingredients-to-video.
- Paste a prompt, or upload a still, or (on standard Veo 3.1) up to three reference images. Set 16:9 or 9:16. Duration on this page is 8 seconds.
- Generate. Wait about two minutes. Preview. Download the take you can use, or change one variable and run it again.
A prompt that actually specifies camera and sound:
Handheld medium close-up, 35mm, overcast. A woman in a navy wool coat walks toward camera on a wet Tokyo side street at dusk. Sodium street lights in the background. She glances left, then says: "The train is in four minutes." Footsteps on wet pavement, a distant crossing signal, light rain.
One subject. One camera. One spoken line. If that miss, change the line or the still. Do not rewrite it into a trailer.
FAQ
Is Veo 3.1 free?
Not on the Gemini API. Google's free tier does not include Veo 3.1; you pay per second of successful video. On Movby.ai, new users get 30 free credits with no card. Free-plan clips are personal and non-commercial, and they may have watermarks. Paid plans drop the watermark and allow commercial use.
How long are Veo 3.1 videos?
A single take is 4, 6, or 8 seconds. Eight seconds is the max for one generation. Scene extension continues from the last second of the previous clip, so you can string takes into a minute or more. Lite cannot extend. Reference images and first/last-frame shots usually need the 8-second setting.
Veo 3 vs Veo 3.1 — should I care?
Yes if you need native audio with the picture, stronger image-to-video, up to three reference images, first-and-last-frame transitions, or scene extension. Length did not increase. At launch Google priced 3.1 the same as Veo 3. Prompt adherence is better. You still cut 8-second pieces.
Can I use the clips commercially?
On Movby.ai, paid plans include commercial use; free-plan outputs are personal and non-commercial. Watermarks come off on paid plans. If you generate in the Gemini app, Flow, or the Gemini API, those products have their own terms. Read them before you run an ad. Movby does not replace Google's license.
Run an 8-second test with audio on the Movby.ai Veo 3.1.
Top comments (0)