Veo 3.1: Specs, Pricing and Prompt Examples for Ads

Veo 3.1 is Google DeepMind's video model that generates picture and sound together. Here is what it means for your ads, what it costs and which prompts work. Status: September 2026; we update this page when Google ships a new version.

Short answer: Veo 3.1 by Google DeepMind creates 4, 6 or 8 second clips in 16:9 or 9:16, up to 4K, always with native audio. Through the Gemini API it costs from $0.05 per second (Lite, 720p) to $0.60 per second (Standard, 4K). It excels at realistic footage with matching sound, but Google only fully supports English, so for ads in other languages you usually add your own voice-over.

What is Veo 3.1?

Veo is Google DeepMind's family of video models. Veo 3.1 launched on October 15, 2025 in Google Flow, the Gemini app, the Gemini API and Vertex AI. Google highlighted richer audio, better prompt adherence and more realism than Veo 3. In January 2026 it gained native vertical video for reference-based clips plus upscaling to 1080p and 4K, and in spring the budget Lite variant followed. There are now three tiers:

• Veo 3.1 (called "Quality" in Flow): the highest quality.
• Veo 3.1 Fast: faster and much cheaper.
• Veo 3.1 Lite: the cheapest, up to 1080p and without reference images.

Note: in the Gemini app, Google is replacing Veo with Gemini Omni, a newer model for generating and editing video. In Flow, the Gemini API and Vertex AI you can still pick Veo 3.1. Google has not announced a Veo 4.

Veo 3.1 specs at a glance

Sources: the Veo documentation for the Gemini API (updated September 17, 2026) and the Vertex AI model page. Extension works with Veo 3.1 and Fast in the API, but only through Lite in Flow.

Strengths and weaknesses for ads

Where it shines:

• Sound is built in. You direct sound effects, ambience and dialogue from the prompt.
• Your product stays recognisable. With up to three reference images, Google says Veo preserves the look of a person or product.
• You control the final frame. Supply a first and last frame and the clip lands exactly on your packshot.
• True vertical. 9:16 is not a crop of widescreen.

Where it falls short:

• Short clips. Eight seconds per generation at most; extending only works in 720p.
• Two formats. Square or 4:5 feed versions need cropping.
• English first. Other languages "may work but results can vary", according to Google.
• People in the EU. In the EU, UK and Switzerland the Gemini API only allows generating adults.
• Top quality costs more. At 720p, Standard is four times the price per second of Fast.

Veo 3.1 pricing and access

Google Flow

Flow, Google's browser-based film studio, is available in the Netherlands and many other countries. Everyone gets 50 Flow credits a day, although without a subscription video generation may be unavailable at peak times. Google charges 10 credits for a Lite clip, 20 for Fast and 100 for Quality. More credits come with a Google AI plan: in the Netherlands, Plus costs €4.99 a month (200 credits), Pro €21.99 (1,000) and Ultra from €99.99 (10,000).

Gemini API and Vertex AI

Developers pay per second of video (pricing page):

An 8 second clip therefore costs $0.80 with Fast at 720p and $3.20 with Veo 3.1 at 1080p. There is no free tier, and you are not charged when a safety filter blocks a video. Download your output within two days, because Google deletes it after that.

Using Veo 3.1 in Genly

In Genly you can pick Veo 3.1 in the video generator and in GenFlow, the canvas where you build a video ad from connected blocks. You get a clip of about 8 seconds in 9:16 or 16:9. Veo 3.1 Fast costs 20 credits per clip, Veo 3.1 Quality 40 credits. In GenFlow you can connect a start image or up to three reference images, such as your product photo.

Because Veo officially only speaks English, this workflow suits non-English markets best: let Veo create the visuals and sound effects without spoken lines, then add a voice-over and captions in GenFlow. A last frame or a 4K option is not part of the Genly version; use Flow or the API for those. You don't need a Google subscription: you pay with the credits in your Genly plan.

Veo 3.1 prompt examples for ads

Google's prompting guide recommends a fixed structure: camera, subject, action, context, style and ambience. Put spoken lines in quotation marks, describe sounds separately ("SFX: ...") and direct multiple shots with timestamps such as [00:00-00:02]. The examples below use our demo brand Tropical Elixir, a bottled tropical juice.

1. Product shot with sound

Vertical 9:16, 8 seconds. Macro close-up, slow dolly-in: a chilled glass bottle of tropical juice stands on a wet slate surface while a hand pours the orange juice into a tall glass full of ice. Bright morning side light, condensation on the bottle, fresh premium commercial style. SFX: glass clinking, juice pouring, ice cracking. Ambient noise: soft birdsong through an open window. No dialogue, no music, clean frame without any on-screen text.

Without music or speech, you pick your own track and voice-over later.

2. Four shots in 8 seconds

Vertical 9:16, 8 seconds, handheld UGC look, natural daylight in a small kitchen.
[00:00-00:02] Close-up of a hand opening a nearly empty fridge. SFX: fridge hum.
[00:02-00:04] The hand grabs a bottle of tropical juice from the counter, seen from the side.
[00:04-00:06] Over-the-shoulder shot: juice pours into a glass with ice. SFX: pouring, ice.
[00:06-00:08] Static packshot of the bottle next to the full glass, empty space above it.
No dialogue, clean frame without any on-screen text.

The last shot leaves room for your headline.

3. Using your product photo as a reference

Keep the bottle exactly as in the product reference image: same shape, colours and label. Vertical 9:16, 8 seconds. Tracking shot at a beach bar at golden hour: a bartender, seen from behind, slides the bottle across the wooden bar towards the camera, where it stops in sharp focus. Warm teal and orange colour grade, shallow depth of field. SFX: bottle sliding on wood, distant waves, soft chatter. No dialogue.

Reference images only work with 8 second clips. In Genly, connect the product photo in GenFlow.

4. One spoken line

Horizontal 16:9, 8 seconds. Medium shot at eye level: a woman in her thirties, seen in profile, sits on a sunny balcony and takes a sip of tropical juice. She lowers the glass and says warmly: "Summer, in a bottle." Soft city ambience, a light breeze, warm natural light, realistic commercial look, no on-screen text.

For a campaign in another language, make the same shot without dialogue and add a voice-over.

Veo 3.1 vs the other video models in Genly

Sources: Gemini API, Kling 3.0 guide, BytePlus and OpenAI.

In short: choose Veo 3.1 for short, realistic shots with sound. For 30 seconds in a single generation, Seedance 2.5 is the better fit, while Kling 3.0 handles up to six shots in 15 seconds. Sora 2 is gone, but the structure in our Sora 2 prompt examples carries over to Veo. New to AI video? Start with how to make AI video ads.

Frequently asked questions