What Is Google Veo 3.1 Lite? Price, Limits and Model IDs

Forty Cents for an Eight-Second Veo Draft

A skincare brand has one serum, twenty hook ideas and a video budget that covers maybe five polished renders. Someone has to decide which fifteen ideas never get made. That call is usually a guess.

Google's cheapest Veo tier changes the price of that guess. On the Gemini API's pricing table, Veo 3.1 Lite costs $0.05 a second at 720p, as of September 21, 2026. Multiply it out and an eight-second 720p draft is $0.40, so twenty hooks as eight-second drafts come to $8. The same twenty clips on Veo 3.1 Standard, listed at $0.40 a second, would come to $64 by the same arithmetic.

Veo 3.1 Lite (often misspelled "Veo 3.1 Light") is the budget tier of Google's Veo 3.1 video model family. Google launched it in preview on the Gemini API on March 31, 2026, under the model ID veo-3.1-lite-generate-preview, and the release note describes it as "our most cost-efficient video generation model, designed for rapid iteration and building high-volume applications." Six months on, it is still a preview model on both of the Google developer surfaces that list it.

This explainer covers the two model IDs, the per-second price next to Veo 3.1 Fast and Standard, what Lite keeps and what it gives up on each surface, what the arena leaderboards say about the quality gap, and where a Lite clip is good enough for ad testing.

What Veo 3.1 Lite Is, and Where Google Sells It

Lite is not a new model family. It is a third price point inside Veo 3.1, next to Standard and Fast, and the useful way to understand it is through Google's own pages, read carefully and one surface at a time.

Google's lowest-priced Veo 3.1 tier

Google's model page for Lite says it "leverages the state-of-the-art Veo 3.1 model" and pitches it squarely at volume: "Veo 3.1 Lite offers the best prices for developers and enterprises building high-volume video applications and workflows at scale." Those are Google's words about its own product. They are a positioning statement, not a benchmark.

The spec sheet is shorter than the pitch. On the Gemini API, Lite takes a text prompt or an image as input and returns video with audio, one clip per request. Read plainly, that makes it the tier for drafts, prototypes and batch jobs, the work where the per-second price matters more than the last increment of fidelity.

Two surfaces, two model IDs, two dates

Google sells Lite to developers in two places, and the details differ enough that you should keep them apart. Its consumer studio, Google Flow, also lists a Veo 3.1 Lite option in its credit table, as of September 21, 2026.

Surface	Model ID	Stage	Date
Gemini API and AI Studio	veo-3.1-lite-generate-preview	Preview	March 31, 2026
Google Cloud Agent Platform	veo-3.1-lite-generate-001	Preview	April 2, 2026

The Gemini API date comes from Google's release notes, and Google's launch post, dated the same day, says you access the model "via paid tier on the Gemini API and Google AI Studio." The Google Cloud date comes from the Veo 3.1 page on Gemini Enterprise Agent Platform, which lists veo-3.1-lite-generate-001 with a launch stage of Preview and a release date of April 2, 2026.

Two days apart, two IDs, two products. A spec you read on one surface's page does not automatically hold on the other, as the extension split further down shows. The Google Cloud page adds two operational details, which apply to Lite and to the other two tiers alike:

• Region: a single one, us-central1.
• Pay-as-you-go: marked Not supported.

Preview everywhere, with no GA release yet

The release stage is where the two surfaces line up, and not in Lite's favor:

• Gemini API: all three Veo 3.1 tiers carry preview IDs, veo-3.1-generate-preview, veo-3.1-fast-generate-preview and veo-3.1-lite-generate-preview, and the Veo guide's feature table marks Lite as preview.
• Google Cloud: Veo 3.1 and Veo 3.1 Fast are listed as GA, both with a release date of November 17, 2025. Lite is listed as Preview.

Put the two pages together and Lite is the one Veo 3.1 tier with no GA release on either surface. For planning, that means the ID, the price and the limits can all still change (the OpenAI video API shutdown is the far end of that risk), which is why every number in this article carries the date it was read.

Google Cloud's page does answer the commercial question for its own surface. Directly above the Lite listing, it says that for this generative AI preview offering, "customers may elect to use it for production or commercial purposes, or disclose generated output to third-parties," subject to the agreement you access Google Cloud under. That notice covers Google Cloud. The Gemini API's terms are a separate document, not covered here.

What Veo 3.1 Lite Costs Next to Fast and Standard

Google prices Veo by the second of generated video, with audio included, on the Gemini API's paid tier. The table lists no free-tier price for any of the three tiers. The pricing page was last updated on September 16, 2026, and these are the figures it showed when we read it on September 21.

The per-second price ladder

Veo 3.1 tier	720p	1080p	4K
Standard	$0.40	$0.40	$0.60
Fast	$0.10	$0.12	$0.30
Lite	$0.05	$0.08	Not supported

All prices are per second, in USD, on the Gemini API paid tier. Two things stand out. Standard charges the same at 720p and 1080p, so on Standard there is no discount for rendering small. On Lite the resolution choice matters most: by our arithmetic, dropping from 1080p to 720p cuts Lite's per-second price by more than a third, where on Fast the same drop saves about a sixth.

What a clip costs, by arithmetic

Google's table is per second, but ads are planned per clip. Multiplying out gives the figures below. They are our calculations from the per-second prices above, not numbers Google publishes.

• Lite, 4 seconds at 720p: $0.20
• Lite, 8 seconds at 720p: $0.40
• Lite, 8 seconds at 1080p: $0.64
• Fast, 8 seconds at 720p: $0.80
• Fast, 8 seconds at 1080p: $0.96
• Standard, 8 seconds at 720p or 1080p: $3.20

By the same arithmetic, Lite is one-eighth of Standard at 720p and one-fifth of it at 1080p. One more line on the pricing page matters when you run drafts in bulk: Google says "you will only be charged if your video is successfully generated." In a batch of two hundred drafts, the ones that fail are not billed.

How "less than half of Fast" reads today

Google's launch post for Lite, dated March 31, 2026, positioned it at "less than 50% of the cost of Veo 3.1 Fast, but with the same speed." The same post announced that Google would be reducing the pricing for Veo 3.1 Fast on April 7.

Against today's table, the comparison holds only at the edge. At 720p, Lite is exactly half of Fast, $0.05 against $0.10. At 1080p it is two-thirds, $0.08 against $0.12. If your drafts need 1080p, the gap between Lite and Fast is narrower than the launch framing suggests, and it is worth pricing both before you commit a pipeline to Lite. For how Veo's per-second price sits against another engine, see our Kling vs Veo cost breakdown, and for Veo bought through a second vendor, our Runway vs Veo comparison.

What Lite Keeps From the Veo 3.1 Family

Cheaper tiers usually strip things out. Lite strips out less than its price suggests, and most of what it keeps is what a short social ad needs.

The specs that matter for ads

Here is what the Gemini API documentation and Google's launch post list for Lite:

• Native audio, always on. The Veo guide's feature table marks audio as always on for Lite, as for Veo 3.1 and Fast. Sound is part of every draft, not an upgrade; for a full song under the cut, Google offers Lyria 3.5 on the same Gemini API.
• Vertical and horizontal. Google's launch post lists landscape 16:9 and portrait 9:16. For vertical social placements, 9:16 is the spec that decides whether a model is usable at all.
• Four, six or eight seconds. The launch post says you can set the duration at 4, 6 or 8 seconds, "with cost adjusting accordingly."
• Text or image in. Lite accepts a text prompt or an image to animate, which covers the product-photo-to-motion route that a lot of ad work starts from.
• First and last frame. The Gemini API parameter table accepts a lastFrame image on Lite, so you can pin both ends of a shot and let the model generate the move between them.
• 24 fps and one clip per request, the same as the other two tiers.

The operational details that apply to every tier

A few limits in Google's Veo guide apply across the family and are easy to miss until they bite:

• Latency. The guide lists a minimum of 11 seconds and a maximum of 6 minutes during peak hours. Plan batch jobs around the slow end.
• Retention. Generated videos are stored on the server for 2 days, then removed. Download what you want to keep.
• Watermarking. Videos created by Veo are watermarked with SynthID, Google's tool for identifying AI-generated content.
• Language. The guide says English is fully supported and other languages "have not been evaluated, so they may work but results can vary." If your ad has dialogue in Spanish, Portuguese or German, test a few clips before you batch a hundred.

What Lite Gives Up, Surface by Surface

On the Gemini API, moving an existing Veo 3.1 call to Lite starts as a model-string change, from veo-3.1-generate-preview to veo-3.1-lite-generate-preview. What changes underneath is which features the model accepts, and on that question Google's two surfaces do not tell the same story.

No 4K

This is the one limit every Google page agrees on. The Gemini API pricing table marks Lite "(4k output not supported)," the Lite model page says the preview "does not support 4K outputs," and the Veo guide notes that 4K is "not available for Veo 3.1 Lite." Google Cloud's Lite listing gives output resolutions of 720p and 1080p.

For vertical social placements that is rarely the binding constraint. For a hero spot on a large screen, it might be. Note too that on the Gemini API, 1080p renders are 8 seconds only, on Lite as on Standard and Fast, so a four- or six-second Lite clip is a 720p clip.

Extension: the two surfaces disagree

Extension means continuing a Veo-generated clip past its original length. It is where the documentation splits:

• Gemini API: not supported. The Lite model page says the preview does not support Extension, and the Veo guide says the feature "is available for Veo 3.1 & Veo 3.1 Fast models only, not Veo 3.1 Lite."
• Google Cloud: marked Supported. The Veo 3.1 Lite listing on Gemini Enterprise Agent Platform shows "Extend videos" as Supported.

Each page speaks for its own surface, and nothing we read reconciles them. If a longer continuous shot matters to your ad, test extension on the exact surface and model ID you will ship on before you build around it. On the Gemini API, extension is a clear reason to stay on Fast or Standard.

Reference images and video input

Reference images let you pass Veo 3.1 up to three images of a person, a character or a product, which Google's Veo guide says is to "preserve the subject's appearance in the output video." Google's documentation marks them unavailable for Lite on both surfaces: Google Cloud's Lite listing shows "Use reference images" as Not supported, and the Gemini API parameter table lists the referenceImages field as n/a for Lite.

Lite gives up video input too. The Veo guide lists text-to-video, image-to-video and video-to-video for Veo 3.1 and Fast, but only text-to-video and image-to-video for Lite. For brand work that depends on the same bottle looking identical across a sequence of clips, the missing reference images are the most expensive thing Lite leaves out. You can still anchor a single clip on a product still, but you cannot hand the model a set of references and ask it to hold them.

Does the Quality Gap Show in the Arena?

Price and feature lists are Google's to publish. Quality is not, so the useful outside check is Artificial Analysis, which runs video arena leaderboards and publishes an Elo score with a 95% confidence interval for each model. We read both of its with-audio boards on September 21, 2026. The scores move as votes come in, so treat them as a snapshot.

Text-to-video: three tiers inside one error bar

On the text-to-video leaderboard (with audio), Veo 3.1 scores 1088 (±7), Veo 3.1 Fast scores 1085 (±6) and Veo 3.1 Lite scores 1083 (±7).

Our read of those numbers: the intervals overlap almost entirely. On this board, the arena does not separate the three tiers, even though Standard costs eight times as much as Lite per second at 720p.

Image-to-video: Lite edges Fast

On the image-to-video leaderboard (with audio), Veo 3.1 scores 1082 (±7), Veo 3.1 Lite scores 1069 (±8) and Veo 3.1 Fast scores 1065 (±7).

Our read: Lite sits four points above Fast, which is well inside the noise, and 13 points below Standard. Image-to-video is the mode closest to how product ads usually start, with a product photo as the first frame, so this is the board we would weight for ad work. On it, the step from Lite to Fast buys nothing visible, and the step to Standard buys a small edge.

What an arena score cannot tell you

An arena Elo is built from viewer preferences on the arena's own prompts. It is not a conversion rate, and it does not measure the things a product ad lives or dies on:

• whether the label on the bottle survives the camera move,
• whether a hand holds the product naturally,
• whether the first second stops the scroll.

So the scores support a narrower claim than "Lite is as good." They say that, on average and on general prompts, Lite is not visibly worse than Fast. The only board that settles an ad question is your own ad account, and the way to get there cheaply is the subject of the next section. If you want a Google model outside the Veo line, Gemini Omni Flash is a separate video model with its own trade-offs, covered in our Gemini Omni Flash explainer.

When a Lite Clip Is Good Enough for Ad Testing

Google's own framing for Lite is "rapid iteration," and that maps onto the stage of ad work where most ideas die: screening. The job of a draft is to kill bad ideas cheaply. It is not to run.

Use Lite when

• You are screening hooks. The opening seconds carry the test, so a four-second 720p draft, $0.20 by the per-second arithmetic, is enough to judge an opening. Our guide to ad hook variations covers how to structure those tests.
• You are choosing between angles or scenes before you commit to a final render or a shoot.
• You are animating a product still into a short vertical clip, the image-to-video job where Lite scored level with Fast.
• You need volume inside a fixed budget, and hearing the sound in the draft helps you judge it.

Step up to Fast or Standard when

• You need 4K, or extension on the Gemini API.
• You need reference images to hold a product or a character steady across shots, or you need video input.
• The clip is the final, the version that will carry spend. Standard scored 13 points above Lite on the image-to-video board on September 21, and a final is where a small edge is worth paying for.
• Your script is not in English and you have not yet tested how Veo handles it.

A worked example: twenty hooks for one serum

Say the skincare brand from the top of this article wants to draft all twenty opening ideas and put three finished variants into a real test. The costs below are our arithmetic from the Gemini API per-second prices, not figures Google publishes.

Plan	Clips	Spec	Cost
Lite drafts	20	4 s, 720p	$4.00
Standard finals	3	8 s, 1080p	$9.60
Draft, then finish	23	Mixed	$13.60
All Standard	20	8 s, 1080p	$64.00
All Fast	20	8 s, 720p	$16.00

The draft-then-finish plan screens every idea and still renders three finals on the most expensive tier, for about a fifth of the cost of making all twenty on Standard. The all-Fast row is the honest alternative: slightly more money, one tier throughout, no separate finishing step, and 720p finals. Either way, the split is the one our ad creative testing guide argues for: spend little to find the idea, spend more on the version that runs.

A clip is still not an ad. Twenty drafts give you twenty moving openings. They do not give you a script, a person on camera, captions, or enough finished variants to test against each other, which we work through in how many ad creatives you need.

How Novoads Solves the Draft-Then-Finish Problem

Novoads offers Google Veo 3.1 as one of its video models. Veo 3.1 Lite is not a separate model you can pick in Novoads, and the difference matters for how you plan the workflow above.

Veo 3.1 in Novoads is priced per clip

Google bills Veo by the second and by tier. In Novoads, Google Veo 3.1 runs at 4, 6 or 8 seconds, in 720p or 1080p, in 9:16 or 16:9, at one flat credit cost per clip whichever of those settings you choose. That makes the full model easy to budget for finals.

It also means Novoads is not the cheaper place to render twenty raw Veo drafts. For bare Veo clips at volume, Lite on the Gemini API costs less per clip, and we would rather say so than pretend otherwise.

The low-cost drafting engine in Novoads

The drafting role that Lite plays in Google's lineup falls to a different engine in Novoads. At the Plus-plan credit rate, a clip runs from about 25 cents for a five-second Seedance 2.0 Mini clip to about $4 for an eight-second Google Veo 3.1 clip, and about $10 for a full 30-second Seedance 2.5 take. Every plan is on the pricing page. The pattern is the same as in the worked example: draft cheap on Seedance 2.0 Mini, then finish the winners on Google Veo 3.1 or Seedance 2.5.

What sits around the clip

The bigger difference is everything that is not the clip. In Novoads you start from a script or an uploaded product image, pick an AI actor to deliver the lines, add a voice and captions, and download a vertical, square or horizontal file for whichever ad platform you run on, with Google Veo 3.1 among the video models you can use along the way. Those are the parts a raw Veo draft leaves open. For engines outside the Veo family that ad makers test against it, see our roundup of Veo alternatives, and for a spec-by-spec comparison, Seedance 2.0 vs Veo 3.1.

The Cheapest Usable Tier Sets How Many Ideas You Test

Veo 3.1 Lite is not the model you finish on. It is the model that decides how many ideas you can afford to look at before you finish anything. At $0.05 a second at 720p on the Gemini API (as of September 21, 2026), the question stops being which three hooks to make and becomes which three of twenty deserve the expensive render.

A better final raises the ceiling of one ad. A cheaper draft raises the number of ideas that get a shot at that ceiling. When the winners are picked, the full Veo 3.1 model is in Novoads, next to a cheaper engine for the drafts.