Most guides to the Google Veo AI video generator still tell you to open the Gemini app and type a prompt. That advice is out of date. In October 2026, the Gemini app makes videos with a newer model called Gemini Omni, and Veo 3.1 lives somewhere else.
Google Veo at a Glance: What Veo 3.1 Can Do in 2026
Veo 3.1 is Google’s named video model for short cinematic clips with sound, and Google says its Flow app alone had generated more than 275 million videos five months after launch (Google Blog, Introducing Veo 3.1). It makes clips of 4, 6 or 8 seconds from text, images or start and end frames.
So where do you actually get it? That is the part most pages get wrong. Google moved Veo around in 2026: the Gemini app switched to Gemini Omni, and the Veo 3.1 Preview models in the Gemini API shut down on October 22, 2026. Today you reach Veo 3.1 through Google Flow, Google Cloud’s enterprise platform, and a few Google apps that build on it.
This guide is built from Google’s own help centers, model pages and pricing pages, read on October 24, 2026. We have no video output of our own to show you, so nothing here scores visual quality. What you get instead is the map: which app runs which Veo model, what each plan’s credits buy, what the API costs, and where the limits sit.
Key Takeaways
Veo 3.1 now runs mainly in Google Flow and on Google Cloud, not in the Gemini app. In Flow, one Veo 3.1 clip costs 10 to 100 credits on non-Ultra plans, and Google AI Pro adds 1,000 credits a month (Google Flow Help). Clips run 4 to 8 seconds, with native audio, in 16:9 or 9:16.
Where You Can Use Veo Now: Flow, Google Cloud, Vids and YouTube Shorts
The main home of the Google Veo AI video generator is Google Flow, whose help page says it offers “a range of features powered by various Veo models, in addition to Gemini Omni” (Google Flow Help, models). Developers and companies reach Veo 3.1 on Google Cloud. Here is the full picture as of October 24, 2026.
| Where | Which video model | Who it’s for |
|---|---|---|
| Google Flow (web and app) | Veo 3.1 Lite, Fast and Quality, plus Gemini Omni Flash | Creators who want to pick the model and build scenes |
| Gemini app | Gemini Omni | Anyone with a Google AI plan who wants quick clips in a chat |
| Google Cloud (Gemini Enterprise Agent Platform) | veo-3.1-generate-001, fast and lite versions | Developers and businesses using the API |
| Gemini API | Veo 3.1 Preview shut down October 22, 2026 | Developers, who are pointed to Omni Flash or Google Cloud |
| Google Vids | AI video clips; Veo extend added June 2026 | Workspace and personal accounts making work videos |
| YouTube Shorts | Veo 3 Fast (announced 2025) | Shorts creators, free |
Why does the Gemini app row say Omni? Because Gemini Apps Help now describes video in the app as made “with Gemini Omni, our latest AI video generation model” (Gemini Apps Help). If a tutorial tells you to find Veo in the Gemini app’s model menu, it describes an older version of the app. The DeepMind Veo page still has a “Try in Gemini” button, but the help center is the more specific source.
The Gemini API change matters for developers. Google’s video page says Veo 3.1 Preview, including the fast and lite preview models, “was shut down on October 22, 2026,” and tells developers to migrate to Gemini Omni Flash “or the GA models available through the Gemini Enterprise Agent Platform” (Gemini API, video generation). GA means generally available, the stable release.
Google Vids is a quieter route. A Google Workspace Updates post from June 17, 2026 says Vids users can now “extend existing video clips using Veo,” and lists personal Google accounts among the editions that get it (Google Workspace Updates). The Vids help page adds that most users can generate up to 50 videos per month, at 720p or 1080p and 24 frames per second (Google Docs Editors Help, Vids).
YouTube Shorts is the free door. In its Made on YouTube 2025 post, YouTube said it had “started integrating Google DeepMind’s Veo 3 Fast into YouTube Shorts so you can easily create video backgrounds or clips with sound, for free” (YouTube Blog). That is an older Veo version tuned for Shorts, not the full Veo 3.1 you get in Flow. If Shorts is your goal, our guide to AI video generator apps for iPhone and Android covers the phone side.
Veo 3.1 Lite, Fast and Quality: Which Model You Get
Veo 3.1 comes in three tiers, and in Flow they cost 10, 20 and 100 credits per clip for non-Ultra subscribers (Google Flow Help, credits). The names describe a trade: Lite is the cheapest, Fast sits in the middle, and Quality is the slowest and most expensive. Which one should you start with?
| Tier in Flow | Credits per clip (non-Ultra) | Credits per clip (Ultra) | Clip lengths | Notable support |
|---|---|---|---|---|
| Veo 3.1 Lite | 10 | 5 | 4, 6, 8 seconds | Ingredients (8 s only), Extend |
| Veo 3.1 Fast | 20 | 10 | 4, 6, 8 seconds | Ingredients (8 s only), no Extend |
| Veo 3.1 Quality | 100 | 100 | Text and frames at 4, 6, 8 s; listed as 8 s on the credits page | No Ingredients, no Extend |
| Veo 3.1 Lite (Lower Priority) | Not available | 0 for Ultra subscribers and family plan managers | 4, 6, 8 seconds | Slower queue |
The Quality row has a wrinkle, and it was the biggest surprise when we read Google’s two Flow pages side by side. The Flow models page lists 4-, 6- and 8-second lengths for Quality text-to-video, while the credits page lists only “8s videos” next to the 100-credit price. Check the length menu in Flow before you plan around a short Quality clip.
On Google Cloud, the same family appears as three model IDs: veo-3.1-generate-001, veo-3.1-fast-generate-001 and veo-3.1-lite-generate-001 (Google Cloud, Veo 3.1). The standard model went GA on November 17, 2025, and the page lists a retirement date of “November 17, 2026 or later.” That date is a reminder to keep an eye on Google’s model versions page if you build a product on Veo.
There is one more model in Flow that is not Veo at all. Gemini Omni Flash 1.1 adds 10-second clips, a cheaper 360p draft mode at half the credit cost of 720p, and video-to-video editing, which none of the three Veo tiers support. Flow’s own docs list it next to Veo, so you can switch per clip.
Native Audio: Dialogue, Sound Effects and Music
Veo 3 and later generate sound with the picture: Google DeepMind says Veo “lets you add sound effects, ambient noise, and even dialogue to your creations – generating all audio natively” (Google DeepMind, Veo). You write the sound into the prompt, for example a line of dialogue in quotes and an “Audio:” note for effects.
DeepMind’s own example prompts show the pattern. One describes a detective interrogating a rubber duck and ends with “Audio: Detective’s stern quack, nervous squeaks from rubber duck.” Another puts a full spoken line in quotation marks so the character says it.
How good is the speech? Google is candid about the weak spot. The same page says natural, consistent spoken audio, “particularly for shorter speech segments, remains an area of active development.” Short lines can come out garbled, so plan to regenerate dialogue shots more often than silent ones.
Audio also changes the bill. On Google Cloud, Veo 3.1 is priced separately for “Video + Audio generation” and plain “Video generation,” and the audio version costs twice as much at 720p and 1080p (Google Cloud pricing). In Flow, the credit cost per clip is the same, and the Flow FAQ says that when Veo produces low-quality audio, the video is not generated and “your credits will be refunded” (Google Flow Help).
Clip Length, Aspect Ratios and Resolution, Including Vertical Video
A single Veo 3.1 clip is 4, 6 or 8 seconds long, in 16:9 or 9:16, at 24 frames per second, per Google Cloud’s model page (Google Cloud, Veo 3.1). Reference-image-to-video only supports 8 seconds. The page also lists up to 4 output videos per prompt and a 20 MB limit for input images.
Resolution depends on the tier and the app:
- Google Cloud, standard Veo 3.1: 720p, 1080p or 4K output.
- Google Cloud, Veo 3.1 Fast and Lite: 720p or 1080p.
- Google Flow: 1080p upscaling costs 0 credits for Plus, Pro and Ultra subscribers and is not available without a subscription; 4K upscaling costs 50 credits and is Ultra-only.
- Google Vids: 720p and 1080p.
What about vertical video? Every surface above lists 9:16, so a Short, a Reel or a TikTok-shaped clip comes straight out of the model without cropping. Flow’s models page simply says “Both aspect ratios” for each Veo 3.1 mode.
Need something longer than 8 seconds? You chain clips. Google’s October 2025 Veo 3.1 announcement said Extend can create “longer videos, even lasting for a minute or more,” with each new clip “generated based on the final second of your previous clip” (Google Blog). In Flow today, only Veo 3.1 Lite can run Extend, though it can extend any Veo 3.1 8-second clip.
Ingredients, Frames and Extend: Image and Reference Controls
Veo 3.1 takes three kinds of visual input: reference images (“ingredients”), a first and last frame, and an existing clip to extend, and DeepMind says these modes “achieved leading results in head-to-head comparisons” by human raters on Google’s internal benchmarks (Google DeepMind, Veo). Treat that as Google’s own claim, last updated October 2025.
Here is what each control does in plain terms:
- Ingredients to Video: you upload images of a character, an object or a style, and Veo keeps them consistent in the clip. DeepMind’s page says this now “includes audio.” In Flow, it works with Lite and Fast at 8 seconds; Quality does not support it.
- Frames to Video: you give a start image, or a start and an end image, and Veo animates the path between them. All three Veo 3.1 tiers in Flow support both versions at 4, 6 and 8 seconds.
- Extend: Veo continues a clip from its last second. Flow allows it on Lite only.
Flow’s help page adds a practical tip for ingredients: provide “subject or product references on a plain or segmented background,” and keep location and style references free of extra subjects (Google Flow Help, create videos). Your text prompt should describe how the ingredients interact, not re-describe them.
The Google Cloud tiers differ here too. The model page marks “Use reference images” as supported for standard and Fast, and not supported for Lite. Want to animate your own product photos rather than generate a scene from scratch? Our guide to animating your own photos with an image-to-video generator compares the tools built for that job.
Veo Limits by Google AI Plan: Flow Credits per Clip
Google no longer prints a fixed number of Veo videos per day; Flow gives everyone 50 credits a day, and Google AI plans add 200 to 25,000 credits a month on top (Google Flow Help, credits). Turn credits into clips and the plans become easy to compare.
The table uses the US prices on Google’s subscriptions page and the per-clip costs on the Flow credits page, both read on October 24, 2026. Clip counts are our arithmetic from the monthly credits alone, not a Google promise.
| Plan | US price | Flow credits | Veo 3.1 Lite clips | Veo 3.1 Fast clips | Veo 3.1 Quality clips |
|---|---|---|---|---|---|
| No plan | $0 | 50 a day | 5 a day | 2 a day | 0 |
| Google AI Plus | $4.99/month | 200 extra a month | 20 a month | 10 a month | 2 a month |
| Google AI Pro | $19.99/month | 1,000 extra a month | 100 a month | 50 a month | 10 a month |
| Google AI Ultra (5x) | $99.99/month | 10,000 extra a month | 2,000 a month | 1,000 a month | 100 a month |
| Google AI Ultra (20x) | $199.99/month | 25,000 extra a month | 5,000 a month | 2,500 a month | 250 a month |
Prices come from Google’s US plan page (Google AI plans). Ultra counts use the lower Ultra credit prices (5 for Lite, 10 for Fast). Ultra subscribers also get a 0-credit “Lower Priority” Lite option that the table leaves out.
Four rules change how far those credits go:
- Costs are “per generation, not per request.” Flow can make 2 videos from one request, which spends credits twice.
- The 50 daily credits refresh after your first generation each day and don’t roll over. Monthly plan credits refresh with your billing cycle and don’t roll over either.
- Without a subscription, Flow says you “may not be able to generate videos” during peak hours, “roughly 2 PM to 5 PM UTC.”
- You are not charged credits for failed generations, per the Flow FAQ.
One honest caveat about the free row. The Flow credits page lists 50 free daily credits and Google’s plan page lists Flow on the free tier, but Flow’s “Get started” page still lists a Google AI Plus, Pro or Ultra subscription as a requirement (Google Flow Help). Expect the free tier to be limited, and check what your own account shows.
What about the Gemini app? Its limits page describes “compute-based usage limits” that refresh “every 5 hours until you reach your weekly limit,” and says video “generally requires the most usage” (Gemini Apps limits). Video generation is not available there without a Google AI plan. Plus gets 2x the standard limits, Pro 4x, and Ultra 5x or 20x the Pro limits. For a wider look at free tiers across tools, see our roundup of free AI video tools and the limits on each free plan.
Veo API Pricing on Google Cloud
Google Cloud lists Veo 3.1 with audio at $0.40 per unit at 720p or 1080p and $0.60 at 4K, with Fast from $0.10 and Lite from $0.05 (Google Cloud, Agent Platform pricing). The page labels each price “/ 1 count,” and we could not find a line on it that defines one count.
| Model | Output | 720p | 1080p | 4K |
|---|---|---|---|---|
| Veo 3.1 | Video + audio | $0.40 | $0.40 | $0.60 |
| Veo 3.1 | Video only | $0.20 | $0.20 | $0.40 |
| Veo 3.1 Fast | Video + audio | $0.10 | $0.12 | $0.30 |
| Veo 3.1 Fast | Video only | $0.08 | $0.10 | $0.25 |
| Veo 3.1 Lite | Video + audio | $0.05 | $0.08 | Not listed |
| Veo 3.1 Lite | Video only | $0.03 | $0.05 | Not listed |
So is it per video or per second? Developer reports point to per second. On Google’s AI Developers Forum, a developer wrote that a single 8-second Veo 3.1 Lite video listed at $0.05 cost $0.40, “which clearly indicates it is billing $0.05 per second,” and a Google staff member replied that they were looking into making the pricing display clearer (Google AI Developers Forum).
If billing is per second, an 8-second clip with audio costs:
- Veo 3.1 at 1080p: 8 × $0.40 = $3.20. At 4K: $4.80.
- Veo 3.1 Fast at 1080p: 8 × $0.12 = $0.96.
- Veo 3.1 Lite at 720p: 8 × $0.05 = $0.40.
Confirm the unit in your own billing console before you scale a workload. The pricing page also says you’re charged “only for requests that return a 200 response code,” so blocked or failed requests don’t bill.
One more detail surprises people. The Veo 3.1 model page lists Provisioned Throughput and Fixed quota as consumption options, and marks Pay-as-you-go and Batch inference as “Not supported.” Provisioned Throughput means you buy reserved capacity in advance. Read Google’s quota pages before you assume you can call Veo on demand from a new project.
SynthID Watermarks and Commercial Use
Every Veo video carries an invisible watermark: the Flow FAQ says “All outputs generated in Google Flow using Veo, Omni, or Nano Banana include invisible SynthID watermarks” (Google Flow Help). On Google Cloud, the Veo 3.1 model page also lists Content Credentials (C2PA), a standard that records how a file was made.
What does SynthID do in practice? Google DeepMind says the watermark “doesn’t change the image or video quality” and is “designed to stand up to modifications like cropping, adding filters, changing frame rates, or lossy compression” (Google DeepMind, SynthID). You can upload a clip to Gemini and ask whether Google AI made it, and DeepMind’s SynthID Detector portal checks files too.
A visible watermark is a separate setting. Flow has a “Visible watermarking” toggle under your profile picture. The FAQ adds that a visible watermark “will be applied automatically for users in India, South Korea, and Vietnam who are not on the Google One AI Ultra 20x Plan.”
Can you use Veo clips commercially? Flow’s answer points to Google’s Terms of Service and quotes one line from them: “Google won’t claim ownership over that content.” That is about ownership, not a full license to use any likeness or brand you put in a prompt. Flow also blocks “generating videos of prominent people” in every region, and the Prohibited Use Policy still applies to everything you make.
Veo vs Kling and Runway: Which to Use
Veo 3.1 makes the shortest clips of the three, 4 to 8 seconds with native audio and up to 4K, while Runway’s Gen-4.5 runs 2 to 10 seconds at 720p and Kling 4.0 lists “native 30s-generation” (Runway Help Center; Kling AI membership plans). All three figures come from the vendors’ own pages, read on October 24, 2026.
| Spec | Google Veo 3.1 | Runway Gen-4.5 | Kling 4.0 |
|---|---|---|---|
| Clip length | 4, 6 or 8 seconds | 2 to 10 seconds | Up to 30 seconds |
| Sound | Native, in the model | Separate music, voice and sound effects tools | Kling 3.0 lists synchronized audio-video |
| Output resolution | 720p, 1080p or 4K (Google Cloud) | 720p, with 4K upscaling on paid plans | 1080p and 4K listed on paid plans |
| Vertical 9:16 | Yes | Image-to-video only | Not stated on the plan page |
| Lowest paid entry | Google AI Plus, $4.99/month | Standard, $12/month billed yearly | Prices shown as first-month promotions, so not quoted |
| Credit math | 10 to 100 Flow credits per clip | 12 credits per second | Credits per plan |
How do you choose? Pick Veo when you want sound and picture from one prompt, or a vertical clip at high resolution. Pick Runway when you want precise camera direction: its help page says Gen-4.5 handles “detailed camera choreography, intricate scene compositions, precise timing of events” in one prompt, and its pricing page says Standard’s 625 monthly credits equal 52 seconds of Gen-4.5. Our breakdown of Runway’s Gen-4.5 plans and credit costs goes deeper. Pick Kling when one continuous shot must run longer than 10 seconds.
None of these tables tell you which one looks best on your prompts. Only your own test can answer that.
Two other jobs point away from all three. If the video needs a person talking to camera for several minutes, a clip model is the wrong tool; see what Synthesia’s presenter videos include on each plan. If you only need a short clip inside designs you already make, read how Canva’s built-in clip generator fits a design workflow before you pay for a separate plan. For how Veo stacks up against avatar tools, script-to-video tools and the rest, see how Veo compares with the other AI video generators we track.
What Happened to Sora, and Whether Veo Replaces It
OpenAI’s help center says “The Sora web and app experiences were discontinued on April 26, 2026” and lists September 24, 2026 as the date the Sora API is discontinued (OpenAI Help Center). So the old “Veo vs Sora” question now has a simple answer: only one of them is still running.
If you had Sora videos, the same page says you can export them by “going to sora.chatgpt.com/sunset and clicking on Export,” and that after any final export window OpenAI “will permanently delete any data associated with your use of Sora.” Purchased ChatGPT and Sora credits can still be spent on Codex.
Is Veo the natural replacement? For short, realistic clips with sound, it is the closest match among the tools still on sale. Be careful with search results, though. Many pages that still rank for “Sora AI video generator” are third-party sites offering a “Sora 2” generator, not OpenAI. Check the domain before you hand over a card.
Sora’s shutdown carries a lesson for Veo users too. Google retired the Veo 3.1 Preview in the Gemini API on October 22, 2026, and Google Cloud lists a retirement date for the current Veo 3.1 model. Keep your prompts, reference images and exported clips in storage you control. For a primer on how clip generators like these work and bill, see the basics of how clip generators bill and what they produce.
How to Make a Veo Video in Google Flow
To make a Veo video in Flow, open a project, pick Video and a Veo 3.1 model in the prompt box, set the aspect ratio, number of outputs and length, then click Generate, per Google’s Flow help page (Google Flow Help, create videos). Flow runs best in a Chromium-based browser such as Chrome or Edge.
The steps, as Google documents them:
- Check you qualify: Flow says you must be 18 or older, age-verified, and in a supported country. The United States, the United Kingdom, Canada and Australia are on the list.
- Go to Google Flow on a computer and open a project or start a new one.
- Write the prompt. Google’s tip reads: “Be specific about the subject, action, environment, lighting, and style.” Add dialogue in quotes and sounds after “Audio:”.
- Click the model name. Flow defaults to Nano Banana Pro, an image model, so switch to Video.
- Optional: add ingredients (Video, then Ingredients) or frames (Video, then Frames, then drag images to “+ Add start frame” and “+ Add end frame”).
- Pick aspect ratio, number of outputs, the model (Veo 3.1 Lite, Fast or Quality) and the length.
- Click Generate. Check the credit cost in Settings first; Flow says “Model costs are evolving fast.”
Would you rather describe the video and let software choose the steps? Flow also has an Agent mode. Turn on Agent in the prompt box, describe the scene, and ask for variations, such as “Give me 5 variations of this video with different lighting.” Each variation is a generation, so each one spends credits.
A tip from Flow’s own FAQ: when you see “You’re requesting generations too quickly,” Google is rate-limiting, “particularly for zero-credit models.” Slow down rather than retrying in a loop.
Before You Pay for Veo: Your Next Step
Before you pay for a Google AI plan for Veo, spend Flow’s 50 free daily credits on five Veo 3.1 Lite clips of one real brief, because at 10 credits a clip that is the full free allowance (Google Flow Help, credits). If Lite looks good enough, Google AI Plus at $4.99 a month adds 20 more Lite clips a month. If you need Quality, budget 100 credits a clip.
Not sure which plan that points to? A short checklist to take with you:
- Which surface do you need? Flow for picking the Veo tier, Google Cloud for the API, Shorts for free vertical clips.
- How long is your longest shot? Veo stops at 8 seconds per clip; longer scenes mean Extend or editing.
- Do you need sound from the model? Veo 3.1 makes it natively; budget extra retries for dialogue.
- Which resolution will you publish? 1080p upscaling is free for subscribers in Flow; 4K needs Ultra.
- Will you use clips for clients? Read Google’s Terms and the Prohibited Use Policy first.
- Where will your exports live? Keep copies outside Flow.
Planning to cut those clips into a longer video? Our guide to editing software for finishing YouTube videos covers the next step.
References
- Google Blog: Introducing Veo 3.1 and advanced capabilities in Flow
- Google Flow Help: Manage your Google Flow credits
- Google Flow Help: Learn about Google Flow models & supported features
- Google Flow Help: Create videos in Google Flow
- Google Flow Help: Get started with Google Flow
- Google Flow Help: Where you can use Google Flow
- Google: Google AI plans (US)
- Gemini Apps Help: Generate videos with Gemini Apps
- Gemini Apps Help: Gemini Apps limits & feature availability
- Gemini API: Video generation in the Gemini API
- Google Cloud: Veo 3.1
- Google Cloud: Agent Platform pricing
- Google AI Developers Forum: Veo 3.1 Lite pricing thread
- Google Docs Editors Help: Use AI to generate video clips in Google Vids
- Google Workspace Updates: Create longer Veo videos and generate multiple at once
- YouTube Blog: Made on YouTube 2025
- Google DeepMind: Veo
- Google DeepMind: SynthID
- Runway Help Center: Creating with Gen-4.5
- Runway: Pricing
- Kling AI: Membership plans
- OpenAI Help Center: What to know about the Sora discontinuation