Gemini Omni Flash Review: Google's Veo 3.1 Replacement at $0.10 a Second

On October 22, 2026, Google shut down every Veo 3.1 preview model on the Gemini API — Standard, Fast, and Lite — and pointed developers at its replacement: Gemini Omni Flash 1.1. The date was listed on Google's deprecations page in advance, but a listed date and a working migration are different things, and plenty of scripts that called veo-3.1-generate-preview stopped working that day.

The replacement is not another Veo. Omni Flash is a Gemini model that generates video from text, images, audio, and video, and then lets you edit the result by describing the change in plain words. It also costs about a quarter of what Veo 3.1 Standard charged per second on the API — $0.10 against $0.40 — while adding the one thing Veo never did in the API: editing real footage. The switch is cheaper and, in some ways, more capable. It is also worse in others, and the free route and the developer route now carry very different price tags.

The short version: Gemini Omni Flash 1.1 is Google's general-availability video model on the paid tier of the Gemini API, at roughly $0.10 per second of 720p video with no free tier. In Google Flow, the same model costs 7 to 15 credits a clip, and every account gets 50 free credits a day. It makes 3- to 10-second clips at up to 4K, and it edits uploaded video — but the API publishes no per-second token rate for 1080p or 4K, and no independent head-to-head test against Veo 3.1 exists yet.

What happened to Veo 3.1

Google's own documentation tells the story in three dates. Gemini Omni was announced at Google I/O on May 19, 2026, as a model that could "create anything from any input — starting with video," with the first model, Gemini Omni Flash, rolling out the same day to the Gemini app, Google Flow, and YouTube Shorts. The stable 1.1 version, gemini-omni-1.1-flash, shipped on August 27, 2026. Then, on October 22, 2026, Google shut down the Veo 3.1 preview models on the Gemini API and the original Omni Flash preview, naming gemini-omni-1.1-flash as the recommended replacement for all of them.

Two things about that table are easy to get wrong, so here they are plainly. First, Veo 3.1 is not dead. The shutdown hit the preview models on the Gemini API only. Veo 3.1 still runs inside Google Flow, and Google's deprecations page points developers who need Veo to the general-availability models on the Gemini Enterprise Agent Platform instead. Second, the Gemini API's video generation documentation now opens with a warning — "The Veo 3.1 Preview in the Gemini API has been shut down" — and tells developers to use Gemini Omni Flash as the model for video generation and conversational video editing. If your code still calls a veo-3.1-*-generate-preview model ID, it stopped working on October 22.

What Gemini Omni Flash actually is

Omni Flash is a fast, multimodal video model. Google's API documentation describes it as offering "superior video coherence, multi-input reasoning (supporting text, images, audio, and video inputs simultaneously), character consistency, factual accuracy, and multi-turn conversational editing." In practice, that means three things Veo 3.1 never combined in one model: you can start from any mix of inputs, you can keep a character or object consistent across shots, and you can change a finished clip by typing what to change instead of re-rolling it.

The published specifications, from Google's model page for gemini-omni-1.1-flash, are:

• Output: 3 to 10 seconds of video at 360p, 720p, 1080p, or 4K, at 24 frames per second, with audio generated alongside the picture. There is no switch to turn the audio off.
• Input: text, images, and video — up to 10 seconds of video for editing and extension. The context window is 1,048,576 tokens.
• Editing: multi-turn conversational editing through the Interactions API, including element replacement and perspective changes.
• Availability: generally available on the paid tier of the Gemini API, in the Gemini app for Google AI Plus, Pro, and Ultra subscribers, in Google Flow, and at no cost on YouTube Shorts and the YouTube Create app.

The 10-second cap is the headline limit. Veo 3.1 topped out at 8 seconds, so Omni Flash buys two extra seconds per clip, but anything longer still means chaining clips. Google has said longer durations are coming, according to its developer posts, but no date is published. The 1-million-token context window matters more than it sounds: it is what lets the model hold a reference video, several images, and a conversation about edits in a single session.

What it costs on the API

This is where the switch gets concrete. Google's Gemini API pricing page lists Gemini Omni Flash as generally available on the paid tier, with no free tier at all:

• Input: $1.50 per million tokens, for text, image, video, and audio inputs alike.
• Output: $9.00 per million tokens for text, and $17.50 per million tokens for video, including thinking tokens.
• The conversion: billing is based on total output token consumption, calculated at 5,792 tokens per second of 720p video. Google's own arithmetic on the pricing page puts that at "an effective price of approximately $0.10 per second."

Run the numbers yourself and they check out. An 8-second clip at 720p consumes 46,336 video-output tokens, which at $17.50 per million tokens is $0.81. A 10-second clip is 57,920 tokens, or $1.01. A full minute of 720p footage works out to about $6.08. The text prompt itself is nearly free — a 100-token prompt costs about $0.00015 at the $1.50 input rate — but reference images and input video are tokenized too, and Google does not publish a per-image or per-second input token count on the pricing page, so budget a small margin when your prompt carries heavy references.

Compare that with what Veo 3.1 charged before the shutdown. On the Gemini API, Veo 3.1 Standard cost $0.40 per second at 720p or 1080p and $0.60 at 4K, with audio included — so an 8-second clip was $3.20, roughly four times Omni Flash's $0.81. Veo 3.1 Fast was $0.10 per second at 720p, essentially the same per-second price as Omni Flash, and Lite was $0.05. The honest reading: Omni Flash does not undercut every Veo tier. It undercuts the quality tier by about four times while matching the Fast tier's price — and it adds video editing, 10-second clips, and multi-input prompts that neither Veo tier offered through the API.

One gap in the pricing page is worth flagging. The 5,792-tokens-per-second figure is stated for 720p video only. Google's model page lists 1080p and 4K as output options, but the pricing page publishes no per-second token conversion for those resolutions, so the "$0.10 a second" figure is a 720p figure. Do not quote it for 4K output until Google publishes the conversion.

What it costs in Google Flow

Most people will never touch the API. They will use Google Flow, Google's creative studio, where Omni Flash is billed in credits instead of tokens. Google's own Flow credits help page lists the per-generation costs:

The allowances behind those costs come from the Google AI plan you hold. Every account gets 50 Flow credits a day, free, with no subscription. Google AI Plus at $9.99 a month adds 200 monthly credits, Pro at $19.99 adds 1,000, and the two Ultra tiers add 10,000 ($99.99 a month) and 25,000 ($199.99 a month). Google notes on its plans page that prices may vary by market.

What that buys, in clips: the free daily 50 credits cover four 8-second Omni Flash clips at 720p (48 credits, 32 seconds of footage) or seven 10-second clips at 360p (49 credits, 70 seconds). A Pro subscriber's 1,000 monthly credits buy about 83 eight-second 720p clips before counting the daily grant. Three rules shrink those numbers in practice. Credits are charged per generation, not per request, and Google's own example is a single Flow request that produces two videos — leave the output count at two and every clip costs double. Unused daily and monthly credits never roll over. And free users can be blocked from video generation during peak hours, roughly 2 PM to 5 PM UTC, while subscribers skip the block.

The cheapest way to use Omni Flash is also the most obvious: draft at 360p, where an 8-second clip costs 6 credits instead of 12, then render the keeper at 720p. A single video edit costs 40 credits — 80% of a free day's allowance — so edits are the expensive operation in Flow, not generations.

Omni Flash vs the field

*Per-second figures are the vendors' published rates as of October 2026, each on a different billing basis — token counts, credit bundles, and clip minimums all differ, so treat them as directional, not exact. Seedance figures are BytePlus ModelArk estimates for a 5-second 16:9 clip with no video input, as compiled in an October 2026 pricing survey. Kling's figure is the standard tier's per-second rate on an aggregator's rate card ($0.12 silent, $0.18 with sound; the Pro tier reaches $0.24). Veo 3.1's rates are the retired Gemini API preview rates, as recorded on October 13, 2026.

Two rows in that table deserve a second look. Seedance 2.5 is the only rival that clearly beats Omni Flash on clip length — up to 30 seconds against 10 — and its 480p rate is essentially the same price per second. But its 1080p rate jumps to about $0.57 a second, nearly six times Omni Flash's 720p rate, and ByteDance's Dreamina subscription is running a first-month promotion ($0.29 for the first month, with promotional per-second rates) whose end date is unconfirmed — the promo rate is not the standing price. Kling 3.0, meanwhile, runs 3 to 15 seconds — longer than Omni Flash — but its standard tier costs $0.12 to $0.18 a second depending on sound, and its Pro tier reaches $0.24.

The comparison that matters most is the one inside Google. Omni Flash at $0.10 a second matches Veo 3.1 Fast's 720p rate, beats Standard by four times, adds two seconds per clip, and edits video. What it does not do is match Veo 3.1 Quality's output. In independent side-by-side testing of 48 renders, the only model that delivered a slow-motion liquid pour and a low car-tracking shot was the Veo line — Omni Flash was not part of that test, so the gap is unmeasured rather than proven, but nobody has published evidence that Omni Flash matches Veo Quality on cinematic hero shots yet. For the wider field of video models and what each one costs, our Sora alternatives roundup and our Kling 4.0 review cover the other contenders, including FLUX 3 Max's 12-second clips.

Pros and cons

Pros:

• About a quarter of Veo 3.1 Standard's per-second API price, and equal to Veo Fast at 720p.
• Edits real video from a text prompt, with multi-turn refinement — Veo never offered this through the API.
• 10-second clips, two seconds longer than Veo 3.1's maximum.
• Free to try properly: 50 Flow credits a day buy four 8-second 720p clips, and it costs nothing on YouTube Shorts.
• 360p drafts cost half the 720p rate in Flow, so iterating is cheap.

Cons:

• No free tier on the API — every developer second is billed, and input tokens for reference images and video add to the bill.
• 10-second cap means any real scene is still a chain of clips.
• Google publishes no per-second token rate for 1080p or 4K output, so the headline $0.10 figure stops at 720p.
• No independent benchmark head-to-head against Veo 3.1 or Seedance has been published; all quality claims are Google's own.
• Flow's per-generation billing punishes retries — a two-output request doubles the cost, and one edit eats 80% of a free day.

The catches before you switch

First, check what your code actually calls. The shutdown was of the preview model IDs, not of Veo itself, and the fix is a one-line model change to gemini-omni-1.1-flash — but the request and response shapes differ, because Omni Flash takes video and audio inputs and returns edited video. A straight model-ID swap compiles; it does not guarantee the same output.

Second, do not budget from the Flow credit table if you are building on the API, or vice versa. Flow's 12 credits for an 8-second 720p clip and the API's $0.81 for the same clip are two different currencies, and neither converts cleanly into the other. Pick the route you will actually use and budget in its units.

Third, the "free" story has fine print. The API has no free tier. Flow's 50 daily credits are real, but they vanish at the next refresh, free users lose video access during peak hours, and the 1080p upscale that costs subscribers nothing is not available without a plan at all. The genuinely free route is YouTube Shorts, where Google says Omni Flash rolls out at no cost — with Shorts' own limits on length and format.

Finally, treat every quality claim in this article as Google's until someone independent tests it. Google's documentation calls Omni Flash's coherence "superior" and its editing "multi-turn," and those are the company's words, not measurements. The model has been generally available for two months, and the independent evaluations that would settle the Veo Quality question have not been published yet.

FAQ

Is Veo 3.1 gone?

No. Only the Veo 3.1 preview models on the Gemini API shut down on October 22, 2026. Veo 3.1 still runs in Google Flow, and Google points developers to the general-availability models on the Gemini Enterprise Agent Platform. What ended is the preview API route, not the model.

Is Gemini Omni Flash free?

It depends on the route. The Gemini API has no free tier — every second is billed at about $0.10 per second of 720p video. Google Flow gives every account 50 free credits a day, which covers four 8-second 720p clips. In the Gemini app it requires a Google AI Plus, Pro, or Ultra subscription. On YouTube Shorts and the YouTube Create app, Google says it rolls out at no cost.

How much does an 8-second clip actually cost?

On the API, about $0.81 at 720p — 46,336 output tokens at $17.50 per million — plus a negligible amount for the text prompt and a variable amount for any reference images or input video. In Flow, 12 credits at 720p or 6 at 360p. A 10-second clip is about $1.01 on the API and 15 credits in Flow.

Can Gemini Omni Flash edit my own video?

Yes. You can upload up to 10 seconds of video for editing and extension, and refine it across multiple conversational turns. In Flow, each edit costs 40 credits regardless of length. On the API, edits are billed on the same input and output token rates as generations.

Does it work in the Gemini app?

Yes, for Google AI Plus, Pro, and Ultra subscribers. That is the same subscription that sets your Flow credit allowance, so the app route and the Flow route share one plan.

Key takeaways

• Google shut down all three Veo 3.1 preview models on the Gemini API on October 22, 2026, and named Gemini Omni Flash 1.1 the replacement.
• On the API, Omni Flash costs about $0.10 per second of 720p video — roughly a quarter of Veo 3.1 Standard's $0.40 — with no free tier.
• In Google Flow, an 8-second 720p clip costs 12 credits, and every account gets 50 free credits a day.
• Omni Flash makes 3- to 10-second clips and edits real video from a text prompt, which Veo 3.1 never did through the API.
• Veo 3.1 is not dead — it still runs in Flow and on Google's enterprise platform — and no independent test has yet shown Omni Flash matching Veo Quality on cinematic shots.

The price is genuinely lower, the editing is genuinely new, and the 10-second cap is genuinely still there. If you were paying Veo Standard rates, switch now. If you were paying Veo Fast rates, the money is the same — the edit button is the reason to move.

Sources: Google's Gemini API pricing page, deprecations page, video generation documentation, and Gemini Omni Flash model page; Google's I/O 2026 keynote coverage; Google Flow's credits help page; Google One's plans page; an October 2026 pricing survey of 19 providers; and independent side-by-side render testing of 48 clips. All prices as published on October 27, 2026, and subject to change — check the live price before committing spend.