Gemini Omni 1.2 Flash is Google’s newest video-generation model, and it changes the arithmetic of AI video in a way that matters for anyone who makes clips for a living. Launched on November 5, 2026, it generates and edits video natively up to 15 seconds long — five seconds more than its predecessor — while cutting the price by 30% to about $0.07 per second at 720p. A full 15-second clip at 720p costs roughly $1.04 in video output on the Gemini API.
That is not a rounding error in a pricing table. It is the difference between AI video as an expensive experiment and AI video as a line item you can budget for a real campaign.
This guide explains what Gemini Omni 1.2 Flash actually does, what a 15-second clip really costs once you count the whole bill, where you can use it — the Gemini API, the Gemini app, Flow, Google Vids and Google Ads — and how it stacks up against Kling 4.0 and HeyGen Video. Everything here is based on Google’s own launch announcement and the official Gemini API pricing page, checked on November 11, 2026, with the company’s own claims labelled as such.
If you are deciding whether to build your next short ad, product demo or social clip on Google’s video model, this is the full picture before you spend a dollar.
Table of Contents
- What Is Gemini Omni 1.2 Flash?
- What Is New in Gemini Omni 1.2 Flash vs Omni 1.1
- Gemini Omni 1.2 Flash Pricing: What a 15-Second Clip Really Costs
- How to Use Gemini Omni 1.2 Flash
- Conversational Editing in Gemini Omni 1.2 Flash
- Reference Following: Characters, Products and Style That Stay Put
- Gemini Omni 1.2 Flash vs Kling 4.0 vs HeyGen Video
- Who Should Use Gemini Omni 1.2 Flash?
- Gemini Omni 1.2 Flash Limitations to Know Before You Pay
- Gemini Omni 1.2 Flash FAQ
- The Verdict
What Is Gemini Omni 1.2 Flash?
Gemini Omni 1.2 Flash is Google DeepMind’s next-generation video generation and editing model, and it is the model behind the video features in the Gemini app, Google Flow and Google Vids. Its model ID on the Gemini API is gemini-omni-1.2-flash, and Google made it generally available on November 5, 2026, across the Gemini API and Gemini for business. “Generally available” is doing real work in that sentence: this is not a research preview or a waitlist.
Developers can call it from the paid tier of the Gemini API, businesses can use it through Google Cloud, and creators can reach it inside Google’s own apps.
At its core, Gemini Omni 1.2 Flash does two things in one model. First, it generates video from a text prompt, with native audio — dialogue, sound effects and music — produced alongside the pictures rather than bolted on afterwards. Second, it edits video: you can hand it an existing clip or image and ask it to change the background, swap a product, alter the lighting or reframe the shot, and it re-renders the result.
The “Omni” in the name is Google’s shorthand for that multimodal reach: text, images, short motion clips and audio cues all feed into a single generation.
The three numbers that define this release are 15 seconds, 30% cheaper, and $0.07 per second. Fifteen seconds is the new maximum clip length in a single generation, up from 10 seconds with Gemini Omni 1.1 Flash. Thirty percent is the price cut on video output. And $0.07 per second is the effective rate for 720p video on the Gemini API — the number Google puts in its launch headline and the number most creators will actually plan around.

It helps to place Gemini Omni 1.2 Flash in the family it comes from. The original Gemini Omni arrived in May 2026 as Google’s first video model built on the Gemini architecture, and it quickly became the engine inside Google’s video products — Flow, Vids and the Gemini app all run on it. Omni 1.1 Flash refined that foundation.
Gemini Omni 1.2 Flash is the first version Google is positioning as the default workhorse: it is the version that ships with a longer clip ceiling, a lower price, stronger consistency and enterprise backing through Google Cloud, including Provisioned Throughput for teams that need guaranteed capacity. Every clip it produces also carries an imperceptible SynthID watermark, Google’s way of marking AI-generated media so it can be identified later.
For a US creator or small business, the practical question is simpler than the spec sheet: can this model make a usable 15-second ad, product demo or social clip, at a price that survives contact with a real budget, without the faces melting halfway through? The rest of this guide answers that question piece by piece.
What Is New in Gemini Omni 1.2 Flash vs Omni 1.1
Gemini Omni 1.2 Flash is an upgrade on four fronts at once: length, price, consistency and control. None of these is a revolution on its own. Together, they change what the model is good for. Here is the side-by-side that matters, drawn from Google’s launch announcement and the official pricing page.
| Capability | Gemini Omni 1.1 Flash | Gemini Omni 1.2 Flash |
|---|---|---|
| Maximum clip length | 10 seconds | 15 seconds |
| Duration control | Fixed steps | 3 to 15 seconds, in one-second increments |
| Video output price | $17.50 per 1M tokens (~$0.10/sec at 720p) | $12.00 per 1M tokens (~$0.07/sec at 720p) |
| Price change | — | About 30% cheaper |
| Multi-turn conversational editing | Limited | Full 15-second clips, turn by turn |
| Reference following | Basic | Character, product and style images, motion clips up to 3 seconds, audio cues |
| Dialogue and lip-sync | Basic | More natural speech, clearer lip-sync, multi-speaker conditioning |
| Camera control | Basic | Continuous FPV flythroughs, zoom-outs, orbits, fewer jump cuts |
| Resolutions | Up to 1080p | 360p, 720p, 1080p and 4K (1080p/4K via upsampling) |
| Availability | API | Generally available: API, Gemini for business, Flow, Vids, Gemini app, Google Ads |
Longer Clips: 15 Seconds Changes the Story You Can Tell
The jump from 10 to 15 seconds sounds modest until you try to fit a real message into a clip. Ten seconds buys you a punchline or a product reveal. Fifteen seconds buys you a setup, a beat and a payoff — a customer problem, the product solving it, and the call to action — inside a single generation. Google’s announcement makes exactly this point: the extra five seconds give you room to tell complete stories, create product explainers or produce full social ad spots without stitching two generations together.
Duration is also more flexible than before. You can specify the length in one-second increments from 3 to 15 seconds, so a 7-second bumper and a 13-second demo cost you exactly what they use rather than forcing you into a fixed bucket. One exception to note: first-and-last-frame workflows — where you give the model a starting frame and an ending frame and it interpolates between them — are limited to 3 to 10 seconds, according to the model documentation. If your workflow depends on keyframe interpolation, plan around that ceiling.
Stronger Consistency Across the Whole Clip
The second upgrade is the one creators will feel most in daily use: less visual drift. Google says Gemini Omni 1.2 Flash holds character, object and scene consistency from the first frame to the last, with less morphing across motion and camera transitions. That matters because drift is the tax every AI video user pays — the face that subtly changes, the logo that warps, the background that rearranges itself between shots.
A 15-second clip is 50% longer than a 10-second one, which means 50% more frames for something to go wrong. The consistency work in this release exists precisely because the length increase would otherwise have made drift worse.
Google illustrates the point with a demanding test: given the same static studio-lighting exhibition poster and a five-layer animation prompt, Omni 1.1 Flash failed to match the reference image as closely as 1.2 does. The newer model builds the poster sequentially, layer by layer, and locks every material, shadow and typographic detail in place through the final frame. That is a company-chosen example, not an independent benchmark, so treat it as a direction of travel rather than a guarantee — but the direction is the right one for anyone making branded video.
More Natural Speech and Smoother Camera Moves
Two more upgrades round out the release. Dialogue is described as more fluid and natural-sounding, with clearer lip-sync and distinct multi-speaker conditioning — the model can route different voices to different characters in the same scene, which is what makes a two-person conversation look like a conversation instead of a monologue with a bystander.
And camera work gets steadier: Google claims complex, continuous movements such as sweeping FPV flythroughs, dramatic zoom-outs and orbits now come with significantly fewer unwanted jump cuts and steadier scale. The launch example is a single continuous FPV drone flight across a piazza, through bronze doors and up through the Pantheon’s coffered dome — with, Google says, zero jump cuts.
Gemini Omni 1.2 Flash Pricing: What a 15-Second Clip Really Costs
Pricing is where Gemini Omni 1.2 Flash makes its strongest case — and where the fine print matters most. Google bills the model by the token, not by the second, and the famous $0.07 figure is a conversion of that token price. Understanding the conversion is the difference between budgeting correctly and being surprised by your bill.
Here is the official pricing, from the Gemini API pricing page as checked on November 11, 2026. All figures are in US dollars.
| What you are billed for | Rate |
|---|---|
| Input — text, image or video | $1.50 per 1M tokens |
| Output — text (including thinking tokens) | $9.00 per 1M tokens |
| Output — video | $12.00 per 1M tokens |
| Free tier | Not available |
The video output rate converts to a per-second price through a fixed token rate: 5,792 tokens per second of 720p video. Multiply that by $12.00 per million tokens and you get $0.0695 per second — the “about $0.07” in Google’s headline. The rate scales cleanly with resolution, because the token rate per second scales with it:
| Resolution | Tokens per second | Price per second | 8-second clip | 15-second clip |
|---|---|---|---|---|
| 360p (draft) | 1,931 | $0.023 | $0.19 | $0.35 |
| 720p | 5,792 | $0.070 | $0.56 | $1.04 |
| 1080p (upsampled) | 8,688 | $0.104 | $0.83 | $1.56 |
| 4K (upsampled) | 17,376 | $0.209 | $1.67 | $3.13 |
Three things about this table deserve emphasis. First, the 360p draft tier costs exactly one-third of 720p, which makes it the sensible place to test a prompt before committing to a full-price render. Second, 1080p and 4K are produced by upsampling rather than native generation at that resolution, so you are paying for a sharper delivery format, not for a fundamentally different generation.
Third, these figures cover the video output only. Your prompt text, reference images and any text output the model produces are billed separately at the input and text-output rates. On a typical short generation those extras are small — a few cents at most — but they are not zero, and they are exactly the kind of line item that surprises a first-time API user.
The 30% Cut, in Real Money
Against Gemini Omni 1.1 Flash, the saving is straightforward. The older model billed video output at $17.50 per million tokens, which worked out to about $0.10 per second at 720p. The new $12.00 rate is a cut of roughly 31% on the token price, matching Google’s “30% cheaper” headline. In practical terms:
- One 15-second clip at 720p: about $1.04 with Gemini Omni 1.2 Flash, versus about $1.51 with the old rate — a saving of roughly $0.47 per clip.
- Ten 15-second clips at 720p: about $10.40 versus about $15.10.
- One hundred 15-second clips at 720p: about $104 versus about $151 — a saving of roughly $47 on a campaign-sized batch.
- Drafting at 360p: a 15-second test costs about $0.35, so you can run three drafts for roughly the price of one 720p final.
That last point is the workflow Google is steering you toward, and it is genuinely good advice: draft cheap at 360p or 720p, fix details with a follow-up edit instead of re-generating the whole clip, and upscale only your best creations to 1080p or 4K. Every re-generation you avoid is money saved, and the multi-turn editing in this release exists precisely to make that workflow work.
What the API Price Does Not Include
Before you multiply $0.07 by your shot list, count the costs that sit outside the per-second rate. The honest budget for a finished, publishable clip includes the failed attempts and the revisions, not just the one generation you keep. A prompt that needs three tries costs three times the headline rate. A clip you edit four times costs the original generation plus four edit passes.
And if you are generating through a partner platform rather than the API directly, you are paying that platform’s rates, not Google’s — Runway and Magnific, both named as launch partners, set their own credit prices.
There is also no free tier for this model on the Gemini API. The pricing page lists the free tier column for Gemini Omni 1.2 Flash as “Not available,” which means every API generation is billed at paid rates from the first second. That is a real difference from some of Google’s text models, where a free tier lets you experiment without a billing account. If you want to try the model without opening a paid API account, the consumer routes — the Gemini app, Flow and Google Vids — are the way in, and we cover those below.
Flow Credits: A Different Currency for the Same Model
If you generate inside Google Flow rather than through the API, you pay in AI credits, not dollars. A 15-second 720p clip in Flow costs 60 AI credits: three credits per second for the first ten seconds, then six credits per second for the extra five. Flow access requires a Google AI Plus, Pro or Ultra subscription, so the “cost” of a clip there depends on which plan you hold and how many credits it includes. API dollars and Flow credits are separate billing systems — do not try to convert one into the other.
The practical rule is simple: if you are a creator working inside Google’s apps, think in credits; if you are a developer or a business building video into a product, think in API dollars.

The December 3 Switchover
One date belongs in every budget note. From December 3, 2026, API requests to the older gemini-omni-1.1-flash model will automatically route to Gemini Omni 1.2 Flash and be billed at the newer model’s rates. If your code or pipeline still names the 1.1 model, it will keep working — but it will silently become 1.2, at 1.2 prices, with 1.2 behaviour. That is good news on price and mostly good news on quality, but if you have prompts tuned precisely to 1.1’s quirks, test them against 1.2 before December 3 rather than discovering the difference in production.
Google also lists a retirement date of November 5, 2027 or later for this version, so treat the model ID as stable for about a year and plan a migration check before then.
How to Use Gemini Omni 1.2 Flash
There are five main ways to reach Gemini Omni 1.2 Flash, and the right one depends on who you are: a developer, a business team, or a creator working inside Google’s apps. They all run the same model, but the price, the interface and the limits differ.
1. The Gemini API and Google AI Studio (for Developers)
The most direct route is the Gemini API, where the model ID is gemini-omni-1.2-flash. You can prototype in Google AI Studio without writing code: pick the model, type a prompt, attach reference images, and generate. For production use, call the API from your own application, where you control resolution, duration, aspect ratio and reference inputs, and you pay the token rates listed above.
This is the only route with no free tier — every call is billed — so set up billing and a budget alert before you start experimenting at scale. Google also offers Provisioned Throughput through Google Cloud for teams that need guaranteed capacity rather than shared, best-effort access.
2. The Gemini App (for Everyone)
Inside the Gemini app, you can direct scenes and make multi-turn edits in plain language — no API key, no code. The 15-second generation is rolling out to Google AI Ultra subscribers, while other supported users get shorter clips: Google’s help pages indicate clips run 3 to 10 seconds for supported users, with up to 15 seconds reserved for AI Ultra.
That tiering is worth knowing before you subscribe for the feature: if 15 seconds is the reason you are paying, Ultra is the plan that unlocks it. Note also that video editing and avatar creation are unavailable in the European Economic Area, Switzerland, the United Kingdom and some US states, so availability depends on where you are.
3. Google Flow (for Filmmakers and Creators)
Flow is Google’s AI filmmaking tool, and it is where the conversational editing in Gemini Omni 1.2 Flash feels most at home. You can direct multi-character dialogue scenes, refine a clip turn by turn, and build sequences from a single starting idea. Generation in Flow costs AI credits — 60 credits for a 15-second 720p clip, as detailed above — and requires a Google AI Plus, Pro or Ultra subscription. If your work is narrative or cinematic rather than a single ad spot, Flow is the interface that makes the model’s strengths visible.
4. Google Vids (for Business Video)
Google Vids, the Workspace video app, offers 15-second generation at no additional cost with a Google account or Google Workspace subscription. This is the cheapest way to try the model: if you already pay for Workspace, the video generation is included rather than metered. It is aimed squarely at workplace video — explainers, training clips, internal announcements, polished videos with avatars, music and narration — rather than at filmmakers, but the underlying model is the same.
5. Google Ads and Gemini for Business (for Teams)
For businesses, the model is available through the Google Cloud console and the enterprise API, backed by Google Cloud’s infrastructure and security. Advertisers can also scale video campaigns in Google Ads, where the multi-turn editing becomes a practical tool: generate one hero clip, then produce variants for different audiences and placements — different backgrounds, different aspect ratios, different lighting — without re-briefing the model from scratch each time. That variant workflow is one of the clearest commercial use cases in the whole release.
A Practical First Workflow
If you are trying Gemini Omni 1.2 Flash for the first time, here is the sequence that wastes the least money:
- Start in the Gemini app or Vids if you have access — no API billing, and you learn the model’s personality for free or at no additional cost.
- Draft at 360p when you move to the API: a 15-second test costs about $0.35, and you will see the composition, the motion and the obvious mistakes at that resolution.
- Fix with edits, not re-generations. Change the background, the lighting or one prop with a follow-up prompt instead of starting over.
- Upscale only the winner to 1080p or 4K once the 720p version is right.
- Save the prompt that worked. The reference images, the duration and the exact wording are your recipe — reuse them for the next clip in the series.
Conversational Editing in Gemini Omni 1.2 Flash
The feature that will change daily workflows most is not the extra five seconds. It is multi-turn conversational editing: the ability to refine the same clip turn by turn, in plain language, instead of starting over from scratch every time something is wrong.
Here is how it works in practice. You generate a clip — say, a 15-second product shot of two characters in medieval armour at a market fair. Then you keep talking to the model: swap the background for an oak forest, change the lighting to golden hour, switch the aspect ratio from landscape to vertical. Each edit applies to the same clip, and the model works to keep everything else intact — the characters’ performances, the reflections in the armour, the camera framing.
Google’s launch example runs exactly this sequence: starting from one generated clip, successive prompts move the scene from a striped pavilion tent to an oak forest, a stable with a white horse, a market fair and a stone castle, while the two characters stay recognisably themselves throughout.
That matters commercially because the old workflow was brutally wasteful. Every change meant a fresh generation, a fresh roll of the dice, and a fresh charge — and the new clip might fix the background while breaking the face. Conversational editing breaks that loop. The edits you will actually use, according to Google’s announcement, are:
- Swapping a product or background — the same shot, a different setting or a different item on the table.
- Changing lighting or weather — golden hour instead of noon, rain instead of sun, without re-shooting the scene.
- Switching aspect ratio — the same creative, reframed from 16:9 landscape to 9:16 vertical for social, without starting over.
- Producing ad variants — one hero clip, then turn-by-turn variations for different audiences and placements.
Multi-turn editing works across the full 15-second clip, which is the detail that makes it practical for real ads. A 15-second social spot is a complete unit of advertising; being able to edit it as a unit, rather than in fragments, is what turns the model from a clip generator into a production tool. The advice from the pricing section applies here too: each edit is a billed generation, so batch your changes into as few turns as you can — but an edit pass is still far cheaper than a full re-generation, and it preserves the parts of the clip that were already right.

Reference Following: Characters, Products and Style That Stay Put
The second pillar of this release is enhanced reference following, and it is the feature that decides whether the model is usable for branded work. You can guide a generation by combining several kinds of reference at once:
- Character images — photos of a person or character, so the same face appears in every shot.
- Product images — your actual product, so the packaging, typography and colours stay true.
- Style images — a reference look, so the lighting, palette and mood match your brand.
- Short motion clips up to three seconds — a movement reference the model can extend.
- Audio cues — voice references that shape how dialogue sounds. Note that audio-reference input remains allowlisted rather than generally open, so not every account will have it on day one.
Google says the model holds onto those details from start to finish: faces, garments, packaging typography and brand colours stay consistent across camera angles.
That claim, if it holds in practice, unlocks three kinds of work that used to be hard or impossible with AI video. Virtual try-on clips — Google’s own example combines one model photo with front-and-back product shots of three outfits to generate a 15-second sequence of 360-degree turns and match cuts, keeping the model’s identity and each garment’s cut, drape and back details consistent. Catalog spins and product showcases, where the product must look like the product, not like a plausible cousin of it. And branded video ads, where the logo and the pack shot are the whole point.
There is an honest limit to keep in mind. Reference following is a statistical achievement, not a guarantee: the model is better at holding a face or a label steady, not perfect at it.
For hero shots where a logo or a face must be pixel-exact, generate, inspect at full resolution, and regenerate the failures. The model’s job is to make the success rate high enough that this inspection loop is quick; your job is still to check. For more on how this compares with the reference systems in rival models, see our Kling 4.0 review, which covers Kling’s 15-reference Omni Reference system in depth.
Native Audio and Multi-Speaker Dialogue
Gemini Omni 1.2 Flash generates audio natively, synchronised with the video, at no separate charge on the pricing page — the audio is part of the video output, not a billed extra. Google claims more fluid, natural-sounding dialogue, clearer lip-sync and distinct multi-speaker conditioning. The launch example is a conservatory scene built from one scene image, four character and mouth references and one-second voice references, with the model routing each sampled voice to the right character, with expressive lip-sync and natural comedic timing across a 15-second animated scene.
For creators, the practical upshot is that a two-person dialogue scene is now a plausible single generation rather than a post-production assembly job. For advertisers, it means a spokesperson-style explainer with a real voice and matching lips, generated in one pass. As with the visuals, verify before you ship: lip-sync that is “clearer” is still AI lip-sync, and a client-facing video deserves a human watch-through at full length.
Gemini Omni 1.2 Flash vs Kling 4.0 vs HeyGen Video
No video model exists in a vacuum, and the honest way to judge this release is against the models US creators are actually choosing between right now. Here is how the three most relevant rivals compare, based on the manufacturers’ own announcements and our own coverage on this site.
| Gemini Omni 1.2 Flash | Kling 4.0 | HeyGen Video | |
|---|---|---|---|
| Max clip length | 15 seconds | 30 seconds | 15 seconds |
| Price signal | ~$0.07/sec at 720p (API) | Credit system; per-second rates vary by tier | Credit system via Creative Studio |
| Native audio | Yes — dialogue, lip-sync, multi-speaker | Yes — stereo audio, improved lip-sync | Yes — synchronised sound |
| Conversational editing | Multi-turn, full 15-second clips | Keyframes and references, not chat editing | Prompt-to-scene, API + Studio |
| Reference control | Character/product/style images, 3-sec motion clips, audio cues | Up to 15 multimodal references, 10 keyframes | Image or reference-based scenes |
| Free way to try | Google Vids (no additional cost); Gemini app tiers | Complimentary credits for new accounts | Free tier via HeyGen plans |
| Best for | Ads, product demos, Google-workflow teams | Longer narrative takes, film-style direction | Avatar-led scenes, presenter video |
| Launched | November 5, 2026 | October 15–16, 2026 | October 21, 2026 |
The comparison turns on one question: what are you making? If your unit of work is a 30-second narrative take — a short film beat, a dialogue scene with a beginning, middle and end — Kling 4.0’s length advantage is real and unmatched right now. Read our full Kling 4.0 review for the detail.
If your unit of work is a 15-second ad, product demo or social spot, Gemini Omni 1.2 Flash is the cheaper, better-integrated choice: about $1.04 per 720p clip on the API, conversational editing for variants, and native slots inside Flow, Vids and Google Ads where the work already lives. HeyGen Video, covered in our HeyGen Video guide, remains the strongest pick when the video is built around a presenter or avatar rather than a generated scene.
Price per second is not the whole story, and it would be dishonest to pretend otherwise. A model that costs half as much per second but needs three attempts to get a usable clip is more expensive than a model that costs twice as much and nails it first time. What Gemini Omni 1.2 Flash offers is a low headline rate plus the editing workflow that reduces the number of attempts — and that combination, rather than the $0.07 figure alone, is the real price story. For a broader view of the whole category, including Runway, Luma, Pika and Google Flow’s free credits, see our best AI video generator 2026 roundup.
Who Should Use Gemini Omni 1.2 Flash?
Gemini Omni 1.2 Flash is not the right model for everyone, and saying so plainly is part of an honest review. Here is who gets the most from it, and who should look elsewhere.
Performance marketers and small agencies get the biggest win. The 15-second format is the native length of a social ad, the $0.07-per-second rate makes test-and-learn affordable, and multi-turn editing turns one hero concept into a dozen placement variants without a re-shoot. If you run Meta, TikTok or YouTube Shorts ads, this is a tool built for your exact job.
E-commerce and DTC brands should look hard at the reference following. A product that keeps its packaging, typography and colours consistent across angles is the difference between a usable product video and a liability. Virtual try-on sequences and catalog spins are the two formats where this release is genuinely new, not just cheaper.
Developers and product teams building video into an app get a generally available API with enterprise backing, Provisioned Throughput and a clear price card — the unglamorous things that decide whether a model survives contact with a production system. The December 3 auto-routing of 1.1 requests to 1.2 also means existing integrations upgrade without a code change.
Google Workspace teams get the easiest on-ramp of all: 15-second generation inside Google Vids at no additional cost, and the same model inside Flow and the Gemini app. If your organisation already lives in Google’s stack, there is no procurement conversation to have.
Content creators and YouTubers making explainers, faceless-channel content or short-form video will find the 15-second ceiling generous for a single beat, and the native audio removes the separate voiceover step. But if your format needs a continuous 30-second take, Kling 4.0 is still the better fit — and if your format is a long-form documentary, no current AI video model replaces a camera.

Gemini Omni 1.2 Flash Limitations to Know Before You Pay
Every honest review names the rough edges, and this release has several that are easy to miss in the launch excitement.
- Fifteen seconds is still a cap. A 15-second clip is not a commercial, a tutorial or a scene. Anything longer is still an assembly job of multiple generations, with all the continuity work that implies. The model helps — consistency and editing are better — but it does not remove the ceiling.
- There is no free tier on the API. Every API second is billed from the start. Test in Vids or the Gemini app first, and draft at 360p on the API, before you commit real budget.
- The $0.07 rate is 720p only. At 1080p you pay about $0.104 per second, and at 4K about $0.209 — and both are upsampled, not natively generated at that resolution. Quote the price with the resolution attached, or your budget will be wrong.
- Regional restrictions apply. Video editing and avatar creation are unavailable in the EEA, Switzerland, the United Kingdom and some US states. Check availability for your location before promising a client a delivery date.
- Fast action and complex physics are still in development. Google says so itself: work continues on fast sports action, rapid script changes and complex physics. If your shot is a sprint, a fight or a collapsing structure, expect to inspect and regenerate.
- First-and-last-frame workflows top out at 10 seconds. If you direct with keyframes, the 15-second ceiling does not apply to you — plan for 3 to 10 seconds.
- Audio-reference input is allowlisted. Voice-reference features may not be available on your account yet, even though the model supports them.
- The model retires. Google lists a retirement date of November 5, 2027 or later for this version. That is a reasonable runway, but build your pipeline so a model swap is a configuration change, not a rewrite.
- Every clip is watermarked. Outputs carry an imperceptible SynthID watermark. That is good for trust and for identifying AI-generated media — and it means you should not present the output as camera footage.
- Google’s quality claims are Google’s. The consistency, lip-sync and camera-control improvements are company-reported, not independently benchmarked at launch. Treat them as a direction, verify them on your own footage, and keep a human in the review loop.

Gemini Omni 1.2 Flash FAQ
How much does Gemini Omni 1.2 Flash cost?
On the Gemini API, video output costs $12.00 per million tokens, which works out to about $0.07 per second at 720p, $0.023 per second at 360p, $0.104 at 1080p and $0.209 at 4K. A 15-second 720p clip costs about $1.04 in video output, plus small input and text-output charges.
Input costs $1.50 per million tokens and text output $9.00 per million. There is no free tier on the API. In Google Flow, a 15-second 720p clip costs 60 AI credits and requires a Google AI Plus, Pro or Ultra subscription.
How long can Gemini Omni 1.2 Flash videos be?
Clips run from 3 to 15 seconds, in one-second increments, for both generation and editing. First-and-last-frame workflows are limited to 3 to 10 seconds. In the Gemini app, 15-second generation is rolling out to Google AI Ultra subscribers; other supported users get clips of 3 to 10 seconds.
Is Gemini Omni 1.2 Flash free?
Not on the API — the free tier is listed as not available, so every API generation is billed. But you can use the model at no additional cost inside Google Vids with a Google account or Google Workspace subscription, and through the Gemini app and Flow under a Google AI subscription. “Free” depends on which door you walk through, not on the model itself.
What is multi-turn conversational editing in Gemini Omni 1.2 Flash?
It is the ability to refine the same clip with follow-up prompts in plain language — swap a background, change the lighting or weather, switch the aspect ratio — without starting over. The edits apply across the full 15-second clip, and the model tries to keep the characters, performances and camera framing intact from turn to turn.
How is Gemini Omni 1.2 Flash different from Gemini Omni 1.1 Flash?
Three headline differences: clips run up to 15 seconds instead of 10, the price is about 30% lower ($0.07 versus about $0.10 per second at 720p), and reference following, lip-sync and camera control are all improved. From December 3, 2026, API requests to the 1.1 model automatically route to 1.2 and are billed at 1.2 rates.
Can Gemini Omni 1.2 Flash make vertical videos for TikTok and Reels?
Yes. It supports 16:9 and 9:16 aspect ratios at 360p, 720p, 1080p and 4K, and conversational editing can switch a clip’s aspect ratio from landscape to vertical without re-generating from scratch. That makes it practical for producing the same creative in both formats.
Does Gemini Omni 1.2 Flash include audio?
Yes — audio is generated natively and synchronised with the video, including dialogue with clearer lip-sync and multi-speaker conditioning. The pricing page lists no separate charge for audio; it is part of the video output. Audio-reference input, where you supply a voice sample, remains allowlisted rather than generally open.
Where can I use Gemini Omni 1.2 Flash?
Through the Gemini API and Google AI Studio, Gemini for business via Google Cloud, the Gemini app (15 seconds rolling out to AI Ultra subscribers), Google Flow, Google Vids, Google Ads, and partner platforms including Runway and Magnific. Video editing and avatar creation are unavailable in the EEA, Switzerland, the United Kingdom and some US states.
Is Gemini Omni 1.2 Flash better than Kling 4.0?
It depends on the job. For 15-second ads, product demos and social spots, Gemini Omni 1.2 Flash is cheaper per second (about $1.04 per 15-second 720p clip on the API), offers conversational editing, and slots into Google’s apps. For longer narrative takes, Kling 4.0’s 30-second native clips and 10-keyframe direction system are stronger. Our Kling 4.0 review and best AI video generator 2026 roundup compare the full field.
Are Gemini Omni 1.2 Flash videos watermarked?
Yes. Outputs carry an imperceptible SynthID watermark, Google’s marker for AI-generated media. It is invisible to the eye but detectable by tools built to read it — including Google’s own SynthID detector site. Do not present generated clips as real camera footage.
The Verdict
Gemini Omni 1.2 Flash is the first version of Google’s video model that feels priced and packaged for production rather than for demos. Fifteen seconds is the right length for the work most people actually do — a social ad, a product beat, a short explainer — and $0.07 per second at 720p makes the experiment cheap enough to run properly: draft at 360p, edit in conversation, upscale only the winner.
The multi-turn editing is the feature that will save the most money in practice, because it attacks the real cost of AI video, which was never the per-second rate but the number of re-generations.
It is not the longest model — Kling 4.0 still owns the 30-second take — and it is not free on the API, so casual users should start in Vids or the Gemini app.
But for marketers, e-commerce brands, developers and Google Workspace teams, Gemini Omni 1.2 Flash is now the default sensible choice in its class: generally available, enterprise-backed, honestly priced, and good enough at holding a face, a product and a brand steady across 15 seconds that the results are usable without apology.
Try it the cheap way first. Draft a 15-second clip at 360p for about $0.35, or open Google Vids and generate one at no additional cost. If the face holds, the product reads and the dialogue lands, you have found your video model for the next year — and at roughly a dollar a clip, you can afford to find out.
Sources: Google’s launch announcement, “Gemini Omni 1.2 Flash: 15-second video and lower pricing” (November 5, 2026), on the Google blog; the official Gemini Developer API pricing page, checked November 11, 2026. Prices and availability are as published by Google and may change; check the pricing page before budgeting a large project. Performance and quality claims in this article are Google’s own unless stated otherwise.
