Google's Gemini Omni 1.2 Flash hits 15-second AI video at $0.07 per second

Google DeepMind has made its Gemini Omni 1.2 Flash video-generation model generally available, and the release doubles down on two things buyers actually ask for: longer clips and lower bills. The updated model generates clips up to 15 seconds long, supports step-by-step conversational editing, and costs 30 percent less than its predecessor.

That combination matters beyond Google's own product sheet. Video generation is turning into a volume business, where the cost per second and the ability to fix a shot without regenerating it decide who wins advertising, e-commerce and enterprise workloads. A 30 percent price cut pushes Gemini directly into that fight.

What's new in numbers?

The concrete changes are easy to list:

• Clip length: 3 to 15 seconds, up from a previous maximum of 10 seconds.
• Pricing: $0.07 per second for 720p video, a 30 percent reduction.
• Availability: Gemini API, Google AI Studio, Google Cloud Console, Flow, Google Vids, the Gemini app for Google AI Ultra subscribers, and Google Ads.
• Provenance: every generated video carries an imperceptible SynthID watermark indicating its origin.

The rollout touches both developer tools and consumer-facing surfaces, which signals Google wants this model in production pipelines, not just in demos. The SynthID watermark also keeps the release aligned with the growing policy expectation that AI-generated media stays traceable.

How does Conversational Editing work?

The feature Google calls "Conversational Editing" is the most structurally interesting part of the release. Users can tweak a generated clip step by step — changing the background, the lighting, or the aspect ratio — without regenerating the entire clip. Google says this works across the full 15-second length.

That's a real shift in how these systems are used. Older-generation tools force a binary choice: accept the output or pay to reroll the whole thing. Editing in place, through an ongoing conversation, turns video generation into something closer to a normal post-production workflow.

The editing feature pairs with a tiered workflow designed to cut costs. Drafts are first created at 360p or 720p, individual details are corrected with a follow-up edit, and only the final result is upscaled to 1080p or 4K. The logic is straightforward: don't pay 4K rates while you're still figuring out the composition.

The takeaway: Google is engineering the cost curve, not just the quality curve — cheaper drafts and targeted edits attack the biggest practical objection to AI video, which is wasted regeneration spend.

What does better reference adherence buy you?

Google also promises significantly better reference adherence in this version. The model accepts character images, product photos, short motion clips of up to three seconds, and audio cues as inputs. According to the announcement, faces, clothing, packaging typography, and brand colors stay consistent across different camera angles.

For anyone who has watched earlier video models slowly mutate a product label or a character's face mid-clip, this is the capability that decides commercial viability. Brands can't run an ad where the packaging changes color at second seven.

Google's own example makes the use case explicit. From a single model photo and product shots of three outfits, the model generates a 15-second virtual try-on sequence with 360-degree turns — while preserving the model's identity and each garment's cut and drape. That's a fashion e-commerce scenario, and Google is clearly aiming the demo at advertisers and retailers rather than at hobbyists.

Overall visual consistency is said to have improved as well. Google reports less visual drift or morphing across motion and camera transitions than with the predecessor model.

What about dialogue and camera work?

Audio and motion get their own upgrades in 1.2 Flash. Google says the model produces more fluid, natural-sounding dialogue with clearer lip-sync, and that it can now distinguish between multiple speakers.

On the visual side, the model handles complex, continuous camera movements — FPV drone flights, dramatic zoom-outs — with what Google describes as significantly fewer unwanted jump cuts. Both fixes address familiar failure modes: robotic speech timing and camera cuts that appear for no reason and break the illusion of a continuous shot.

Together these changes point at a specific target format: short-form advertising and product video, where 15 seconds, clean dialogue, stable branding and a single dramatic camera move are the standard brief.

Why the price cut matters more than the specs

The headline specs will draw the comparisons, but the 30 percent price drop to $0.07 per second at 720p is what changes purchasing decisions. At that rate, a 15-second 720p clip costs just over a dollar before the tiered workflow discounts come into play — the kind of arithmetic that makes marketing teams treat AI video as a line item rather than an experiment.

The tiered resolution workflow compounds the effect. Because drafts run at 360p or 720p and only the finished clip gets upscaled to 1080p or 4K, the expensive steps happen once, at the end, instead of on every iteration.

Distribution explains the rest of the strategy. By shipping the model simultaneously through the Gemini API, Google AI Studio, Google Cloud Console, Flow, Google Vids, the Gemini app for AI Ultra subscribers, and Google Ads, Google puts the same generator in front of developers, enterprise users, subscribers, and advertisers. Very few competitors can match that distribution breadth in a single release.

The open questions are the usual ones for vendor claims: Google's consistency, adherence, and dialogue improvements are the company's own characterization, and independent testing will show how they hold up outside curated demos.

What's clear is the direction. With 15-second clips, editable outputs, reference-driven consistency, and a price cut landing at the same time, Google is positioning Gemini Omni 1.2 Flash for high-volume commercial video — and betting that cost per second, not raw spectacle, will decide the next phase of the AI video market.