Gemini Omni 1.2 Flash: 15-Second Clips, 30% Price Cut

Google DeepMind announced general availability of Gemini Omni 1.2 Flash on November 5, 2026. The release extends native clip length to 15 seconds at 720p and drops the per-second price by 30%, to $0.07 from the Omni 1.1 rate, according to Google DeepMind. Those two figures change the calculus for video-generation API pipelines: longer atomic outputs reduce stitching work, and the lower unit cost shifts break-even throughput for production use.

It is rolling out across the Gemini API and Gemini for business, with entry points in Google AI Studio, Google Cloud console, Flow, Google Vids, the Gemini app, and Google Ads.

Clip length, duration control, and reference adherence

Google DeepMind reports that Omni 1.2 Flash generates and edits clips natively up to 15 seconds, up from 10 seconds in Omni 1.1. Duration can be specified in one-second increments from 3 to 15 seconds, so a 6-second bumper and a 15-second pre-roll are both reachable without over-generating and trimming.

According to Google DeepMind, reference conditioning now accepts character, product, and style images, plus motion clips up to three seconds and audio cues, in combination. The maker says faces, garment details, packaging typography, and brand colors stay consistent across camera angles within a single generation—aimed at virtual try-on clips, catalog spins, and branded video ads built from product photography.

Multi-turn conversational editing

The interaction model is a meaningful architectural shift. Instead of re-generating a full clip after each revision, Omni 1.2 Flash lets a developer or creative refine the same clip with successive natural-language prompts: swapping backgrounds, changing lighting or weather, or switching aspect ratio while keeping character performance, camera framing, and material reflections intact. Google DeepMind states this works across full 15-second clips.

For production pipelines, that changes the cost profile of ad-variant generation. If background swaps and placement-ratio changes can happen conversationally without full re-generation, the effective cost per delivered variant drops below the headline per-second rate. The workflow Google describes—draft at 360p or 720p, refine with follow-up edits, then selectively upscale to 1080p or 4K—is a staged cost-control pattern worth making explicit in high-volume work. This kind of pipeline architecture decision often matters more than raw model capability for throughput and spend.

Pricing and availability

Google DeepMind reports the $0.07/second figure directly. The model is now generally available on the Gemini API and Google Cloud, with Provisioned Throughput as an option. Outputs carry an imperceptible SynthID watermark. Access includes Google AI Studio for prototyping, the Google Cloud console for enterprise deployment, Flow for scene direction, Google Vids with 15-second clips at no additional cost for a Google account or Google Workspace subscription, the Gemini app for Google AI Ultra subscribers, and Google Ads for campaign scaling.

AI Mastery analysis

The 30% price reduction Google DeepMind reports is notable because it arrives alongside capability gains—longer clips, multi-turn editing, and stronger reference adherence—rather than as a standalone commodity cut. That pattern, where token and compute efficiency gains fund simultaneous feature expansion, has recurred across the frontier model tier and suggests the underlying generation cost curve is still falling faster than list prices.

The more consequential detail is multi-turn editing. Stateless re-generation was the primary cost amplifier in earlier video API workflows: every refinement restarted the full generation clock. If Omni 1.2 Flash's conversational editing genuinely avoids full re-generation on background swaps and aspect-ratio changes, video API pricing shifts from a flat per-second rate toward session cost with marginal edit pricing. That favors iteration-heavy creative workflows over batch generation without refinement.

The source acknowledges direct limitations: fast sports action, rapid script changes, and complex physics are still under development. Teams building in those categories should evaluate accordingly rather than extrapolate from the branded-content and dialogue-heavy demos.

The broader signal is that general availability, backed by Google Cloud's enterprise-grade infrastructure and security plus Provisioned Throughput, removes the beta-risk argument for holding video generation at the prototype stage. At Google's stated $0.07/second with 15-second clips and multi-turn editing, the remaining question for engineering teams is integration architecture, not model readiness.

Sources

• Primary: Google DeepMind: Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips