Gemini Omni 1.2 Flash: 15-second clips, conversational editing and a 30% lower price

Google DeepMind has introduced Gemini Omni 1.2 Flash, a new version of its video generation model that is now generally available in the Gemini API and in Gemini for enterprise (Google Cloud). The post, signed by product lead Anish Nangia and dated November 5, 2026, focuses on three ideas: longer clips, greater consistency with references and lower cost.

Pricing and workflow. Google says it has cut the price by 30% compared with Gemini Omni 1.1, to $0.07 per second of 720p video. Beyond the price cut, the company proposes a workflow designed to spend less: generate quick drafts in 360p or 720p, fix a detail with a follow-up edit instead of regenerating the whole clip and, finally, upscale the best pieces to 1080p or 4K.

Clips of up to 15 seconds. The model natively generates and edits video up to 15 seconds long, compared with 10 for Omni 1.1. Duration can be set in one-second increments, from 3 to 15. Google pitches this as enough room to tell complete stories, create product explainers or produce social ads in a single generation.

Better reference adherence. Generation can be guided by combining images of characters, products or style, short motion clips (up to three seconds) and audio prompts. According to Google, faces, garments, packaging typography and brand colors remain consistent across camera angles. This makes it possible to turn product photos or character sheets into virtual try-ons, catalog spins or brand ads. As an example, the company describes a 15-second virtual try-on sequence built from a photo of a model and front and back images of three outfits, with 360-degree spins and match cuts, preserving the model's identity and the cut, drape and back details of each garment.

Greater consistency. The model reduces visual drift and warping between the first and last frames, both when interpolating between a start and end frame and when extending a clip. Google compares 1.1 and 1.2 with examples such as an exhibition poster animated in layers and a camera moving from an exterior to an interior in a single continuous shot.

Multi-turn conversational editing. The same clip can be refined turn by turn: swapping a product or the background, changing the lighting or weather, or changing the aspect ratio, without starting from scratch. It works with full 15-second clips, making it easier to create ad variants for different audiences and formats without rewriting the brief. In one demo, the background changes successively (pavilion, oak forest, stable, fair, castle) while the characters' performances, the reflections on the armor and the framing are preserved.

Voice and dialogue. Google says the model delivers smoother, more natural dialogue, better lip sync and distinct conditioning for multiple speakers. In its demo, using a scene image plus one-second character, mouth and voice references, the model assigns each voice to the correct character in a 15-second animated scene.

Cinematic look and camera control. According to the company, it is possible to direct continuous camera moves, such as FPV flights, dramatic zoom-outs or orbits, with far fewer unwanted abrupt cuts, more stable scale and a more polished finish. The example cited is a continuous FPV flight across the Pantheon's piazza, through the bronze doors and up to the dome's oculus, generated from two reference photos.

Availability and rollout. For the first time, Gemini Omni is generally available, backed by Google Cloud's enterprise infrastructure and security. Provisioned Throughput is available and outputs include an imperceptible SynthID watermark. Developers can prototype in Google AI Studio and consult the documentation and pricing; enterprises can use it from the Google Cloud console. For the general public, users can direct scenes and edit over multiple turns in Flow, generate 15-second clips at no extra cost in Google Vids with a Google account or Workspace subscription, create 15-second videos in the Gemini app (rolling out to Google AI Ultra subscribers) and scale campaigns in Google Ads.

Acknowledged limitations. Google admits it is still working to improve fast sports action, rapid script changes and complex physics. All comparisons and improvements described come from the company itself; the post provides no independent metrics or external evaluations to back them up.