AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Google's Gemini Omni 1.2 Flash stretches AI video clips to 15 seconds and cuts pricing by 30 percent

Collected Nov 5, 2026

Google DeepMind has made Gemini Omni 1.2 Flash generally available. The video-generation model produces clips of three to 15 seconds, up from a previous 10-second cap, and is priced at $0.07 per second for 720p output, 30 percent less than its predecessor. It is rolling out through the Gemini API, Google AI Studio, the Google Cloud Console, Flow, Google Vids, the Gemini app for Google AI Ultra subscribers, and Google Ads.

The headline capability is what Google calls Conversational Editing. Instead of re-rendering a whole clip to change a background, the lighting or the aspect ratio, users issue follow-up instructions in an ongoing dialogue and the model adjusts only the requested elements. Google says this holds across the full 15-second length. This is a meaningful architectural shift from the usual text-to-video pattern, where each prompt yields a fresh generation and small corrections mean starting over. Treating a clip as a persistent object that can be revised in place is closer to how image editors with inpainting and region-based prompts already work.

Google pairs that with a tiered workflow aimed at cost control: draft at 360p or 720p, fix individual details with an edit pass, then upscale only the finished output to 1080p or 4K. Because edits apply at the draft stage, teams avoid paying higher-resolution rates for iterations that get thrown away, which the source describes as a further saving beyond the sticker price.

Reference adherence is the other major claim. The model accepts character images, product photos, motion clips of up to three seconds and audio cues as inputs, and Google says faces, clothing, packaging typography and brand colors stay consistent across different camera angles. Its demonstration is a virtual try-on: from one model photo and product shots of three outfits, it renders a 15-second sequence with 360-degree turns while keeping the model's identity and each garment's cut and drape. For product and commerce teams, holding brand typography and color constant across shots is often the difference between a usable asset and a reshoot, so this is the claim worth testing first. Consistency across angles has been a persistent weakness of diffusion-based video models, which tend to drift or morph as the camera moves.

Google says overall visual consistency improved, with less drift and morphing across motion and camera transitions than the predecessor. It also claims more natural dialogue with clearer lip-sync and the ability to distinguish between multiple speakers, and improved handling of continuous camera moves such as FPV drone flights or dramatic zoom-outs, with notably fewer unwanted jump cuts. Those are self-reported results; independent comparisons would be needed to judge the size of the gain.

Every generated video carries a SynthID watermark, an imperceptible signal Google embeds to mark AI-generated content. It is a provenance measure rather than a hard restriction, so teams building pipelines should assume outputs are detectable as synthetic.

In practice, the 15-second ceiling and conversational editing point at a specific workflow: short-form social and ad spots, product demos, try-on sequences and iterative concepting where a creative revises a shot rather than re-rolling the dice. The per-second price of $0.07 makes quick math possible, but the tiered draft-first approach matters as much as the headline figure, since a 15-second 1080p or 4K final is far cheaper when only the last step pays for high resolution. A likely trade-off is that conversational edits may still introduce small inconsistencies the model does not fully control, and that repeated edit passes could accumulate artifacts or cost. The published consistency claims are Google's own and deserve in-house testing against real brand assets before teams commit.

Why it matters: Teams already generating video through Gemini, Flow, Vids or Google Ads get longer clips and cheaper output without changing platforms, and API builders can now build correction loops around a single clip. Product and creative teams weighing vendors may find the draft-edit-upscale pattern changes how they budget generation, though the strongest consistency claims are still Google's to prove.

Read at The Decoder

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Google Deepmind is making Gemini Omni 1.2 Flash generally available. The video-generation model now creates clips up to 15 seconds long, can be edited step by step via conversation, and costs just $0.07 per second at 720p, 30 percent less than its predecessor. Users who draft in low resolution and upscale only the final result can save even more. The article Google's Gemini Omni 1.2 Flash stretches AI video clips to 15 seconds and cuts pricing by 30 percent appeared first on The Decoder .