Google has recently officially released a new AI video generation model, Gemini Omni1.2Flash, marking a key step forward in its push to bring AI video into large-scale commercial production. As an upgraded iteration of the previous generation, the model has been comprehensively optimized in generation duration, scene consistency, multi-turn conversational editing, and API pricing.

In terms of core technology and features, Gemini Omni1.2Flash increases the maximum duration of a single video generation from 10 seconds to 15 seconds, effectively reducing long-form video creation's reliance on multi-step stitching. At the same time, thanks to underlying efficiency optimizations, the cost of generating 720p video has dropped significantly by 30%, from $0.10 per second in the previous generation to $0.07 per second. The API cost of generating a single 15-second 720p video has fallen to $1.05.

QQ20261106-143004.jpg

In addition, the new model performs better at preserving products, people, logos, and text across multiple viewpoints, and supports multi-turn conversational editing, allowing users to locally adjust lighting, background, or camera angle without regenerating the entire video. The model also improves natural speech processing, lip sync, and multi-speaker interaction, while retaining fast 360p draft previews and output up to 1080p or 4K.

Currently, Gemini Omni1.2Flash is officially available to developers and creators through Google AI Studio, Google Cloud, Google Flow, and Google Vids. This update not only greatly lowers the commercial barrier for large video models, but also accelerates the evolution of AI video generation technology toward a refined, highly coherent, industrial-grade production stage.