Artificial intelligence video generation and editing has received a major update. Google DeepMind has officially released a brand-new multimodal video generation model—Gemini Omni1.2Flash. The model brings comprehensive major upgrades across generation duration, visual consistency, multi-turn conversational editing, and API pricing.
Core Performance Upgrades: Doubled Duration, Improved Consistency, and Multi-Turn Conversational Editing
In terms of core features, Gemini Omni1.2Flash has achieved several technical breakthroughs:
Increased generation and editing duration: The upper limit for native video generation and edited clips has been officially extended from 10 seconds in the previous generation to 15 seconds, and users can specify any duration between 3 and 15 seconds.
Image quality and draft strategy: The model supports fast generation at 360p or 720p draft quality first, then seamless upscaling to 1080p or 4K HD resolution once confirmed.
Scene consistency and lip sync: Compared with previous versions, the new model performs better in the visual consistency of characters, objects, and scenes over time, effectively reducing flaws such as jump cuts, while greatly improving lip-sync accuracy for multiple speakers and the coherence of camera movement.
Conversational progressive editing: It introduces powerful multi-turn conversational editing. Users no longer need to regenerate from scratch each time; instead, they can precisely replace and adjust the subject, background, or lighting of an existing video round by round directly in the conversation.
Comprehensive Multi-Platform Ecosystem Coverage and Highly Competitive Pricing Strategy
On launch day, Gemini Omni1.2Flash achieved large-scale rollout across the entire ecosystem:
Official platform matrix: The model is simultaneously available on Gemini API, Google Cloud, Flow, Google Vids, Google Ads, and Google AI Studio. In the Gemini App on mobile and web, the model is available to all Google AI subscribers, with the ability to generate clips of up to 15 seconds exclusive to Ultra subscribers; Google also confirmed that generating 15-second videos in Google Vids will incur no additional fees.
Third-party platforms launch simultaneously: Several mainstream third-party platforms, including fal, OpenRouter, Krea, and Higgsfield, also launched the model the same day. Among them, OpenRouter prices by resolution tier: $0.023 per second for 360p, $0.070 per second for 720p, $0.104 per second for 1080p, and $0.209 per second for 4K, and promises to keep prices consistent with direct official access.
Significant price cut: Overall, the API price for 720p video has dropped by about 30% compared with the previous generation Omni1.1, adjusted to $0.07 per second, which works out to about $1.05 to produce a 15-second 720p video.
Authoritative Benchmark Debut: Tops Both Text-to-Video and Image-to-Video Leaderboards
In the evaluation results subsequently released by the authoritative third-party blind testing platform Arena, Gemini Omni1.2Flash stunned everyone in its debut, taking first place on both the Text-to-Video and Image-to-Video leaderboards.
Among them, its score in the text-to-video category reached 1541 points, 27 points higher than the previous generation Gemini Omni1.1Flash. With the model’s official debut and full rollout, Google’s Omni series has successfully swept the top three spots on the text-to-video leaderboard, further consolidating its leading position in the field of multimodal generation.
