Google DeepMind 全量开放视频模型 Gemini Omni 1.2 Flash，片段最长 15 秒并降价 30%

Google Deepmind is making its Gemini Omni 1.2 Flash video-generation model generally available. The updated model creates clips up to 15 seconds long, supports conversational editing, and costs 30 percent less than its predecessor.

Google says Gemini Omni 1.2 Flash 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. Every generated video carries an imperceptible SynthID watermark to indicate its origin.

Clips can now run between three and 15 seconds, up from a maximum of 10 seconds previously, and pricing drops 30 percent to $0.07 per second for 720p video. Generated clips can also be edited through an ongoing conversation.

A standout feature is what Google calls "Conversational Editing." Users can tweak a generated clip step by step, changing the background, lighting, or aspect ratio without regenerating the entire clip. Google says this works across the full 15-second length. Paired with a tiered workflow, the feature is supposed to cut costs. Drafts are first created at 360p or 720p, individual details are corrected with a follow-up edit, and only the final result is upscaled to 1080p or 4K.

Google promises more consistent results with reference images

Google also promises significantly better reference adherence. According to the announcement, the model can take character images, product photos, short motion clips of up to three seconds, and audio cues as inputs. Faces, clothing, packaging typography, and brand colors are supposed to stay consistent across different camera angles.

As an example, Google shows a virtual try-on clip. From a single model photo and product shots of three outfits, the model generates a 15-second sequence with 360-degree turns while preserving the model's identity and each garment's cut and drape.

Overall visual consistency is said to have improved as well. Google says there is less visual drift or morphing across motion and camera transitions than with the predecessor.

Beyond that, Gemini Omni 1.2 Flash is supposed to produce more fluid, natural-sounding dialogue with clearer lip-sync and the ability to distinguish between multiple speakers. The model further handles complex, continuous camera movements such as FPV drone flights or dramatic zoom-outs, with significantly fewer unwanted jump cuts.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.