正文 · AI 翻译
IT Home reported on November 6 that Google has officially launched the Gemini Omni 1.2 Flash video generation model and made it generally available through the Gemini API and Gemini for business platform.
The new model offers native 15-second video generation, multi-turn conversational editing, and stronger adherence to references. It is also 30% cheaper than version 1.1, with 720p video generation priced at $0.07 per second (IT Home note: approximately 0.47 yuan at the current exchange rate).
To reduce costs at scale, the new workflow lets users quickly draft at 360p or 720p, refine details through subsequent edits without regenerating the entire clip, and then upgrade the finished video to 1080p or 4K, further reducing costs.
In terms of clip length, Gemini Omni 1.2 Flash extends native video generation from 10 seconds in version 1.1 to a maximum of 15 seconds. It also allows users to set a duration from 3 to 15 seconds in 1-second increments, enabling them to generate a complete story, product explainer, or social media ad clip in one go.
To guide generation, users can combine images of characters, products, or styles with motion clips of up to 3 seconds and audio cues. The model can maintain consistency in faces, clothing, packaging fonts, and brand colors across different camera angles throughout the video, making it easier to turn existing product images and character design images into virtual try-ons, rotating product showcases, and branded video ads.
For visual consistency, the new model can keep characters, objects, and scenes highly consistent from the first frame to the last, reducing visual drift and distortion during motion and camera transitions. Whether interpolating between the first and last keyframes or extending clips into longer sequences, it can keep products, characters, props, and environments consistent across shots.
For conversational editing, users can adjust the same clip over successive turns, such as replacing a product or background, changing the lighting or weather, or switching the aspect ratio, without starting generation from scratch. This multi-turn conversational editing capability covers the full 15-second clip, making it easier to create ad variants for different audiences and placements without rewriting prompts or regenerating the video.
For audio, the new model can generate smoother, more natural dialogue and clearer lip synchronization, and supports control over multiple speakers. It is suited to directing conversations between multiple characters in Flow, creating explainer videos in Gemini for business, and producing professional videos with digital presenters, music, and narration in Google Vids.
For cinematic visual control, users can direct complex, continuous camera movements, such as panoramic FPV fly-throughs, dramatic pullbacks, and orbiting shots, substantially reducing unnecessary jump cuts while maintaining a more stable zoom ratio and a more polished cinematic look.
Gemini Omni is now fully generally available for the first time. Built on Google Cloud's enterprise-grade infrastructure and security capabilities, it supports Provisioned Throughput, and generated content includes an imperceptible SynthID watermark.
Developers can prototype and debug in Google AI Studio and view documentation and pricing; enterprise users can try it in the Google Cloud console. General users can direct scenes and make multi-turn edits in Flow and generate 15-second clips for free in Google Vids. It is gradually rolling out in the Gemini App to Google AI Ultra subscribers, while Google Ads supports video ad delivery at scale.