Omni 1.1 draws on 10 seconds of prior scene context per extension, up from 1 second, and added 4K upscaling, first, and last frame interpolation, and conversational editing. Google bills Gemini Omni 1.2 Flash as a faster, cheaper tier for longer clips via the Gemini API. The transformer handles text, vision, video, and audio natively and was trained on Google TPUs with a JAX and ML Pathways stack. #AI #Technology
Omni 1.1 draws on 10 seconds of prior scene context per extension, up from 1 second, and added 4K upscaling, first, and ...
