Google launches Gemini Omni Flash "world model"

Omni Flash signals Google's pivot from simple media generators to "world models" that understand physical forces and context.

• –Native any-to-any architecture enables seamless generation across text, image, audio, and video modalities without discrete sub-models
• –Conversational video editing allows for iterative, natural language adjustments to camera angles, lighting, and characters
• –Model is grounded in physical laws like gravity and kinetics, significantly reducing common "AI hallucinations" in motion
• –Integration with Google Flow and YouTube Shorts makes high-end video production accessible to millions of mobile creators
• –Mandatory SynthID watermarking by default addresses the growing need for provenance in a "video-first" AI era