Gemini Omni Flash R2V

• 🆕 Reference-to-video variant in Google's new Gemini Omni Flash family
• 👁️ Multimodal video generation exposed through the Gemini API
• 🧠 Combines physics understanding with Gemini's world knowledge for storytelling
• 💬 Supports conversational, stateful editing that carries context across turns
• 🔧 Accepts text, image, audio and video inputs in one workflow
• ⚡ Positioned as a fast default for video generation
• 🔒 Generated media carries SynthID watermarking
• 🏢 Aimed at collapsing multi-tool production into plain-language instructions

Google

Google is an American multinational technology corporation and one of the world's most valuable brands. A subsidiary of parent company Alphabet Inc., Google operates across search, cloud computing, consumer electronics, and artificial intelligence. Its…

Gemini Omni Flash R2V is the reference-to-video configuration of Google's Gemini Omni Flash line, a fast, multimodal model for video generation and conversational editing exposed through the Gemini API. The reference-to-video mode conditions generation on supplied reference material, complementing the family's other entry points such as Gemini Omni Flash text-to-video and Gemini Omni Flash image-to-video.

The most notable shift versus Google's earlier video systems is statefulness. Where clip generators typically restart from a blank prompt, Omni Flash keeps video context across a conversation, so each turn builds on the prior result and applies incremental changes—adjusting lighting or swapping backgrounds—without re-describing the whole scene. It also treats text, image, audio and video as combinable inputs rather than a single text prompt.

Compared with the earlier Veo 3.1 Full Quality and Veo 3 Full Quality generations, which remain available for video work, Omni Flash emphasizes conversational, multi-turn editing and Gemini's world knowledge—historical, scientific, cultural and physical context—to move from photorealism toward narrative. Google notes generated media includes SynthID watermarking.

Because primary documentation for this specific reference-to-video variant is limited, capability and benchmark specifics beyond the family-level description above are not detailed here.