Gemini Omni Flash 1.1 Reference Video generates video guided by reference assets. Add reference images for subject or style, and up to three short reference clips for motion.
In your prompt, use <IMAGE_REF_0> to refer to the first image and <VIDEO_REF_0> for the first video clip. These labels connect each instruction to the reference it describes. Reference clips must be at most 3 seconds each.
Resolution affects the cost per second. 360p costs the least, 720p is the default, and 4K costs roughly ten times as much per second as 360p. Video Duration ranges from 3 to 10 seconds.
Example: a camera pass through a university courtyard
Gemini Omni Flash 1.1 reference-to-video result generated from the image above and a tracking-shot prompt.
Pick a model and start generating
These models are tuned for detail — open one and put this guide to work.
