Image-to-Video
Gemini Omni Image-to-Video API
Add public image URLs to a Gemini Omni request and use the same asynchronous task flow as text-to-video.
Request shape
{
"model": "gemini-omni/video",
"input": {
"prompt": "Camera slowly pushes in while studio light sweeps across the product",
"image_urls": ["https://example.com/product.jpg"],
"mode": "pro",
"aspect_ratio": "1:1",
"duration": "5",
"sound": false
}
}Best practices
Use public image URLs
The API needs to fetch the image from the generation backend, so private local files are not suitable.
Guide the motion
Prompts such as camera push-in, slow pan, subject turns, or background motion help control the output.
Match aspect ratio
Choose 1:1, 16:9, or 9:16 based on your target placement and source image framing.
Prefer concise tasks
Short, concrete motion instructions work better than multi-scene prompts for image-guided tasks.