Gemini Omni Flash
Precisely edit your input video with a simple prompt.
About
This workflow lets you precisely edit an input video with a single, well-structured prompt using Google’s Gemini Omni Video through the GeminiVideoOmni node. It starts by loading your source clip via LoadVideo, then routes the frames and your instructions to GeminiVideoOmni, which performs the video-to-video transformation. A small, focused subgraph called Prompt Constructor (subgraph id: 331103d2-d660-4a88-abf0-1117dc46b2df) helps you assemble a clean, model-friendly prompt that includes the two required parameters for Gemini Omni Video: duration and aspect ratio.
What you provide
- Reading the workflow
Runs in ComfyUI. This workflow can't run on this page. Open it in Comfy Cloud with everything loaded, or download it for your own ComfyUI.
Try in CloudQuestions & answers
The Gemini Omni Video model expects both fields to resolve temporal and spatial constraints. Providing them helps prevent timing drift, cropping, or letterboxing and improves overall consistency. Use the Prompt Constructor subgraph to add them correctly.
You may see truncation, repetition, or timing artifacts. For best results, set the prompt duration to the exact clip length (in seconds) from LoadVideo so the edit aligns frame-for-frame.
Match the prompt’s aspect ratio to the source video’s aspect ratio. If the input is 1920x1080, use aspect ratio: 16:9. Mismatches can introduce letterboxing or unintended crops.
This workflow focuses on visual edits. If you need audio, export the edited video and remux it with the original audio track in a video editor or command-line tool.