Do I have to specify duration and aspect ratio?

Yes. The GeminiVideoOmni node expects both in the prompt so the model knows the output canvas and time budget. Use clear formats like "duration: 5s" and "aspect ratio: 9:16" in the Prompt Constructor.

Does this workflow keep audio?

No. This setup focuses on visual edits, and the output from SaveVideo is typically silent. If you need audio, re-sync your original track or music in a video editor after export.

Does my video stay on my machine?

GeminiVideoOmni calls the Gemini API, so your media and instructions are sent to the service provider. Avoid sensitive content and be mindful of your API’s usage, privacy, and content policies.

How can I get more reliable results?

Keep clips short (a few seconds), be explicit and constrained in your instructions, always include duration and aspect ratio, and iterate. Use concrete guidance like "keep the subject centered," "subtle fog only," or "retain original motion" to reduce ambiguity.

Gemini Omni Flash: Video Edit

This ComfyUI tutorial workflow lets you edit a single input video using natural language instructions powered by the GeminiVideoOmni node. The graph is straightforward: LoadVideo ingests your source clip, the Prompt Constructor subgraph (node id 331103d2-d660-4a88-abf0-1117dc46b2df) builds a well-formed prompt with the required duration and aspect ratio, GeminiVideoOmni sends the video plus prompt to the Gemini Omni Flash API, and the results are displayed with PreviewAny and saved with SaveVideo. A MarkdownNote in the canvas provides quick tips and reminders.

Technically, LoadVideo decodes frames and metadata, then the Prompt Constructor concatenates your natural language edit with explicit settings like "duration: 5s" and "aspect ratio: 9:16". GeminiVideoOmni uses these to guide the model: the aspect ratio determines the output canvas, and the duration constrains the target clip length. The model returns an edited clip that adheres to your instructions (e.g., background changes, mood/atmosphere adjustments, reframing), which you can quickly review in PreviewAny and export via SaveVideo. For best results, keep prompts concrete, always include duration and aspect ratio, and iterate in short passes.

API

Use this workflow from code

Every Comfy workflow is a JSON graph. The payload below is this workflow, exactly as ComfyUI runs it — fetch it from the URL, keep it in version control, or load it in ComfyUI and run it node by node.

2620e5135442.json

Fetching workflow JSON…

Run it from TypeScript or Python with the Comfy SDK. The same code targets Comfy Cloud or a ComfyUI you host yourself — only the base URL changes.

// Install (beta)
npm i @comfyorg/sdk

// Run this workflow (TypeScript)
import { Comfy } from "@comfyorg/sdk";

const client = new Comfy({ apiKey: "comfyui-..." });
const wf = await client.workflows.fromFile("workflow_api.json");
const job = await client.run(wf);
await job.getOutputs("<output-node-id>")[0].toFile("output.png");

The SDK takes a workflow in API format: open this workflow in ComfyUI and use File → Export Workflow (API).

Comfy Cloud API access requires a plan with an API key.

SDK docs Get an API key

FAQ

Frequently Asked Questions