Gemini Omni Flash: Conversational Video Generation

Gemini Omni Flash is Google DeepMind’s conversational video generation and editing model, part of the Gemini Omni family first introduced at Google I/O 2026. It combines Gemini’s multimodal reasoning with native video creation: you describe what you want in plain language, attach reference images or videos, and the model generates a clip with synchronized audio. The current version, Gemini Omni Flash 1.1, reached general availability on August 27, 2026 and adds keyframe interpolation, 360p/4K output options, and scene extension.

​

What Gemini Omni Flash 1.1 is good at

• Conversational video editing: Refine and edit videos using natural language: swap characters, relight scenes, alter angles, add or remove objects while maintaining the original audio and video tracks
• Multimodal input: Combine text, images (up to 14), and videos (up to 3, 10 seconds each) to guide generation. Every output carries a native audio track
• Keyframe interpolation: With the image_to_video task, attach a starting frame and an optional ending frame, and the model generates the footage in between
• Reference to video: Bind reference images to roles with tags like <IMAGE_REF_0> so characters, products, and objects from your images appear in a new scene without the image itself being used as a frame
• Scene extension: The extend task appends up to 10 seconds of new footage to a clip, reading the last 10 seconds of context for character and motion consistency, up to about 40 seconds cumulatively
• Resolution control: Draft at 360p (about one third of the 720p cost), then re-render at 720p, 1080p, or 4K. 16:9 and 9:16 aspect ratios are supported
• World knowledge and simulation: Combines physics understanding with Gemini’s knowledge of history, science, and cultural context, enabling meaningful storytelling beyond photorealism
• Text and action synchronization: Render legible text and graphics directly into video, syncing kinetic typography with on-screen movements

​

Available workflows

​

Text to Video (Omni Flash 1.1)

Generate cinematic video from natural language prompts with Gemini Omni Flash 1.1. Describe the desired length (3 to 10 seconds) directly in the prompt, and pick the aspect ratio and output resolution in the node: 16:9 or 9:16, 360p for cheap drafts, up to 4K for final renders. Every clip includes a generated audio track.

​

Image to Video (Omni Flash 1.1)

Animate an image with Gemini Omni Flash 1.1. With the image_to_video task, the first attached image becomes the starting frame and an optional second image becomes the ending frame: the model generates the footage in between, which makes camera orbits, zoom transitions, and looping clips predictable.

​

Reference to Video (Omni Flash 1.1)

Generate video that incorporates specific subjects from up to 14 reference images. In reference mode, bind each image to a role with tags like <IMAGE_REF_0> and refer to the tags in your prompt: characters, products, and objects from your images appear in the scene while the image itself is never used as a frame. Combine character references with style references for brand-consistent content.

​

Video Edit (Omni Flash 1.1)

Edit videos with natural language using Gemini Omni Flash 1.1. With the edit task, the node takes exactly one input video (10 seconds or less) and rewrites it based on your instructions: swap backgrounds, restyle scenes, add or remove elements. The edit and extend tasks keep the aspect ratio of the input video. Simple prompts work best; adding “keep everything else the same” maximizes consistency.

​

Video Extend (Omni Flash 1.1)

Extend an existing video by up to 10 seconds per step with the extend task, building stories up to about 40 seconds total. The model analyzes the last 10 seconds of context, keeping characters, motion, and audio coherent as the scene continues from where it left off. Optionally attach reference images to introduce new characters mid-story. Extension appends new content only at the end of the clip, but the model may revise the final source frames to make the transition seamless.

​

Get started

1. Update ComfyUI to the latest version (0.34.2 or later for the Omni Flash 1.1 workflows)
2. Double-click the canvas and search for “Gemini Omni Flash” nodes
3. Or go to the Template Library to use the ready-to-go workflows
4. Choose the workflow that matches your input type (text, image, or video)
5. Enter your prompt and generate