New on Runway
Gemini Omni Flash
Google's newest video model, on Runway. Generate 3–10 second clips from a prompt, a first frame and up to 10 reference images — or edit existing footage with plain-language instructions.
How to use Gemini Omni Flash on Runway
Choose Frame or Edit mode
Frame mode generates a new video from a prompt and optional images. Edit mode changes existing footage — upload a clip and describe the adjustment you want.
Add your prompt and references
Set the duration between 3 and 10 seconds, optionally pin a first frame, and add up to 10 reference images to guide the look and feel.
Generate and share
Gemini Omni Flash picks 16:9 or 9:16 automatically from your inputs — or ask for a format directly in your prompt — and returns a clip ready to post.
What you can do with Gemini Omni Flash
Two modes in one model
Generate brand-new clips from text and images, or upload existing footage and edit it with a plain-language instruction.
Reference-guided looks
A first frame becomes the literal opening of your video, while reference images guide style and feel without appearing verbatim — combine them however you like.
Prompt-controlled framing
The model chooses 16:9 or 9:16 automatically based on your inputs; steer it by asking for a widescreen or vertical clip in the prompt.
Be first to create with Gemini Omni Flash on Runway.
Try nowMore models on Runway
Runway gives you access to the world's leading video, image and audio models — Runway's own frontier models plus the best from Google, OpenAI, Kling, ByteDance and more — all in one workspace.
Veo 3.1
Google's cinematic video generation with native audio and speech.
Explore the model
Kling O3
Multi-modal video with audio, keyframes and up to 4K output.
Explore the model
Seedance 2.0
Multi-modal video generation with sound from text, images, video or audio.
Explore the model
Nano Banana Pro 2
Google's latest image model — the best balance of quality, speed and intelligence.
Explore the model
Try Gemini Omni Flash on Runway today.
Google's newest video model, alongside every other leading AI model — in one workspace.
Try now
