Natural-language direction
Describe the scene, action, camera, style, lighting and requested transformation in one prompt.
Gemini Omni multimodal video
Gemini Omni AI Video Generator
Create a video from a prompt and visual references, or transform an existing clip with natural-language direction. Choose the format and follow the asynchronous result online.
Output
Your video will appear here
Write a prompt, add optional reference media and submit a Gemini Omni task.
The model, explained
What is Gemini Omni?
Gemini Omni is a multimodal AI video generator that can interpret a written prompt together with reference images and an optional source video.
Use the same workspace to create a new clip from text and images or transform existing footage while describing the elements that should change and remain consistent.
The page sends the request to Kie as an asynchronous task, follows its progress, and privately stores the completed video for authenticated preview and download.
Multimodal video direction
Create and transform video with Gemini Omni
Combine natural-language direction with visual context, then choose the output settings that fit your delivery channel.
Mixed visual references
Add images for visual guidance and optionally one source video for transformation or editing.
Up to 4K output
Choose horizontal or vertical framing, 720p, 1080p or 4K output, and a supported duration when no source video is present.
Private result tracking
Monitor the Kie task and access the finished video through private preview and download links.
Three focused steps
How to create a Gemini Omni video
Give the model a clear instruction, add only the visual context it needs, and review the completed result.
- 01
Write the creative direction
Define the subject, action, visual treatment and camera behavior, including what must stay consistent during an edit.
- 02
Add media and settings
Upload reference images or an optional source video, then choose aspect ratio, resolution and duration when available.
- 03
Generate and download
Submit the multimodal task, follow its status, and preview or download the privately stored output.
Gemini Omni prompting tips
Use visual references selectively and make every requested transformation explicit.
- Describe the scene, subject action and camera movement before adding style and atmosphere.
- Choose references with a clear purpose; the page allows seven media units in total and a source video uses two.
- Pick 16:9 for landscape delivery or 9:16 for vertical video, then select the resolution your final use needs.
- When editing video, list what should be preserved separately from what should change.
Questions, answered
Gemini Omni FAQ
What can Gemini Omni create on this page?
It can create a video from a prompt and reference images, or transform one uploaded video with natural-language direction.
How many reference files can I add?
The page supports up to seven media units. Each image uses one unit and an optional source video uses two, leaving room for up to five images beside that video.
Can I choose the video duration?
Without a source video, you can choose 4, 6, 8 or 10 seconds. When a source video is included, Gemini Omni determines the output duration automatically.
How is the finished video delivered?
FramePack stores the completed result privately and exposes authenticated preview and download links after the Kie task succeeds.
Turn multimodal direction into video
Combine a prompt with focused visual references and create your next clip with Gemini Omni.
Create with Gemini Omni