Gemini Omni 1.1 Flash is Google’s multimodal model for AI video generation and editing. Direct your scenes with up to 10 seconds of video context, first and last frame control, 360p drafts, video references and output up to 4K.
Bring prompts, character images and video references into one creative direction. Animate a still image, plan an opening and ending, or build on an existing scene by refining movement, lighting and camera language.
Provide an existing clip and describe what should happen next, giving characters, environments and story direction a continuous visual reference. The model supports up to 10 seconds of scene-extension context; video input lets the model determine output duration.
Use a first frame to establish the subject, product and setting, then a last frame to define the destination. Combine them with motion and camera instructions for product transitions, character entrances and loops.
Images can guide appearance, clothing, setting and style, while video can guide motion and camera rhythm. Describe each reference’s role so Gemini Omni can distinguish details to preserve from actions to reinterpret.
Explore prompts, movement and framing in 360p, then choose 720p, 1080p or 4K for your next generation. Drafts help compare directions; higher quality helps inspect subjects, materials and settings. Regeneration can change visual details.
| Duration | 4 / 6 / 8 / 10s |
|---|---|
| Resolution | 360p / 720p / 1080p / 4K |
| Prompt limit | 20,000 characters |
| Reference images | 7 |
Describe the subject, action, setting and camera movement. Explain what each reference image or video should guide.
Start with text, first and last frames, or multimodal references. Choose the aspect ratio and available quality, then review the credit estimate.
Open Results to follow progress and view your video. Refine the prompt based on motion, framing and detail before generating another version.
Start from a product image and add viewing angles, lighting changes and camera movement for product clips or vertical social content.
Give a character their next action and connect portrait references with existing shots to develop a story clip by clip.
Explore pushes, pulls, tracking or orbiting moves in one scene, using short clips to compare atmosphere, composition and pacing.
Say who does what and where, then add camera movement and lighting. Avoid packing unrelated actions into a short clip.
Specify the appearance, clothing or product materials to preserve, alongside the movement, camera or setting to change.
Choose landscape or portrait first, followed by duration and quality. With a video reference, the model determines output duration.
Explore creation modes, output settings and credit usage.
It is Google’s multimodal model for video generation and editing, using prompts, images and video references to guide scenes, subjects and motion, with frame control and output up to 4K.
Yes. Leave the image slots empty in frame mode, describe your shot, and choose duration, aspect ratio and quality.
Use the modes separately. Choose frame control for a defined opening and ending; a last frame requires a first frame. Choose multimodal references to combine image and video information.
Use 360p to explore composition and movement, and 720p, 1080p or 4K for different detail needs. This selects generation quality; it is not a lossless one-click conversion of a draft to 4K.
Without a video reference, choose 4, 6, 8 or 10 seconds. With video input, the model determines output duration and the manual duration selection does not apply.
Credits vary with model, quality, duration, references and task count. Review the estimate in the creation panel before submitting.
Nano Banana 2.5