Video model guide

Gemini Omni 1.1 Flash video generation from a description, frames or references

Create video from text, first and optional last frames, or reference images and a video. Choose 4, 6, 8 or 10 seconds without a reference video; otherwise output duration is automatic.

Open text to video Open image to video All model guides

Controls and specifications

Text mode needs only a description. In the image workspace, choose frame mode with one or two images, or reference mode with up to seven images. Adding one reference video reduces the image limit to five.

The workspace uses up to the first ten seconds of a reference video, whose source can be up to 30 seconds. Output duration is automatic and is not guaranteed to match that input segment.

Clip length: 4–10 s · 4–10s without a reference video; automatic with one
Quality tiers: 360p · 720p · 1080p · 4k
Audio: No audio on/off control
Shape options in text mode: 16:9 · 9:16
Shape options with frames or references: 16:9 · 9:16
Images with one reference video (uses two slots): 5
Reference images: 7
Reference videos: 1
Prompt limit: 20000 characters

Choose a mode, resolution and 16:9 or 9:16 shape; an optional seed is available. A description of up to 20000 characters is required. Select 4, 6, 8 or 10 seconds only when there is no reference video.

What this model is good at

A first frame, and where it ends

Upload one picture to set the opening frame, or two to fix the closing frame as well. The shape selector stays available in this mode, so the clip keeps the proportions you chose.

Several references and one clip

Combine up to seven pictures, such as a product, a setting and a character, and describe how they belong in one shot. A single reference clip can join them and takes two of those seven units.

Four lengths and four tiers

Pick 4, 6, 8 or 10 seconds and 360p, 720p, 1080p or 4k, then set an optional seed when you want to repeat a run with the same settings. The length applies whenever the task carries no reference clip.

Typical projects

Start from a picture you already have

Open a story or an ad on a finished key visual, keep 16:9 for the feed or 9:16 for vertical placements, and let the clip play out from that exact frame.

One subject in a new setting

Supply pictures of the same product or character together with a location and describe the scene, so a new clip still shows the subject your audience knows.

Move like a clip you own

Use an owned video as a visual reference, optionally with up to five images, and describe the intended action. The model determines output duration automatically.

Limits to know before you start

Frame mode and reference mode cannot be combined. Moving between the two tabs clears the pictures or the clip that the other mode cannot accept, and a last frame only works together with a first frame.
The workspace uses up to the first ten seconds of a reference video, whose source can be up to 30 seconds. Output duration is automatic and is not guaranteed to match that input segment.
Pictures and the clip share one quota of seven units, and the clip counts as two. Without a clip you can use seven pictures; with one, five.
Images must be JPEG, PNG or WebP up to 20 MB. The clip must be MP4 or MOV, up to 100 MB and 30 seconds of source material, and only one clip fits in a task.
There is no sound on or off switch. The finished task we checked carried an audio track next to the video, and we have not reviewed how that track sounds, so judge it from your own result.
Voice or character material cannot be uploaded for this model here; the workspace accepts only the inputs listed on this page.
Quality tier names are settings, not measured sizes. Our paid check covered one 360p, 4-second, 16:9 task with a first frame; the other tier and clip combinations come from the published settings, not from measurements.

How to start

Open the workspace

Use a button on this page. It opens the right workspace with this model already selected, in the language you are reading now.

Describe the result

Write what should happen in the shot or image. Concrete subjects, framing and action give the model more to work with than short labels.

Set the controls and confirm

Choose length or quality tier, shape and sound where offered, then read the credit quote for those exact settings before you generate.

What it costs

Generation uses account credits. Cost depends on the model, quality and output size or duration. Reference images, video or audio may add charges in some modes. Check the workspace quote for your selected inputs and settings before generating; the pricing page lists credit packages.

See credit packages

Questions about this model

How many pictures can I use?

Seven per task in reference mode. If you also add a reference clip, the clip takes two of those seven units, so the ceiling drops to five pictures.

Can I combine frames with reference pictures or a clip?

No. First and last frame is its own mode with one or two pictures and nothing else. Switching modes clears the material the other mode cannot use.

How long is the video when I use a reference clip?

The workspace uses up to the first ten seconds of a reference video, whose source can be up to 30 seconds. Output duration is automatic and is not guaranteed to match that input segment.

Is a description required?

Yes, in every mode, up to 20000 characters. Describe what should happen, and in reference mode also say how the pictures belong together.

Does it generate sound?

There is no sound switch. The finished task we checked carried an audio track alongside the video, so check the track in your own result and mute it in the edit if you need silence.

What does a task cost?

Credits depend on the length and the tier you pick, and a task with a reference clip is quoted as one generation instead of per second. The workspace shows the exact amount for your settings before you confirm.

Related model guides

Other models on this platform that do a similar job, or that pair well with this one.

All model guides