Updated

Google DeepMind

Gemini Omni

Google DeepMind's Gemini Omni: video from text, reference images or a clip, at up to 4K on Flashloop.

Create with Gemini OmniView pricing

What is Gemini Omni?

Gemini Omni is Google DeepMind's multimodal video model, announced at Google I/O on May 19, 2026, with Gemini Omni Flash as its first release.

It reads text, images and video together and produces one consistent clip, and it can rework an existing clip. On Flashloop it renders 4 to 10 second clips at up to 4K in 16:9 or 9:16.

Google positions it as a different kind of model from the Veo line rather than the next Veo. Instead of treating a prompt and a reference image as separate instructions, it reasons across everything you give it at once, which helps when a scene has to respect a character from one image, a location from another and a motion from a clip. On Flashloop you get up to 7 reference slots, where each image uses one slot and a reference clip uses two.

The first release is capped at 10 seconds, which Google describes as a product decision with longer lengths planned. When you hand it a clip, it also decides the output length itself. That makes it best for short, reference-heavy shots and for building on footage you already have, rather than long takes, where Seedance 2.5 or Wan 3.0 are the better fit.

Provider
Google DeepMind
Length
10 seconds
Resolution
720p, 1080p, 4K
Aspect ratios
16:9, 9:16
Audio
Silent output

What you can do with Gemini Omni

  • Three ways in

    text, up to 7 reference images, or a reference clip

  • Video-to-video

    trim a source clip and let the model build on it

  • Up to 4K output on Flashloop

  • One model that reads all inputs together instead of one at a time

Prompts to try with Gemini Omni

“The woman from the first reference image walks into the sunlit greenhouse from the second image, runs her hand along the leaves and looks up at the glass roof. Slow dolly forward, soft morning light. 8 seconds, 16:9, 4K.”

Try this prompt

Where Gemini Omni fits best

  • Storyboards with a fixed cast

    Attach reference images for each character and location and keep them consistent from shot to shot.

  • Extending an existing clip

    Trim a source clip to the moment you want and let Gemini Omni continue or rework the scene from there.

  • 4K hero shots

    Short, detailed cinematic clips rendered at 4K for landing pages, trailers or large screens.

  • Vertical and landscape from the same brief

    Render the same scene in 16:9 for web and 9:16 for social without rewriting the prompt.

How Gemini Omni compares

Gemini Omni compared with other video models on Flashloop
ModelLengthTop resolutionAspect ratiosNative audio
Gemini Omni10 seconds4K16:9, 9:16No
Veo 3.18 seconds1080p16:9Yes
Kling 3.015 secondsPro (1080p)16:9, 9:16, 1:1No
Kling 3.0 Turbo15 seconds1080p16:9, 9:16, 1:1No
Kling 4.030 seconds1080p16:9, 9:16, 1:1 +1Yes

Good to know before you generate

  • With a reference clip, the model decides the output length
  • On Flashloop it offers 16:9 and 9:16 only
  • Images and clips share a 7-slot quota: each image uses 1 slot and a clip uses 2
  • Clips stop at 10 seconds; Google has said longer lengths are planned

Start creating with Gemini Omni

Create with Gemini Omni

  1. Select Gemini Omni

    Open Flashloop and choose Gemini Omni from the model selector in the video creator.

  2. Enter your prompt

    Describe what you want to see in your video.

  3. Generate and download

    Hit generate and your video will be ready in seconds. Download or share directly.

Questions about Gemini Omni

What input modes does Gemini Omni support on Flashloop?

Three: text alone, up to 7 reference images, or a reference clip with a trim range for video-to-video. You can mix images with a clip as long as you stay within the 7-slot quota, where each image is 1 slot and the clip is 2.

Can Gemini Omni generate at 4K?

Yes. On Flashloop it renders at 720p, 1080p or 4K. 4K costs more per generation because the render is heavier.

What durations does Gemini Omni support?

4, 6, 8 or 10 seconds. When you attach a reference clip, the model picks the output length itself and your length setting is ignored.

Which aspect ratios are available?

16:9 landscape and 9:16 portrait on Flashloop.

How does the 7-slot reference quota work?

Each reference image uses 1 slot and a reference clip uses 2, with a maximum of 7. So you can use 7 images alone, 5 images plus a clip, or a clip by itself.

Who makes Gemini Omni?

Google DeepMind. Google also offers it in the Gemini app, Google Flow and YouTube Shorts. Flashloop gives you access to it alongside other video models.

How is Gemini Omni different from Veo 3.1?

Both come from Google DeepMind. Veo 3.1 is built around native audio in 8 second 16:9 clips at up to 1080p on Flashloop. Gemini Omni focuses on combining text, images and video as inputs, goes up to 10 seconds and renders up to 4K in 16:9 or 9:16.

More video models on Flashloop

Browse all models