Create Video from Text, Images and Footage with Gemini Omni
Last verified: November 3, 2026
Gemini Omni Flash does more than build scenes that look real. It works out what should happen next and grounds each video in Gemini's knowledge of how the world works, so a prompt can set up a cause and leave the consequence to the model. It is the first model in Gemini Omni, Google DeepMind's family of multimodal generative models built to create from any input, starting with video. It debuted at Google I/O on May 19, 2026, as a transformer-based model with native multimodal support. Gemini Omni 1.1 Flash followed on August 27, 2026, as a production-ready update for professional video work.
World knowledge sits at the center of the line. Gemini Omni 1.1 Flash combines an understanding of physics with knowledge of history, science and cultural context, and Gemini Omni Flash also draws on biology and narrative logic, so its scenes make sense as well as look right. Omni Flash is built for video coherence, character consistency, factual accuracy and conversational editing, such as replacing an element or changing the perspective of an existing clip in plain language. The 1.1 release added scene extension, first and last frame control, video input references, fast 360p drafting and upscaling up to 4K.
Each Gemini Omni 1.1 Flash generation tops out at 10 seconds. Google is making Omni the successor to Veo inside its Gemini app. On this page, Gemini Omni runs Text to video and Image to video, and Gemini Omni 1.1 Flash runs First to last frame, Reference to video and Edit video.
Explore Gemini Omni's Models
What You Can Set and Upload
Clip lengths, resolutions, framing and inputs for the five Omni modes on this page.
Clip length (Video Duration)
4, 6, 8 or 10 seconds; starts at 4
Output resolution
720P, 1080P or 4K; Gemini Omni 1.1 Flash also offers 360P
Frame shape (Aspect Ratio)
9:16 vertical by default, or 16:9 widescreen
Photo inputs
1–7 images in Image to video (Gemini Omni) and Reference to video (Gemini Omni 1.1 Flash)
Keyframe pair
One First Frame and one Last Frame image in First to last frame (Gemini Omni 1.1 Flash)
Footage to edit
One MP4 or MOV clip up to 10 seconds, plus up to 5 optional reference images (Gemini Omni 1.1 Flash)
Check These Before You Hit Generate
Six checks that prevent the wrong mode, cramped timing, and soft or badly framed results.
Match the mode to the version first
Text to video and Image to video run on Gemini Omni. First to last frame, Reference to video and Edit video run on Gemini Omni 1.1 Flash. Each mode has its own uploads and prompt limit, so pick it before you start writing. A long shot-by-shot brief written for a 1.1 Flash mode won't fit the shorter Gemini Omni prompt box.
Give the model a cause to play out
Omni Flash is built to reason about what happens next. Describe the setup, the trigger and the order of events (the kettle boils, the lid rattles, steam fogs the window) instead of describing one finished frame. Set Video Duration to 8 or 10 seconds for multi-step action, because the 4-second default only has room for about one beat.
Change the default frame and resolution
Aspect Ratio starts at 9:16 and Resolution at 720P. If you want a widescreen clip or a 4K master, change both before you generate. If you leave them, you get a vertical 720P video.
Draft at 360P with Gemini Omni 1.1 Flash
In First to last frame and Reference to video, test a new prompt or image set at 360P to check motion and staging. Google added this tier to 1.1 Flash for fast drafting. The 1080P or 4K run you do next is a new generation, not an enlarged copy of the draft.
Prepare keyframes and references that match
Images for First Frame, Last Frame and Add Image must be JPG, PNG or WebP, at least 300×300 pixels and no larger than 10 MB. Compose both keyframes for the Aspect Ratio you plan to use, and keep the lighting and lens look the same in both so the transition moves through one consistent world.
Trim your footage before Edit video
Edit video on Gemini Omni 1.1 Flash takes an MP4 or MOV of up to 10 seconds and 50 MB. Besides the Prompt, it only has Resolution and Seed settings, so cut and frame the clip before you upload it. Describe one clear change, such as swapping an object, shifting the viewpoint or restyling the setting, and attach reference images for any new element.
Gemini Omni or Veo 3.1: Pick the Right Generator for Each Shot
The first five rows show what you can set and upload for each generator on Vidofy, version by version. The last three cover what Google publishes about Gemini Omni and Veo 3.1 as models.
| Feature/Spec | Gemini Omni | Veo 3.1 |
|---|---|---|
| Clip lengths you can pick | 4, 6, 8 or 10 seconds (Gemini Omni in Text to video and Image to video; Gemini Omni 1.1 Flash in First to last frame and Reference to video) | 8 seconds, fixed (Veo 3.1 and Veo 3.1 Fast, every mode) |
| Resolutions you can output | 720P, 1080P or 4K (Gemini Omni); 360P, 720P, 1080P or 4K (Gemini Omni 1.1 Flash, every mode) | 720P, 1080P or 4K (Veo 3.1 and Veo 3.1 Fast) |
| Frame shapes you can choose | 9:16 or 16:9 (Gemini Omni in Text to video and Image to video; Gemini Omni 1.1 Flash in First to last frame and Reference to video) | 9:16 or 16:9 (Veo 3.1 and Veo 3.1 Fast) |
| Images you can add in Reference to video | 1–7 images (Gemini Omni 1.1 Flash) | 1–3 images (Veo 3.1 Fast) |
| Prompt length you can write | Up to 2,048 characters (Gemini Omni); up to 20,000 (Gemini Omni 1.1 Flash in First to last frame and Reference to video); up to 5,000 (Gemini Omni 1.1 Flash in Edit video) | Up to 2,048 characters (Veo 3.1 and Veo 3.1 Fast) |
| What the model takes in and puts out | Gemini Omni 1.1 Flash: text, image and video in (video up to 10 s for editing and extension); video out | Text and image in; video is output only; audio input not supported |
| Frame rate the model outputs | Gemini Omni 1.1 Flash: 24 FPS | 24 FPS |
| When the model was released | Gemini Omni Flash: May 19, 2026 ; Gemini Omni 1.1 Flash: August 27, 2026 | General availability: November 17, 2025 |
Where the Choice Changes Your Workflow
Clip timing and the resolution ladder
Gemini Omni and Gemini Omni 1.1 Flash let you fit each clip to the beat it has to fill, whether that's a short loop, a mid-length product reveal or a full ten-second story moment. Veo 3.1 returns every clip at one length. If eight seconds is exactly what your edit needs, both get you there. If it isn't, the Omni versions are where you get other lengths.
Both generators reach 4K at the top end. The difference is at the bottom: Gemini Omni 1.1 Flash adds a 360P tier that Google built for fast drafting. Use it to check a keyframe pair or a reference set before you commit to a 4K delivery run. Both sides offer 9:16 and 16:9, and both models output 24 FPS, so clips from either one fit the same vertical or widescreen timeline without frame-rate conversion.
References, prompt depth and existing footage
For scenes with several subjects, Reference to video on Gemini Omni 1.1 Flash takes more than twice as many images as Veo 3.1 Fast. That's enough to give a character, their wardrobe, a hero prop and a location their own reference images in a single brief. The 1.1 Flash keyframe and reference modes also accept much longer prompts than the limit Gemini Omni's text and image modes share with Veo 3.1, so shot-by-shot direction with timing, lens and lighting notes fits in one run.
Existing footage is the other dividing line. Google lists video as an input for Gemini Omni 1.1 Flash, and the Edit video mode on this page relies on it. Veo 3.1's published modalities take text and images and list video as output only. If the job starts from a clip you already have, use Gemini Omni 1.1 Flash. If it starts from a prompt or still images, either generator works.
Deciding Between Gemini Omni and Veo 3.1 for Your Next Clip
Use this quick guidance to pick the best option for your workflow.
How to choose: Choose Gemini Omni or Gemini Omni 1.1 Flash when a clip needs to run 4, 6 or 10 seconds instead of 8. Choose Gemini Omni 1.1 Flash in four cases: you want a 360P draft tier below 720P; a scene needs more than three reference images (it takes up to seven); your brief runs past 2,048 characters (its First to last frame and Reference to video modes accept up to 20,000); or you're starting from existing footage, which it accepts as video input. On 720P to 4K output, 9:16 and 16:9 framing, and 24 FPS, the two generators are the same.
Go from Idea to Omni Clip
Five steps from picking a mode to a finished clip.
Pick a version and mode in the model list
Choose Gemini Omni for Text to video or Image to video, or Gemini Omni 1.1 Flash for First to last frame, Reference to video or Edit video.
Add images or a clip with the upload buttons
Use Upload Images for Image to video, Upload first frame and Upload last frame for keyframe transitions, Add Image for Reference to video, or Click to upload the video to edit (with optional Add Reference Image) for Edit video. Text to video needs no upload.
Write your Prompt or use Enhance Prompt
Describe the subject, the action, the camera and the lighting. The Enhance Prompt button under the prompt box suggests fuller versions of a rough idea, and your text only changes if you pick one.
Set Video Duration, Resolution and Aspect Ratio
Choose the clip length, the output resolution and the frame shape. Edit video only asks for Resolution. You can also enter a number in Seed if you want.
Click Generate
Each run returns one video. Watch it, adjust the prompt or settings, and run again if you want a different take.