Create Video From Text, Photos and Footage With Gemini Omni
Last verified: November 3, 2026
Gemini Omni 1.1 Flash is Google DeepMind's multimodal model for video generation, editing and cinematic control. What sets it apart is world knowledge: it combines physical realism with Gemini's understanding of history, science and culture. It belongs to Gemini Omni, a model family Google introduced at I/O 2026. Google describes the family as able to generate in any output modality from any input, and its first model was Gemini Omni Flash. The 1.1 update arrived in August 2026 as a production-ready release that gives more control over generative video. In Google Vids it goes by the shorter name Omni 1.1.
Google presents this grounding as the step from photorealism to meaningful storytelling. It matters most in scenes where reality has to hold up: how objects move, how events unfold and whether a setting fits its era. The 1.1 release adds a new set of creative controls. These include first and last frame interpolation, which builds a smooth transition between a starting image and an ending image, and output options from 360p up to 4K. You also get generation from text, images and video, plus editing in plain language.
Gemini Omni Flash is already built into other creative tools. Adobe, for example, has added it to Adobe Firefly. This page runs two versions: Gemini Omni, for text and image prompts, and Gemini Omni 1.1 Flash, which also handles frame-to-frame, reference and editing workflows.
Explore Gemini Omni's Models
What Each Run Takes In and Delivers
Clip length, framing, accepted files and how the highest resolutions are produced.
Clip length (Video Duration)
4, 6, 8 or 10 seconds in every mode except Edit video
Output resolution
720P, 1080P or 4K on both versions; Gemini Omni 1.1 Flash also offers 360P
Framing
9:16 vertical or 16:9 widescreen
Image uploads
JPG, PNG or WEBP, up to 10 MB each, at least 300×300 px; up to 7 per run in Gemini Omni Image to video and Gemini Omni 1.1 Flash Reference to video
Footage for editing
MP4 or MOV up to 10 seconds and 50 MB, plus up to 5 optional reference images (Gemini Omni 1.1 Flash Edit video)
How 1080p and 4K are rendered
Gemini Omni 1.1 Flash upscales its 1080p and 4K output; 720p is the model's default
Six Checks Before You Press Generate
Catch the problems that cause cut-off briefs, clips that ignore your references and soft results before you run either version.
Match the version to your prompt length
The Gemini Omni version takes prompts of up to 2,048 characters. Gemini Omni 1.1 Flash takes up to 20,000, or 5,000 in Edit video. If your brief is a full shot breakdown with blocking, camera path and lighting notes, choose Gemini Omni 1.1 Flash so none of it is cut off.
Name the real era, place or phenomenon
This model is built around real-world knowledge of history, science and culture. Name things outright, like "Trajan-era Roman Forum" or "the diamond-ring phase of a total eclipse", instead of using general adjectives. That gives the model something concrete to depict accurately.
Tell Reference to video how to use your images
Google's prompt guide for Gemini Omni 1.1 Flash suggests saying that the images are references for the video and should not be used as literal opening frames. Add a line like that to your Prompt. If you need the clip to open on an exact picture, use First to last frame instead.
Upload sharp images and describe the motion
Uploads only need to be 300×300 pixels, but for image-to-video Google recommends high-resolution images and specific motion descriptions. Say who moves, how they move and where the camera goes, rather than leaving the motion for the model to guess.
Make sure the First Frame can lead to the Last Frame
Gemini Omni 1.1 Flash animates the scene from the first image to the last one. The two images should share a subject, setting and lighting so the transition makes sense. Use the Prompt to name the move between them, such as an orbit, a push-in or a reveal. Uploading the same image to both slots asks for a clip that starts and ends on the same picture.
Prepare your footage for Edit video
Trim the clip to 10 seconds or less and keep it under 50 MB as an MP4 or MOV before you upload. In the Prompt, say exactly what should change, such as wardrobe, weather or background, and what has to stay the same. If the new element needs a specific look, add up to five reference images.
Gemini Omni or Kling O3: Match the Generator to Your Shot
Gemini Omni and Kling O3 both run on this site, so the real question is which one's controls suit your shot. Each row shows what the two generators let you choose or upload here, mode by mode, so you can match clip length, framing, inputs and prompt length to the job before you start.
| Feature/Spec | Gemini Omni | Kling O3 |
|---|---|---|
| Modes you can run | Gemini Omni: Text to video, Image to video. Gemini Omni 1.1 Flash: Text to video, Image to video, First to last frame, Reference to video, Edit video | Text to video, Image to video, First to last frame, Reference to video |
| Video Duration choices | Gemini Omni and Gemini Omni 1.1 Flash: 4, 6, 8 or 10 seconds (default 4) | 3 to 15 seconds in 1-second steps (default 3) |
| Resolution choices | Gemini Omni: 720P, 1080P, 4K. Gemini Omni 1.1 Flash: 360P, 720P, 1080P, 4K (720P default on both) | 720P, 1080P, 4K (default 720P) |
| Aspect Ratio in Text to video and Reference to video | Gemini Omni and Gemini Omni 1.1 Flash: 9:16 or 16:9 (default 9:16) | 1:1, 9:16 or 16:9 (default 1:1) |
| Images per Image to video run | Gemini Omni: 1 to 7. Gemini Omni 1.1 Flash: 1 | 1 |
| Images per Reference to video run | Gemini Omni 1.1 Flash: 1 to 7 in Add Image | 1 in Image plus 1 to 7 in Add elements |
| Prompt length limit | Gemini Omni: 2,048 characters. Gemini Omni 1.1 Flash: 20,000 characters (5,000 in Edit video) | 2,048 characters |
| Generate Audio switch in First to last frame and Reference to video | Gemini Omni 1.1 Flash: no such setting in these modes | Shown in both modes, on by default |
What These Differences Mean for Your Next Clip
Long single takes vs. heavily directed short shots
If one unbroken take has to run past ten seconds, or land on an odd length like 7 or 13 seconds, Kling O3's longer and finer duration scale is the practical choice. If the clip is short but every detail is planned, such as period wardrobe, the order of a natural event or a camera path with lens and lighting notes, Gemini Omni 1.1 Flash's much larger prompt limit lets you write all of it out. That suits the detail-heavy briefs where its grasp of the real world matters most.
Animating your own photos
If you have several angles of one product or character, the Gemini Omni version can use the whole set in a single Image to video run. Gemini Omni 1.1 Flash and Kling O3 each animate one picture in that mode. For reference-based work, Kling O3 gives one picture its own Image slot next to Add elements. Gemini Omni 1.1 Flash takes all references through Add Image, so say in your prompt which image does what. Both accept the same image formats and size limits, and both offer First to last frame with a two-image upload.
Editing footage, sound and framing
Gemini Omni 1.1 Flash's Edit video mode starts from footage you already have. Its 360P option is useful for a quick test of how the model reads your prompt. Each run makes a new result, though, so treat a 360P run as a check on your wording, not a preview of a final clip. Kling O3 gives you control over sound with a Generate Audio switch in its First to last frame and Reference to video modes. It also offers square 1:1 framing in Text to video and Reference to video for feed posts.
Choose by Clip Length, Inputs and Editing Needs
Use this quick guidance to pick the best option for your workflow.
How to choose: Choose Gemini Omni 1.1 Flash if you want to edit an existing clip with Edit video, write long, detailed prompts of up to 20,000 characters, or test ideas at 360P. Choose the Gemini Omni version if one Image to video run should use up to seven photos. Choose Kling O3 if a clip needs to run longer than 10 seconds or land on an in-between length. It's also the pick for 1:1 framing in Text to video or Reference to video, a Generate Audio switch in First to last frame or Reference to video, or a Reference to video run with a main Image plus up to seven elements. In other ways the two are the same: both go up to 4K, default to 720P, accept the same image formats and return one video per run.
From Prompt to Finished Clip in Five Steps
Five steps take you from picking a version to a finished video.
Choose Gemini Omni or Gemini Omni 1.1 Flash in the model list
Pick a version and a mode. Gemini Omni runs Text to video and Image to video. Gemini Omni 1.1 Flash also runs First to last frame, Reference to video and Edit video.
Add files with Upload Images, Image, First Frame, Last Frame, Add Image or Video
Text to video doesn't need any files. For other modes, upload what the mode asks for: photos for Image to video, a First Frame and Last Frame pair, reference pictures through Add Image, or a clip through Video, with optional Add Reference Image files.
Write your Prompt and refine it with Enhance Prompt
Describe the subject, action, camera move and lighting. Enhance Prompt, the button under the prompt box, suggests fuller versions of a rough idea. Your text only changes if you pick one of the suggestions.
Set Video Duration, Resolution and Aspect Ratio
Choose the clip length, the output size and 9:16 or 16:9 framing. You can also fill in Seed if you like. In Edit video, only Resolution and Seed appear.
Press Generate
Each run makes one video. Watch it, then adjust the prompt or settings and generate again for a new take.