Gemini Image to Video: Making Videos with Gemini Omni

If you used Veo in the Gemini app, the video tool has changed. Google's Gemini overview page says Gemini Omni replaces Veo in the Gemini app, describing it as a multimodal model that combines Gemini's understanding with image-to-video and video-to-video editing. Google calls it "Nano Banana for video": you describe what you want, and keep shaping the result in conversation. This guide covers what Omni does in the Gemini app, how to animate photos well, and how it relates to Veo and to Google Flow.

What Gemini Omni can do

Google's page lists Gemini Omni 1.1 Flash as the model in the app. A Google AI subscription is required, features vary by tier and location, and the service is for adults. Students may have promotional access; check the page for your country.

Turning a photo into a video

1. Start a chat in the Gemini app and add your photo (or up to five photos).
2. Describe what should happen: the motion, the camera and the sound.
3. Review the 10-second result and refine in the same chat: "slower", "keep the camera still", "add light rain".
4. When the clip is right, ask what happens next to extend the scene.

Write the prompt as a short scene description: "The woman in the photo turns towards the window and smiles as the curtains move in a breeze. Slow push in. Soft morning light. Sound: birdsong and a distant kettle." The photo already shows who and where; the prompt says what changes.

Editing by chat

Google gives examples of edits you can ask for: swap characters, adjust lighting, stabilise the video, change the background, change the wardrobe or transfer a style while keeping the details. The habits that work for image edits apply here too:

• Name what changes and what stays: "Change the background to a beach at sunset; keep the person, their clothes and the camera move the same."
• One change per message, so you can see what each one did.
• Go back to the last good version if an edit drifts, rather than editing the edit.

Working with several photos

With up to five photo references you can put a person into a place, combine a product with a setting, or show a character from more than one angle. Choose photos that share light and style, describe each one's role in words ("the woman from the first photo, in the café from the second"), and keep the action simple. Photos of other people need their consent, and Google's safety rules apply to everything you upload.

Gemini, Flow and Veo

Gemini Omni replaces Veo in the Gemini app, but Veo has not disappeared everywhere. In Google Flow, Google's AI filmmaking tool, Veo 3.1 (Lite, Fast and Quality) sits alongside Gemini Omni Flash, and Flow adds frames, ingredients, extension of Veo clips and Scenebuilder for longer pieces. For a single clip from a photo, the Gemini app is the quickest route; for a sequence of shots, Flow gives more control. See Google Flow vs Veo.

Planning a short sequence

Ten seconds is one shot, so a longer story is a series of clips. Write the shots first, one line each, then make them in order in the same chat: generate the opening shot, extend it while the action continues, and start a new clip when the camera or the place changes. Keep the same photo references and the same description of each person across every clip so they stay recognisable, and keep the light consistent unless the story needs a change. When you join the clips in an editor, short cuts hide small differences in detail far better than long dissolves.

Getting better results

• Use a sharp, well-lit photo with the subject clear of the edges; motion exaggerates flaws.
• One action per 10 seconds. A turn, a walk, a pour, a smile.
• Name the camera, including "static camera" when you want none.
• Describe sound as things you could hear in the room, and say "no music" if you will add your own.
• Keep the original photo and prompt with the video you keep; a remake starts from them.

For still images, the same app uses the Nano Banana models; see Nano Banana prompts.

Watch every result with the sound on before you approve it: native audio is part of the clip, and a wrong sound is as visible to viewers as a wrong frame.

Try it

Animate this photo: the dog on the sofa lifts its head, looks towards the door and wags its tail as someone comes home. Static camera, warm evening lamp light. Sound: keys in the door, a happy whimper, no music.

Use a sharp photo of a dog on a sofa. When the 10 seconds look right, ask "what happens next?" to extend the scene, then try an edit such as "change the room to a cosy cabin interior; keep the dog the same".