How to Generate and Edit Videos with Gemini Omni

Gemini Apps no longer treat video as a one-shot “describe it and hope.” Gemini Omni is the current video generator and editor in the Gemini app. Google’s own overview describes it as conversational video: start from text, photos, or an existing clip, then keep editing in the same thread.

This guide follows the official Gemini Apps Help article and the Gemini Omni product page. It is a consumer walkthrough, not an API cookbook. For developers, Google documents Gemini Omni Flash and Veo 3.1 separately in the Gemini API.

What you need before you start

Google’s help page is explicit:

• A personal Google AI plan (Pro or Ultra, depending on what your region offers), or a qualifying Workspace license for work or school accounts
• You must be signed in to Gemini Apps
• You must be 18 or over — the feature is not available to users under 18
• You must have the rights to any photo or video you upload

Features vary by plan and country. Video-to-video uploads for edits are not available in the EEA, Switzerland, the United Kingdom, and some U.S. states.

Do not generate or share clips to deceive, harass, or harm anyone. Generated videos in the Gemini app carry SynthID, Google’s invisible watermark.

What Omni adds over older Veo-only flows

On the official overview, Google states that Gemini Omni will replace Veo in the Gemini app. Help Center lists the product-level differences:

• Video-to-video edit — upload a clip and tell Omni what to change (where the region allows uploads)
• Multi-turn editing — keep refining one conversation; each turn can include image or video references
• Aspect ratio control — pick the frame before generation; text-only defaults to landscape; a reference image or video sets the ratio to match that file
• Templates — apply a curated style in one tap
• Avatars — type @ plus your Google username to put your saved avatar in the shot

You can also ask Gemini to generate audio with the clip.

Watch how native audio video works

Google’s official Veo 3 announcement clip is still the clearest public demo of text-to-video with synced sound. Omni is the current Gemini-app model; the clip shows the same job: describe a scene, get picture and audio together.

Generate a video on desktop

1. Open gemini.google.com and sign in.
2. Open the sidebar and choose Create video. You can also use the Gemini Omni entry point.
3. Optional: pick a template if you want a starting style instead of a blank prompt.
4. Write one concrete prompt: subject, action, camera, lighting, and sound.
5. Optional: click Add image and/or Add video. Help Center allows one video and up to five images.
6. To use your avatar, include @ and your Google username in the prompt.
7. Click Submit and wait. Generation can take a few minutes. You cannot keep chatting in that same thread while it runs — open a new chat and come back.

A prompt that matches how Omni is meant to be used:

Wide tracking shot of a ceramic mug sliding across a sunlit oak table. Steam rises. Soft kitchen ambience, no dialogue. Keep the mug logo readable. Landscape.

If you only need a still product photo, do not burn a video slot. If the motion is the point, name the camera move and the audio in the same sentence.

Generate a video on Android

1. Open the Gemini app or gemini.google.com on the phone.
2. At the bottom, tap More, then Video (wording can appear as Video under the extra tools).
3. Type the prompt. To start from a photo, tap Add image and pick a file you have rights to.
4. Submit and wait one to two minutes on a typical run.

Vertical source photos stay vertical and use the closest available aspect ratio. You can still brainstorm the prompt in a normal chat first; Help Center notes that if the chat already understands the brief, you may not need to tap Video again.

Edit without starting over

After the first clip lands, stay in the thread. Help Center examples of follow-up prompts:

• Remove or replace an object or character
• Change the camera angle
• Modify the scene (wardrobe, background, lighting)

Google’s overview adds scene extension (“what happens next”) and style transfer while keeping subject details. Treat each turn as one change. Stacking five unrelated edits in one message is how clips fall apart.

Good second turns:

• “Keep the mug and table. Switch to a 35mm close-up. Rain on the window behind.”
• “Remove the extra spoon on the right. Do not change the steam.”
• “Same shot, add a quiet kettle whistle under the room tone.”

Templates, avatars, and audio

Templates are for when you want a house style before you write a long brief. Open Create video, pick a template, then add only the details that must be unique (product, location, line of dialogue).

Avatars are optional. Create and manage them from Gemini’s avatar help. Only you can use your avatar. Do not treat an avatar as a license to impersonate someone else.

Audio is a prompt, not a separate mixer. Name dialogue with a colon after the speaker action, and avoid quotation marks if you do not want on-screen text. Google’s video prompt guidance for Omni and Veo warns that quotation marks often render as burned-in captions.

Download, share, and verify

On desktop, under the finished video click Share, then Share on YouTube or Download video. On Android, open the video and tap Download.

If you need to check whether a file came from Google’s generators, Gemini Apps include a verification tool that looks for SynthID and Content Credentials. Upload one file at a time (video under 90 seconds and 100 MB in the help article’s limits) and ask whether it was generated or edited by Google AI.

Gemini may refuse or later remove a clip if systems detect a Terms of Service or Prohibited Use Policy issue.

A concrete example

Suppose you need an eight-second hero loop for a landing page: a desk plant in late-afternoon light, no people, soft room tone.

1. Create video on the web.
2. Skip templates.
3. Prompt: “Static medium shot of a monstera on a white desk, late sun from the left, dust in the beam, quiet room tone, no music, no text, landscape.”
4. If the first pass invents a window sticker, comment: “Remove all text and logos. Keep the plant and light.”
5. Download the MP4. Overlay your real product UI in an editor if the page needs a screen.

That loop — one job, one constraint list, one edit — is more reliable than a paragraph that tries to storyboard three scenes.

When to leave the Gemini app

• Google Flow if you are assembling a longer sequence with tools and project assets
• Google Vids if you need a scripted Workspace video with scenes you will keep editing as a document
• Gemini API / Vertex AI if an app must call Omni Flash or Veo 3.1 (extension, first/last frame, 1080p or 4K on Veo 3.1) from code

The Gemini app is the right surface for a short clip you will download today.

Limits to plan around

• Paid plan or qualifying Workspace license; not for under-18 accounts
• Generation blocks the same chat until the clip is ready
• Upload caps: one video and up to five images per Help Center
• Video-to-video upload blocked in several regions
• Daily or plan caps exist; Google notifies you as you near them rather than publishing a single global number in Help
• Watermarks and policy filters are not optional

Conclusion

Gemini Omni is the video path inside Gemini Apps: Create video, write a tight prompt, attach only files you own, then edit in the same conversation. Use templates when you want a preset look. Use an avatar only for yourself. Download or send the clip to YouTube when the frame is right, and treat SynthID as part of the file, not a bug.

Start at gemini.google.com or the Gemini Omni page. Confirm plan availability in your account before you promise a volume of clips to anyone else.

Sources

• Generate videos with Gemini Apps — Gemini Apps Help
• Gemini Omni video generation overview — Google
• Verify AI-generated images, videos, and audio — Gemini Apps Help
• Video generation in the Gemini API — Google AI for Developers
• Create videos with Veo 3 in the Gemini app — Google on YouTube