Gemini Omni video and avatars

Create one simple video with Gemini Omni, revise a detail in the same conversation, and download the result. If you want to appear in it, the optional avatar steps use your own recorded face and voice.

Before you start

• A signed-in personal Google Account, age 18 or over, with a Google AI plan and access to Create video. Features and limits vary by plan and region.
• Gemini at gemini.google.com and an idea for one short scene. You can start with text only; use only reference material you have the right to upload.
• For an optional avatar: a personal account in a supported region, English instructions, and a phone or tablet with a camera and microphone. Avatars are unavailable in the EEA, Switzerland and the UK.
• A quiet, well-lit place if recording yourself, and time to watch and listen to the result before using it.

Want to do it yourself?

Difficulty

4 out of 10

Estimated completion time

30–60 minutes, plus generation waits

Skills you’ll need: Describe a scene, optionally add an avatar, and review the entire downloaded video for visible and audio mistakes.

I’ll do it myselfI’d rather get help

How this helps your brand

A reviewed clip can explain a process or introduce an offer on a relevant website page. Giving it a clear purpose and a nearby next action helps interested visitors continue after watching.

1. 1

   Step 1: Describe one scene you can judge

   Write down the subject, action, setting, camera movement and sound. For a first trial, use a simple scene such as hands shaping a clay bowl in a bright pottery studio. Avoid trying to explain your entire business in one clip.

   Choose one thing the viewer should understand. If you need an exact price, web address or legal line, plan to add and check that text separately rather than judging the whole result only by how cinematic it looks.

   Illustrated exampleTurn a scene idea into details you can judge

   One short sceneHands shape a clay bowl

   Bright pottery studio · calm, tactile mood

   01 · establish

   Hands meet the clay

   02 · action

   Shape one bowl

   03 · finish

   Hold on the rim

   Camera

   Slow push in, then steady

   Sound

   Soft wheel hum · no dialogue

   Viewer should notice

   The hands turn the clay into a bowl

2. 2

   Step 2: Open Create video

   At gemini.google.com, open the sidebar and choose Create video. Paste the prompt below with your own scene. Omni supports conversational generation and edits, so begin with a clear first version you can refine.

   You may add reference images with Add image. For this first attempt, one relevant image is enough. If Create video is missing, check account eligibility and plan access before going further.

3. 3

   Step 3: Add your own avatar if you want to appear

   Skip this step for a video without your likeness. To create an avatar on the web, use Add Files, More uploads, then Avatar. Scan the QR code with your phone or tablet and follow the recording instructions. Return to the computer and choose Use Avatar.

   Record your own face and voice with camera and microphone access enabled. Keep your face visible and the background free of other faces. If recording fails, improve the lighting or audio and use Retake.

   For later videos, select the saved avatar using Add files, then Avatar, or type @ and choose your Google username from the menu. A name typed as plain text is not a substitute for selecting your avatar.

   Illustrated exampleRecord and select your own avatar

   1. 01
      QR

      Move to your phone

      Choose Avatar on the web, scan the QR code, and follow the recording prompts.

   2. 02

      Record only yourself

      Use a quiet, bright place. Keep your face visible and other faces out of frame.

   3. 03
      ↗

      Select the saved avatar

      Back in Gemini, choose Use Avatar. Later, select it from the Avatar menu.

4. 4

   Step 4: Generate and watch the whole result

   Submit the request and let generation finish. Play the result from beginning to end, with sound on if you requested audio.

   Check that the action makes sense across the whole clip. Look for changing objects, unexpected people, distorted lettering and audio that does not match the intended scene. For an avatar, inspect both resemblance and speech; do not accept a clip solely because the first frame looks right.

   Illustrated exampleReview the clip across its full timeline

   ▶

   Watch from first frame to lastSound on · one complete pass

   00:08

   START

   Right subject?Hands + clay

   MIDDLE

   Action holds?No object drift

   END

   Finish makes sense?Bowl stays intact

   AUDIO

   Listen for sound that matches the scene; check avatar speech and resemblance too.

   • Unexpected people
   • Distorted lettering
   • Audio mismatch

5. 5

   Step 5: Revise one visible problem

   In the same conversation, describe one change, such as “Keep the studio and subject, but make the camera stationary.” Compare the revised clip with the original request before asking for another edit.

   If the result keeps drifting, simplify the scene and remove competing actions. If a reference-video edit is unavailable, note that Google restricts video uploads for editing in the EEA, Switzerland, the UK and some US states; text-only creation or an allowed image reference is a different workflow, not a way to enable that upload feature.

6. 6

   Step 6: Download and check the saved video

   Under the generated video, open Share and choose Download video. Open the downloaded file and watch it again to confirm you saved the version you intended.

   Give the file a useful name and keep the prompt with your project notes. If you no longer want a saved avatar, Google provides deletion through Settings, Personal avatar and your Google Account recording controls. Deleting recordings does not remove videos you already made.

   If the clip will support your website, choose its page and purpose before adding it: explain a process, introduce a service, or answer one customer question. Add it using your site’s supported media controls, keep the essential information available as text, and check playback, sound controls, captions where needed, and phone loading. Place a relevant contact or booking action nearby and test it. A finished video file does not by itself bring visitors or capture enquiries.

Create one short video of [subject] doing [one action] in [setting]. The mood is [mood] with [lighting]. Use [camera framing and movement]. Audio: [sound or no dialogue]. Keep the scene focused on [the one thing viewers should understand]. Do not add logos, captions or readable text.

If I selected my avatar, use it as [the avatar’s role in the scene]. Otherwise, create the scene without my likeness.