TL;DR
A Google Flow AI avatar is a reusable digital version of you built from your face and voice recordings. Once it is set up, you can use @me inside Google Flow prompts to place yourself into new AI-generated scenes.
Before you start, know these six things:
- Google Flow and the Avatar feature do not have the same regional availability. Avatars are currently unavailable in the EEA, the UK, and Switzerland.
- You can start avatar setup on a computer, but you need a phone or tablet to finish the face and voice capture.
- Your first test should be simple: one short line, one small action, one camera angle.
- The same avatar can be reused, but consistency can still drift when you change too many things at once.
- Saved frames and references can help you build more consistent scenes.
- The best use of an AI clone is to remove content friction, not to replace every real video you make.
If you have ever thought, “I would post more video if I did not have to set up the camera again,” this is where the feature gets interesting.
You may have seen people search for an AI avatar generator, an AI clone, a digital twin, or an AI avatar for TikTok and Reels. Google Flow now gives eligible users a way to build that idea around their own face and voice.
You create yourself once. Then you can place that digital version of you at a desk, in a coffee shop, beside a product, in a hotel room, or inside a short-form video without filming each scene from scratch.
That sounds a little like cheating at content creation.
It is not.
It is more like giving your camera-shy Tuesday afternoon a backup plan.
The trick is learning where Google Flow helps, where it can still get weird, and how to keep the finished content feeling like you.
Before You Create Your Google Flow AI Avatar
There is one thing to check before you spend ten minutes hunting for a button that may not exist on your account.
Flow Access and Avatar Access Are Different
Google Flow is available in many countries, but the Avatar feature has extra regional limits.
Google currently says Avatars are not available in:
- the European Economic Area
- the United Kingdom
- Switzerland
Google Flow itself is available in places such as the UK and Switzerland, so having Flow does not automatically mean you have Avatars. Check Google’s Flow availability by region before blaming your browser, your account, or the poor refresh button.
Google’s current Flow access guidance also says users must be 18 or older and meet its account and plan requirements. Those requirements can change, so the safest approach is to check the current Google Flow access guide before paying for anything.
What You Need
For the avatar setup itself, keep it simple:
- an eligible Google account
- access to Google Flow
- a supported Avatar region
- a phone or tablet
- decent, even light
- a quiet room
- your full face visible
- a few uninterrupted minutes
You do not need a studio, a cinema camera, or a light that could signal passing aircraft.
For a Cleaner Face Capture
Hold your phone around eye level.
Keep your eyes, nose, and mouth easy to see. Normal glasses are fine, but avoid sunglasses, masks, hats, harsh backlight, or anything else that hides your features.
Google recommends clear facial visibility and lighting that is not too dark or too bright. Google avatar recording tips
Move slowly when the screen asks you to turn your head.
The goal is not to look impressive.
The goal is to give the system a clean reference.
For a Cleaner Voice Capture
Use the voice you actually use.
Speak clearly, but do not suddenly sound like you are announcing the final round of a boxing match.
Reduce:
- television noise
- music
- other people talking
- loud fans
- echo
- whispering
- shouting
Google asks you to recite a short audio sequence as part of voice verification.
Quick tip: If your room is noisy, move closer to curtains, a couch, or a closet full of clothes. You are not recording a studio album. You just want less echo and background noise.
What Is a Google Flow AI Avatar?
A Google Flow AI avatar is a personal avatar made from your own visual and audio likeness. Google describes it as a unique digital version of you created from selfie and voice recordings.
Think of it as a reusable version of your face and voice that can appear inside new scenes.
What Can Stay Recognizable
Your avatar can carry forward your:
- facial likeness
- voice
- general appearance
- recognizable identity
What Can Change From Scene to Scene
| You can change | Example |
|---|---|
| Location | Home office, café, hotel, street |
| Outfit | T-shirt, hoodie, suit, casual shirt |
| Activity | Talking, walking, holding a product |
| Camera | Close-up, medium shot, phone-camera feel |
| Dialogue | Tip, story, FAQ, product explanation |
| Mood | Calm, funny, thoughtful, energetic |
| Background | Kitchen, office, shop, travel setting |
That is the useful part.
You are not building one talking head that has to live forever in the same digital room.
How to Create Your Google Flow AI Avatar
The setup is short.
Most of the useful work happens after the avatar exists.
Step 1: Open Google Flow
Open Flow and start a new project or enter one you already have.
Step 2: Open Your Profile
Select your profile image or account initial.
Choose Create avatar.
If that option is missing, check the regional limits above before assuming something is broken.
Step 3: Start the Setup
Choose Get started.
If you begin on a computer, Google gives you a QR code.
Scan it with your phone or tablet to continue.
Step 4: Record Your Voice
Follow the on-screen directions and recite the short audio sequence.
Speak normally.
The system is trying to capture you, not your “professional voice.”
Step 5: Record Your Face
Follow the prompts as the camera captures several angles of your face.
Keep your face inside the frame and move slowly.
Step 6: Finish Verification
Google verifies your likeness and identity before the avatar is created. Google’s current Avatar setup instructions document the voice sequence, facial capture, and verification process.
Once that process is complete, your personal avatar becomes available for supported uses.
Worth knowing: Google says Avatar data cannot be shared through Flow public project links or Flow Tools.
How to Use @me With Your Google Flow AI Avatar
Once your avatar exists, remember one tiny command:
@me
Type it in the prompt box, then tell Flow what your avatar should be doing.
Google officially documents @me as the reference for your personal avatar in Flow prompts.
A first prompt could look like this:
@me sitting at a small home-office desk, looking toward a phone camera and saying, “Here are three things I wish I knew before I started creating content.” Soft daylight, normal room sound, relaxed posture and vertical social-video framing.
You do not need to write like a film-school textbook.
You need to tell Flow what matters.
A Simple Google Flow AI Avatar Prompt Formula
Use:
@me + location + action + camera + dialogue + environment + mood
For example:
@me sitting beside a window in a neighborhood coffee shop with a laptop open on the table. A phone camera sits across from me at table height. I close the laptop, look toward the camera and say, “This is the part nobody tells you when you start working for yourself.” Natural café sound, ordinary daylight and relaxed movement.
That prompt tells Flow:
- who is in the scene
- where the person is
- what is happening
- what the camera sees
- what is being said
- what the environment should feel like
That is enough direction without turning the prompt into homework.
Make Your First Google Flow AI Avatar Video Boring
This sounds like bad creative advice.
It is excellent testing advice.
Your first video has one job:
Does this look and sound enough like me to build on?
Good First Tests
Try:
- sitting at a desk
- sitting on a couch
- standing in a kitchen
- walking slowly
- holding one object
- saying one short sentence
Save the Chaos for Later
Avoid starting with:
- running
- dancing
- several outfit changes
- long dialogue
- several people moving around you
- complicated hand movements
- multiple camera moves
- huge facial reactions
Every extra action gives the generation another chance to drift.
Get one clean result first.
Then start making life difficult for it.
How to Make Your Google Flow AI Avatar Look More Like You
A verified identity helps, but your prompt still affects how natural the result feels.
Keep Movement Smaller Than You Think
Try:
@me slowly turns toward the camera.
before:
@me spins around, points at the camera, laughs, grabs a bag and runs toward the door.
The second prompt is not impossible.
It is simply asking the model to solve more problems at once.
Use Simple AI Clone Poses First
If you are testing different AI clone poses, start with ordinary body language:
- sitting naturally
- standing with relaxed arms
- holding one object
- leaning lightly against a desk
- walking slowly
- looking toward the camera
These are easier to judge because you can focus on the face, body and voice instead of wondering why your digital twin suddenly developed three new hobbies in one shot.
Give Your Hands One Job
Good:
I pick up the coffee mug and take one sip.
Harder:
I pick up the mug, turn the logo toward the camera, point to it, put it down, open my laptop and wave.
Hands already have a difficult reputation in AI video.
There is no need to give them a side hustle.
Keep Facial Reactions Natural
Small expressions often feel more believable:
- slight smile
- short laugh
- calm look
- thoughtful pause
- mild surprise
If you ask for a huge reaction, you may get a face that looks like it just remembered the oven was left on three cities away.
Change One Thing at a Time
If the mouth looks odd, shorten the dialogue first.
Do not also change:
- outfit
- room
- camera
- lighting
- movement
- mood
If you replace everything, you will not know what helped.
A Simple One-Change Test
- Keep the scene.
- Keep the outfit.
- Keep the camera.
- Change one problem area.
- Generate again.
- Compare.
That tiny habit can save a surprising amount of wasted testing.
How to Make Your AI Clone Sound Like You
The cloned voice is only part of the job.
Your script still has to sound like something you would say.
If you normally speak plainly, do not give your avatar this:
“Today I am going to share three transformative strategies designed to optimize your content-production workflow.”
Your own voice can say that sentence and still sound like somebody borrowed your face for a webinar.
Try:
“Here are three things that made creating content easier for me.”
Read every spoken line out loud.
Ask:
Would I actually say this?
Pay attention to:
- sentence length
- humor
- pauses
- slang you genuinely use
- how formal you are
- how quickly you get to the point
- phrases your audience already connects with you
Your digital twin should copy your habits, not only your cheekbones.
7 Copy-Ready Google Flow AI Avatar Prompts
These are starting points.
Change the spoken words so they fit your real voice.
1. Educational Talking-Head Video
@me sitting at a normal home-office desk, speaking casually toward a phone camera. I glance briefly at my laptop, look back and say, “Here is one thing I wish I understood earlier.” Relaxed posture, soft daylight, normal home-office background, vertical framing and natural room sound.
Best for: creators, coaches, consultants, founders
2. Coffee-Shop Creator Story
@me sitting beside a window in a neighborhood coffee shop with a laptop and coffee on the table. I finish typing, close the laptop, look toward the camera and share one short lesson about working for myself. Casual smartphone framing, normal café customers in the background and soft daylight.
Best for: founders, freelancers, personal brands
3. Product Demo
@me standing in a normal bathroom during a morning routine, using the featured product while looking toward the mirror. The phone sits beside the mirror at chest height. I finish using the product, glance toward the camera and explain one thing I genuinely like about it. Natural bathroom light and small hand movements.
Best for: ecommerce, beauty, lifestyle content
Do not claim you bought, tested, loved, or recommend a product unless that is true.
4. Founder Story Video
@me walking slowly along a quiet neighborhood street while a friend films from a few steps away. I look toward the camera and share one real lesson from building my business. Natural walking pace, everyday clothing, realistic outdoor sound and vertical framing.
Best for: entrepreneurs, service businesses, personal brands
5. Customer FAQ Video
@me sitting at a desk with a phone beside the laptop. I look toward the camera and say, “I get this question all the time, so here is the short answer.” Calm delivery, simple background, one steady camera angle and natural daylight.
Best for: service businesses, consultants, educators
6. Quick Social Tip
@me standing beside a simple whiteboard in a small office. I point once toward one short note on the board and explain the idea in one sentence. Clear delivery, soft daylight, vertical video framing and light room ambience.
Best for: Reels, TikTok, Shorts, educational content
7. Lifestyle Travel Clip
@me walking into a hotel room with a small suitcase, placing it near the door and opening the curtains. I pause at the view, turn slightly toward the camera and say one short line. Natural daylight, normal room ambience and relaxed movement.
Best for: travel creators, hotels, lifestyle brands
How to Keep the Same Avatar Across Different Videos
Consistency matters more than novelty when you are building recognizable creator content.
Keep These Fairly Stable
Try to keep your:
- personal avatar
- facial identity
- hairstyle
- approximate age
- general body build
- speaking style
Change These When the Story Needs It
You can vary:
- location
- clothing
- activity
- camera angle
- product
- dialogue
- mood
Google Flow supports visual references, ingredients, saved frames, start frames and end frames in supported workflows.
If one frame nails your face, lighting and general look, save it.
Do not throw away a good result just because the Generate button is sitting there looking tempting.
Use Saved Frames as Anchors
A useful workflow is:
- Generate a simple clip.
- Pause on the strongest frame.
- Save that frame.
- Reuse it as a reference, ingredient, start frame or end frame where supported.
- Keep the next prompt focused.
Google’s current Flow model support shows that different models support different reference modes, aspect ratios and clip lengths.
Google also said at I/O 2026 that Gemini Omni Flash improved character consistency, but “improved” should not be read as “perfect every time.”
How to Build a Full Video From Short Avatar Clips
One generated clip does not need to carry your whole story.
In many cases, three shorter clips are easier to control and more pleasant to watch.
A Simple Three-Clip Story
Clip 1: Hook
Introduce the problem.
@me sitting in my parked car after a meeting, looking toward the phone and saying, “I made the same mistake again today.”
Clip 2: Story
Move the idea forward.
@me walking into my home office, putting my bag down and explaining what happened.
Clip 3: Payoff
Give the lesson.
@me sitting at my desk and explaining what I will do differently next time.
Now you have a beginning, middle and end.
More important, each clip has one job.
Join Your Clips in Scenebuilder
Google Flow’s desktop Scenebuilder can:
- arrange several clips
- change their order
- trim beginnings and endings
- preview a full sequence
- download the finished scene
Google explains the current workflow in its Google Scenebuilder instructions.
A simple process is:
- Generate three short scenes.
- Choose the strongest version of each.
- Add them to Scenebuilder.
- Put them in story order.
- Trim awkward starts and stops.
- Watch the full sequence.
- Check for face, outfit or background drift.
- Download the scene.
- Add captions, branding or music if your publishing workflow needs them.
The point of Scenebuilder is not to create another complicated system.
Put the good clips in the right order.
What to Do When Your Google Flow AI Avatar Looks Weird
It will happen.
Do not take it personally.
Your digital twin is having a day.
| Problem | Try this first | Why it may help |
|---|---|---|
| Face changes | Reduce movement | Fewer moving details are easier to keep stable |
| Mouth looks odd | Shorten the spoken line | Long dialogue adds more lip movement |
| Body feels stiff | Ask for smaller gestures | Simpler movement can look more natural |
| Hands look wrong | Reduce object interaction | Fewer hand actions reduce complexity |
| Outfit changes | Describe clothing clearly | Stable wardrobe details help continuity |
| Voice feels wrong | Rewrite in your normal style | A script can feel wrong even if the cloned voice is close |
| Background shifts | Simplify the location | Busy scenes create more details to maintain |
| Person drifts between clips | Reuse strong references | Saved anchors can help continuity |
| Scene feels fake | Use a normal activity | Everyday movement can look more believable |
| Too much happens | Remove actions | One clip does not need five jobs |
Creators using current Flow and Omni workflows still report face, wardrobe and scene drift in longer or more complicated sequences. That is community experience, not a guaranteed result for every user, but it is useful to know before you spend an hour trying to fix one nine-minute generation.
Optional Gear If You Mix AI Clips With Real Footage
You do not need extra gear to create the avatar itself.
But if you plan to mix your AI clips with real talking-head footage, product demos or behind-the-scenes video, a few simple tools can help the real footage look and sound more consistent.
| Tool | Why it helps | Amazon USA option |
|---|---|---|
| Phone tripod | Keeps framing steady and repeatable | Ulanzi phone tripod |
| Wireless microphone | Gives your real clips cleaner speech | DJI Mic Mini |
| Small LED light | Helps keep your face evenly lit | Ulanzi VL49 light |
A phone tripod is useful when you want the same framing across several real clips. Ulanzi’s current MT-44 line is designed for phones and compact creator setups.
A small wireless microphone matters more when you are mixing AI video with real speaking footage. DJI describes the Mic Mini as a compact wireless microphone designed for creator recording.
A pocket LED light is optional, but it can help if your room is dim or your face changes dramatically from one recording setup to another. Ulanzi lists its VL49 line as a compact creator/video light.
You do not need all three.
If your phone is already stable, your room is bright and your audio sounds clean, congratulations.
You just saved money.
Use Your AI Avatar Without Losing the Human Part
The biggest risk is not that your clone looks slightly strange.
It is that your content starts feeling like nobody is home.
AI can help you produce more.
It cannot supply your lived experience, taste, judgment or point of view.
Buffer’s 2026 creator trust outlook makes a similar point: as realistic AI content becomes easier to produce, audiences have more reason to look for human signals such as authorship, point of view and lived experience.
Hootsuite’s recent AI content workflow guide also recommends being selective about what gets handed to AI and keeping human review, judgment and brand voice in the process.
Social Media Examiner’s AI content scale guide shows the same useful balance: AI can help with speed and consistency, but the expertise and audience understanding still need to come from the person or business behind the content.
So use this rule:
Let AI carry some of the production load. Keep the meaning human.
Use your avatar for:
- recurring tips
- simple FAQs
- educational videos
- content variations
- short story setups
- B-roll featuring you
- simple product explanations
- repeatable creator series
Film yourself when:
- the real emotion matters
- something just happened
- the story is personal
- you are showing proof
- you need to demonstrate a real physical experience
- your natural reaction is the content
AI Avatar vs Real Filming
| Use an AI avatar when | Film yourself when |
|---|---|
| You need faster variations | The moment itself matters |
| You cannot film that day | You are reacting in real time |
| You need another location | The location is part of the proof |
| You are answering routine questions | The story is deeply personal |
| You want recurring educational clips | Trust depends on seeing the real event |
| You are testing ideas | Your spontaneous reaction is the value |
The strongest personal-brand system may use both.
You are not trying to convince people that the digital version is the only version.
You are using it to help your real ideas travel farther.
What Happens to Your Face and Voice Data?
This part deserves more than a quick “yes, it is safe.”
Google says the face and voice recordings used for avatars are stored in your Google Account and used for content generation, safety and identity verification.
Google also says it does not sell those recordings to third parties and does not use the recordings to train the generative AI models powering its services, except to address abuse or harm.
Some face and voice data may count as biometric data under local laws.
Read Google’s current avatar privacy details before creating the avatar.
Flow Data Has a Separate Setting
Google Flow also has a Help improve Google Flow setting.
When that setting is enabled, Google says Flow interactions, uploads, outputs, feature-use information and feedback may be used to improve its products, services and machine-learning technologies.
That is separate from the more specific statement about the face and voice recordings used to create your personal avatar.
Can You Delete the Avatar?
Yes.
Google lets you manage and delete the avatar recording.
Deleting the avatar does not automatically erase videos you already created or published with it. Those need to be removed from the service where they were created.
Can You Use a Google Flow Avatar for Business Content?
Google says its Terms of Service govern commercial use of Flow outputs and that Google does not claim ownership over original content generated through supported services. Its current Flow commercial-use guidance points users back to those terms.
That does not mean every use is automatically cleared for every situation.
If you are creating:
- sponsored content
- client work
- product claims
- ads
- testimonials
- branded partnerships
make sure your script is accurate and follow the rules of the platform where you publish.
Realistic AI-generated content may also trigger platform disclosure rules. Buffer now supports native AI-generated disclosure for several major social networks, which is another sign that disclosure is becoming part of normal publishing workflow. AI disclosure publishing guide
Transparency is especially useful when your audience could reasonably think a synthetic scene really happened.
What Should You Create First?
Do not start by asking your AI clone to parachute into Tokyo while reviewing a moisturizer.
Run three useful tests.
Test 1: Talking Head
Check:
- face
- voice
- mouth movement
- eye contact
- natural expression
Test 2: Lifestyle Scene
Check:
- body movement
- clothing
- environment
- camera framing
- background consistency
Test 3: Product Interaction
Check:
- hands
- object consistency
- natural movement
- timing
Then compare the results.
Ask:
- Which angle looks most like me?
- Which lighting works best?
- Which script sounds most natural?
- Which movements cause problems?
- Which frame should I save?
- What should I keep the same next time?
You will learn more from those three controlled tests than from twenty random generations.
A Simple Weekly Content Plan for Your AI Avatar
Once you know what works, give your avatar a repeatable job.
| Day | Simple content idea | Keep it easy |
|---|---|---|
| Monday | One useful teaching point | One sentence and one example |
| Tuesday | Short lifestyle clip | One location and one action |
| Wednesday | Product or service FAQ | Answer one real customer question |
| Thursday | Personal lesson | Use a story that is actually yours |
| Friday | Quick opinion or myth-buster | One strong point, not a lecture |
Your AI avatar does not need to post seven times a day just because it never gets tired.
You still have to live with the content.
Frequently Asked Questions
Can Google Flow Create an AI Clone of Me?
Yes. In supported regions, Flow can create a personal avatar from your face and voice recordings so you can use a digital version of yourself in AI-generated content.
Why Do I Not See Create Avatar in Google Flow?
Regional availability is one major reason. Avatars are currently unavailable in the EEA, the UK and Switzerland even though Flow itself is available in many of those places. Account eligibility and product access can also affect what you see.
Does the Google Flow Avatar Use My Real Voice?
The setup includes a short voice recording along with facial capture. Google describes the avatar as a digital version of you that can look and sound like you.
Do I Need a Computer to Make the Avatar?
You can start on a computer and continue through a QR code, but the secure face and voice capture requires a phone or tablet. Google also supports mobile Flow apps in eligible regions.
What Does @me Do in Google Flow?
@me tells Flow to use your personal avatar as the reference for the prompt. Add it, then describe the scene, action, dialogue, camera and mood.
Can My Avatar Wear Different Clothes?
Yes. You can describe different clothing in your scene prompts. For better continuity, change major visual details gradually and keep successful references.
Can I Make Vertical Reels, TikToks or Shorts?
Yes, supported Flow video modes currently offer both aspect ratios. Exact clip lengths and reference features depend on the model you select.
How Long Can a Google Flow Avatar Video Be?
Current Flow models support different clip lengths, including 4, 6, 8 and 10 seconds in different modes. Longer stories can be built by combining clips, using supported extensions, or working with Scenebuilder.
How Do I Keep My Avatar Consistent Across Videos?
Reuse the same personal avatar, keep key appearance details stable, save strong frames, use references where supported and avoid changing several major things at the same time.
Can I Use the Same Avatar in Gemini and Flow?
Google says an avatar created for Gemini and Google Flow can be used across those two services. YouTube manages its avatar separately.
Does Google Train Its AI on My Avatar Face and Voice Recordings?
Google says it does not use those Avatar recordings to train the generative AI models powering its services, except to address abuse or harm. Flow interactions and uploads have separate data controls, so review those settings too.
Can I Delete My Avatar Later?
Yes. You can delete the Avatar recording. Videos already created with the avatar do not disappear automatically.
Can I Use My Google Flow Avatar Commercially?
Google says its Terms of Service govern commercial use and that it does not claim ownership over original generated content. Check the current terms plus the rules of the platform where you publish.
Should My AI Avatar Replace Real Filming?
No.
It works best as another way to create, especially for repeatable content, variations and scenes that are hard to film.
Keep real filming for moments where lived experience, trust or emotion is the reason people care.
Your AI Clone Should Save Time, Not Create More Work
The point of a Google Flow AI avatar is not to prove that you can make a digital copy of yourself.
The point is to remove a barrier.
Maybe you have ideas but no energy to set up the camera again.
Maybe you need three versions of the same lesson.
Maybe you want to answer the same customer question in a more visual way.
Maybe you need a scene that would be expensive, awkward or impossible to film today.
Create your avatar once.
Start with one simple scene.
Pay attention to what looks like you and what does not.
Save the good frames.
Build from what works.
And keep enough of the real you in the process that your audience still knows who they are listening to.
That is the sweet spot.
Your AI clone does not need to replace you.
It just needs to make it easier for your real ideas to show up.