# Make your first HeyGen avatar and check it before you post

Pick a photo avatar or a digital twin, record it once, then render the answer you'd post this week and check it the way a client will see it.

Source page: https://postmosis.com/vault/heygen-avatar. By Leo at Postmosis. Last checked 2026-10-29. Adapted from Keepy Labs: [Your First AI Avatar in HeyGen](https://keepylabs.com/ai-avatars/first-ai-avatar-in-heygen).

In short: A HeyGen avatar is a video version of you, made from a photo or a recording, that reads new scripts without new filming, so your weekly answer to a client's question goes out without a shoot [1]. Make a digital twin if it'll read your videos under your own name, or a photo avatar if you only want to try the idea today [2]. In HeyGen, open Avatars, choose New Avatar, then Upload Photo, or Clone a Real Person to record a 15-second video and a consent clip for a twin on Avatar V [2][3][4][5]. Render a 30 to 60 second test of a video you'd really post, check it on your phone, and switch on the AI label when you post it [6][7].

## Pick one route

Start with the video your clients will see each week: your answer to the question every new client asks, this month's market update, the one mistake you keep fixing. If those go out under your name, you want the avatar to move the way you do, and HeyGen says a video avatar is usually the better option for that [2]. If you only want to try the idea this week, or give a drawn character a voice, a photo avatar is enough. You need one route for your first video, not both [6].

HeyGen avatar routes:

| | What you give HeyGen | What your clients see | Where to start |
| --- | --- | --- | --- |
| Photo avatar | One photo, uploaded or designed with AI | Your photo, animated and lip-synced to your script | Avatars, New Avatar, Upload Photo or Design with AI [2] |
| Digital twin | A recording of you, live or imported, plus a consent clip | You, with the body language, expressions and delivery from your footage | Avatars, New Avatar, Clone a Real Person [3][5] |

The free plan gives you one digital twin slot [3] and up to 3 photo avatars [2], so you can make your first one on the free plan. Paid plans can add more twin slots [8].

The catch with starting on a photo: it won't copy your real gestures the way a twin built from video does [9]. Clients who know you from video calls may notice. If your test looks like a talking passport photo, record the twin.

## One video, from script to post

Here's one avatar video from a normal week of yours, from a client's question to the posted Reel. The example is a career coach; in another trade, swap the question and keep the steps. Each step has its own section further down.

Example: The question. Say you're a career coach. On your discovery calls this month, three people asked you the same thing: 'Should I tell my manager I'm interviewing?' It goes on the list of client questions you keep beside your laptop, and it becomes Tuesday's video.

Example: The script. You type it the way you'd say it across a table: one point and one next step, plus an exception. It's 88 words, about 35 seconds read aloud, which sits inside the 30 to 60 seconds a first test should run.

Example: a career coach's Tuesday script:

```
Should you tell your manager you're interviewing somewhere else?

Most of the time, not yet.

Once your manager knows, you're the person who's leaving, even if you stay. New projects go to someone else. Promotions wait.

So tell them once you've signed an offer.

One exception. If your manager asks you straight out, don't lie. Say you're thinking about your next step, and you'd rather talk it through with them first.

If you're weighing a move right now, book a call with me. The link's in my bio.
```

Example: The setting. You recorded your twin once, back in September, in the spare room you take video calls in. You wore the navy jumper your clients know from those calls, and you use that look every week. You skip the studio backgrounds, because a client who books should see the same room on the call.

Example: The voice. Your own voice clone reads the script. In the preview it trips over your surname, so you double-click it, choose Pronunciation and type it the way it sounds. Then you add your surname to your Brand Glossary, so next week's video gets it right too.

Example: The check. You render it and watch it on your phone, first with the sound off, then with it on. The twin stares silently for two seconds at the end, because you left a blank line after your sign-off. You delete the line and render again. Then you send it to your sister, who says the voice is right and the smile at the start looks like you.

Example: The post. Your caption opens with the question and ends: 'Read by my AI avatar. The answer is mine.' You switch on Add AI label for the Reel and post the same video on LinkedIn with the same line. You don't pin it. The video pinned to your profile is one you filmed on your phone, so a new visitor meets the real you first.

Example: What your clients see. Your own face and voice in the room they know from calls, giving your advice, with a label saying how it was made. You answer the comments yourself that evening, and anyone who books meets you live on the call.

Example: The week it won't work. A script about a salary offer needs you to sound exasperated. The twin reads it evenly and the joke falls flat. You film that one on your phone in four takes and go back to the twin the week after.

After the setup, that's your week: one client question, a script, a test render, a check on your phone and the label. The sections below set it up once, in the same order.

## Which avatar suits your trade

Coaches, consultants, agents, freelancers and lawyers each have a weekly video an avatar can read, and a few it should never touch. Find your row, then pick the route.

Avatar videos by trade:

| | A weekly video it can read | Start with | Keep it out of |
| --- | --- | --- | --- |
| Coach | Your answer to the question from this week's discovery calls | A digital twin, filmed where you take calls | Anything a client told you in a session, and the moments they hire you to be there for  |
| Consultant | One mistake you keep fixing in client work, explained in under a minute | A digital twin, or a photo avatar to test the idea | Client names and results you haven't been cleared to share  |
| Real estate agent | This month's market update for your area, in the same format every week, with one next step | A digital twin | Property tours, which buyers need to see filmed for real, and anything about one buyer or seller [1] |
| Freelancer | How a project with you runs, or one tip from your work | A photo avatar from your portfolio headshot, then a twin once you post weekly | A client testimonial read by an avatar, or a made-up teammate [10] |
| Lawyer | One rule change or one common misunderstanding, in plain words, with 'general information, not advice' on screen | A digital twin, labelled as AI | Advice on anyone's own case, and anything your professional body's advertising rules don't allow [1][11] |

Two rules hold for every row. Anything health, legal or financial still needs real expertise behind the script [1]. And legal, medical and financial work often has its own advertising rules, so check yours before the first post [11].

If the last column of your row is most of what you post, an avatar won't save you much time. Film those posts yourself and give the avatar the one recurring format that's left. Plenty of businesses run it that way, with the avatar on the weekly slot and real footage for launches, events and personal moments [1].

## What to expect from a twin

A twin changes one part of your week. You record once, and after that each new video starts as a script you type [3]. Here's what that looks like in practice.

- What you record: one uninterrupted take of at least 2 minutes [12] and a consent clip under 30 seconds [5]. Set aside a morning for the setup and a retake.
- What it makes: short scripts, explainers about what you sell and any format you repeat [6]. For a coach that's the weekly tip, for an agent the market update, for a lawyer one rule change a week.
- What it copies: whatever was in your footage, coughs and stiffness included [3]. A flat recording gives you a stiff twin [13], and prompts can't add hand movements the clip never showed [9].
- Where it slips: eye line, gesture variety and the way it says names, so check those in every render [9].
- What a client sees: you, saying words you wrote, with the AI label switched on [7]. Say so in the caption too, so nobody feels tricked later.
- What it won't do: live Q&As, client calls or a hard conversation. Those still need you there [1].

Judge the twin after a few weeks of posts, not on the first render. If something looks off, fix one input at a time with short tests [6] before you spend your redo, because you can only re-record once each billing cycle [8].

Example: If you're a property agent, your week with a twin might run like this. On Monday night you type a 45-second script about what sold on your street this month, and the twin renders it while you're at a viewing. On Wednesday it reads your answer to 'how long does a sale take here?'. On Friday you film a real walkthrough on your phone and answer the comments yourself.

## When to film it yourself

An avatar saves you the filming, and that's all it saves. Some videos carry your clients' trust in a way a render can't, and those are the ones to film on your phone.

- The first time a prospect sees you. Someone deciding whether to book will compare the video with you on the call, so pin a video you filmed yourself to the top of your profile and let the first face they meet be the real one.
- A hard conversation with a client, a crisis update or a thank-you that should feel personal: show up in person for those [1].
- Anything live. Q&As and livestreams still need you there [1].
- Advice on one person's situation, which belongs on a call where you can ask questions back. Keep avatar scripts to answers that hold for most of your clients.
- Testimonials and client stories. An avatar should never play a client, and the FTC's rule on fake reviews covers testimonials from someone who doesn't exist, AI-generated ones included [10].
- A week when you've got nothing to say. An avatar makes a generic script faster, but it can't make it worth watching [1].

A rhythm that suits a practice of one is one avatar video and one phone video each week. The avatar keeps the weekly answer going on the weeks you can't film, and the phone video keeps you in the feed as yourself.

And if a client tells you the avatar put them off, believe them. Stop using it for that kind of post and film it instead.

## Other ways to be on video

A twin isn't the only way to post as yourself without filming every week. Weigh it against these before you record anything.

Other ways to put yourself on video:

| | Good for | The catch |
| --- | --- | --- |
| Batch-film real videos | Block a morning, film a dozen scripts back to back and post them over several weeks | Every script change means filming again [9] |
| Clone only your voice | Narration and voiceovers in your own voice, from a voice tool such as ElevenLabs | No face and no lip sync, so the video needs a second tool [9] |
| A personal avatar in Synthesia | A photo or short video of you, plus a live consent recording | The consent step can't be skipped or uploaded later [14] |
| A HeyGen photo avatar | One photo of you, animated to your script | It won't move the way you do as closely as a video avatar [2] |

Batch-filming suits you if you like being on camera and only need a few weeks of posts [9]. A twin makes more sense once you're posting under your own name every week for months.

## Route A: a photo avatar

Start here if you want to try the idea before you record anything. The easiest photo is the one your clients already know, such as your website or LinkedIn headshot, as long as it's clear and front-facing [2]. A portrait made for the job usually animates better, though. Start with 3 to 6 clear photos of yourself [6] and an AI tool that can read images. Run two prompts: the first writes a precise description of your face, the second uses that description and your photos to make the new portrait.

Prompt 1: describe your face:

```
You're a visual identity analyst. Look at the person who appears in all of the photos I've uploaded and write a Facial Identity Report I can reuse in image prompts.

Describe only what you can see. Don't guess ethnicity, nationality, age, personality, health or anything else sensitive.

Cover:
- Face shape
- Skin
- Eyes
- Nose
- Mouth and smile
- Hair
- Facial hair and distinctive features
- Camera and lighting notes
- Ten likeness anchors: the small details that make this face this person

Finish with a line that starts "Portrait Identity Prompt:" followed by one specific paragraph describing the face for future image prompts. Don't mention the photos, don't use vague words like "attractive" or "normal", and leave out clothing, setting and mood unless they're needed to keep the likeness.
```

Prompt 2: make the portrait:

```
Make a realistic, high-resolution photograph of the person described below. Treat the identity details as fixed: keep the exact face shape, eye shape and colour, eyebrows, nose, lip proportions, skin tone and texture, hairline, hair density and hair texture. Natural anatomy and realistic camera rendering.

Portrait Identity Prompt:
[PASTE THE PARAGRAPH FROM PROMPT 1]

The new photo:
[DESCRIBE ONLY THE NEW SCENE: outfit, hairstyle, pose, expression, camera angle, lighting and location. For HeyGen, ask for a front-facing head-and-shoulders shot, face filling most of the frame, relaxed expression.]

Realistic skin detail, natural depth of field and a photographic lens look. No text, logos, watermark, extra people, distorted hands, beauty-filter skin or changed facial proportions.
```

Compare the result with your real photos before you go further [6]. If it looks like your cousin, adjust the description and run the second prompt again.

Check your headshot or the new portrait against HeyGen's advice before you upload it:

- Keep the face close to human proportions, with the eyes and lips clearly visible [2].
- Fill the frame. A character that's too small in the picture is a common reason a photo avatar doesn't work [2].
- Use a clear, front-facing picture with a natural posture for the most natural movement [2].

Then in HeyGen, open the Avatars tab, choose New Avatar, then Upload Photo. Name it, submit it, and wait while HeyGen validates the look [2]. For photo looks, HeyGen uses its Avatar IV engine, which also suits a drawn, 3D or non-human character [4]. If it asks you to pick a voice during setup, skip down to the voice section and do it there; you don't have to make the avatar again [6].

Keep the portrait honest. If it looks better than you do on a call, start again from a plainer one: a client who books after the video should recognise you when you turn your camera on.

Postmosis can also turn one photo of you into a talking video: type up to 30 seconds of script and your photo speaks it, up to 2 a day. [See videos without filming](https://postmosis.com/features/ai-video)

## Route B: record a digital twin

This is the recording every future video of yours comes from, so film it the way you'd want a new client to see you. In HeyGen, open Avatars, choose New Avatar and then Clone a Real Person, and either record live or import your footage [3]. A twin learns from that footage, so the recording decides most of the result: coughs, odd movements and an echoey voice all show up in the avatar [3]. Avatar V, HeyGen's engine for real people, works on video looks only [4].

Set up where your clients already picture you:

- The room you take client calls in often works, as long as it's quiet, the light is even from the front and the camera sits at eye level [6], so your twin looks at your clients, not past them [9].
- Soft, even light with no shadows on your face. HeyGen's guide for its live avatars suggests two diffused lights, one on either side of you [15]; two shaded lamps do the job at home.
- At least 1080p at 30 frames a second. A phone or webcam is fine, landscape or portrait [3].
- What you'd wear to meet a client. Large logos, text and busy patterns cause glitches in the looks HeyGen makes later [12].
- On a phone, tap and hold your face on the screen until the lock icon appears, so focus and exposure stay put [12].

Then record one continuous take. HeyGen's filming tips ask for at least 2 minutes without a cut [12], while its newer Avatar V can learn your movement from a 15-second clip [4]. Give it the 2 minutes if you can.

- Start with a pause of 2 to 3 seconds, and pause 1 to 2 seconds between longer sentences with your lips closed [12].
- Look into the lens, keep your hands below your chest, and turn your head no more than 30 degrees either way [12].
- Hold eye contact with the lens from start to finish, and sit still between gestures: no swaying, no heavy breathing into the mic [15].
- Be more expressive than feels natural. A flat recording gives you a stiff avatar [13].
- Talk out loud even if you plan to use a different voice later, because HeyGen needs your speech to get the lip sync right [3].
- If you stumble, keep talking or start again from the top. Never cut and stitch clips together [3].

You don't need a script. A natural ramble about your work beats a memorised one [9]. If you're worried your mind will go blank, answer these out loud:

What to talk about for two minutes:

```
Talk to the camera as if a new client just asked you these, one after another. Pause with your lips closed between answers.

1. What do you do, and who for?
2. What's the mistake you see most often?
3. Tell me about a week when the work went well.
4. What do people get wrong about your line of work?
5. What would you tell someone before their first session with you?

Smile when it fits, keep your hands below your chest, and keep looking into the lens.
```

Submit the footage with your consent clip (next section). Processing takes about 10 to 20 minutes for footage that isn't 4K [3].

Watch the take back before you submit it, and if you wouldn't send it to a client, record it again now: once it's a twin, a redo means deleting it and uploading new consent and footage, which you can do once each billing cycle [8].

## Record the consent clip

This is the part clients ask about, so it helps to know what it proves. Every video-based twin needs a short consent video before HeyGen builds it [5], so an avatar is only made with the person's knowledge and permission. Record it straight after your footage, in the same room and the same light.

- Read the consent text HeyGen shows you exactly as written, and keep the clip under 30 seconds [5].
- It has to be the same person as the footage, filmed in decent light with a decent camera [5][16].
- Read the four-letter code at the end clearly. If it keeps failing, read the paragraph and the code a little faster [16].
- Don't film a screen playing the statement [5].
- Making a twin of a colleague? They record their own consent, from their phone through a QR code if they're somewhere else, at the same time as the footage [5].

If HeyGen rejects the avatar, the reason shows on the avatar card on the Avatars page [5]. A consent clip that doesn't follow these rules is the most likely cause [5].

The clip also answers a question clients sometimes ask: is this a deepfake? A deepfake uses someone's likeness without permission, while a twin needs a live consent recording from the person themselves [9].

## Set up the voice

Your clients know your voice from calls, so it's the first thing they'd notice if it's wrong.

1. Pick the voice for your avatar; for a first test, any voice from HeyGen's library will do [6].
2. For your own voice, make the standalone voice clone HeyGen offers during setup. It's optional and HeyGen strongly recommends it [4], because it sounds more realistic than the audio from your footage [13].
3. Record at least 30 seconds for the clone [17]. Talk the way you would to a client across a desk, in your own phrases and your own accent, because HeyGen wants a speaking voice rather than a reading voice [17]. Record again until it sounds like you on a good day [13].
4. Preview a short script and listen for pronunciation, accent and pacing [6]. Listen hardest to names, yours, your business's and the places you work in, which HeyGen suggests checking in every render [9].
5. When a word comes out wrong, double-click it in the script, choose Pronunciation and type it the way it sounds [18]. HeyGen's own pages disagree on what happens next: its older script tips say the fix saves to your Brand Glossary by itself [19], while its newer scripts guide says it fixes that one place only [18]. So add the names you say every week, like your surname and your suburb, to the Brand Glossary yourself, where you can type how each sounds or record yourself saying it [20].
6. Write numbers and dates out as words, since voice engines often stumble on them [18], and keep each script in one language, because mixing languages in one script can make the voice unpredictable [18].
7. Confirm the voice before you render, so the test you watch is the voice you'll actually use [6].

If the clone still sounds flat after a re-record, record a short, high-energy sample of yourself and add it as a Custom Emotion [17], or compare HeyGen's voice engines using the same test sentence [17]. A photo avatar has a simpler way out: record the words on your phone and upload the audio file instead of typing a script [21].

## Render a short test

Keep the first video short, so a bad take costs you a few minutes and few credits: Avatar V uses 48 credits per minute, so a short test protects your monthly balance [9]. Make it the video you'd post this week, such as your answer to the question clients ask most, 30 to 60 seconds long and under 100 words [6]. The [script length calculator](https://postmosis.com/tools/speech-time-calculator) tells you how long yours runs.

A first test script:

```
Hi, I'm [name]. I help [who] with [what].

The question I get asked most is [question].

Here's my honest answer: [two or three sentences, the way you'd say it across a table].

If that's where you are, [one next step: book a call, read the guide, send me a message].
```

Fill the brackets with a question you heard this month. If none comes to mind, borrow one from your trade:

A first test question for each trade:

| | The question to answer | The next step to end on |
| --- | --- | --- |
| Coach | Should I tell my manager I'm interviewing? | Book a call  |
| Consultant | Why does the plan from our offsite never happen? | Ask for the one-page checklist  |
| Real estate agent | Is this a good month to sell on [your street or area]? | Message me for this month's sales nearby  |
| Freelancer | Why do I ask for a brief before I quote? | Send me your brief  |
| Lawyer | Do I need a will if I don't own a home? | Get the list of what to bring to a first meeting  |

- Press play beside the script to hear it before you submit, and fix any awkward phrasing there [18].
- Add a pause after the question so it lands. Each pause is half a second, and you can lengthen it [19].
- Trim blank space and line breaks from the end of the script before you generate, so the video doesn't end on dead air [9].
- Ignore a blurry mouth in the preview. HeyGen only renders the lip sync when you submit the video [8].
- Watch the finished test once all the way through before you change anything [6].

If the lip sync or movement still feels off after a shorter test, the script isn't the problem. Go back to the photo or footage the avatar came from [6].

## Check it before you make more

Watch the test on your phone, the way a client will see it, then show it to someone who knows you well. Ask five questions [6]:

- Does the face look like you, straight away, to someone who's met you?
- Does the voice suit the script?
- Do the mouth movements match the words?
- Do the eyes, head and hands move naturally?
- Does the framing suit the place you'll post it?

Then one for you: would a client who meets you next week feel they'd already met you in the video?

Fix one input at a time, and render another short test before you make anything longer [6]. If the person you showed it to says 'that's not how you talk', the fix is in the script, not the avatar. Read it aloud, rewrite the lines you'd never say, and render the test again.

## When it looks off

Each problem below is something a client would spot on their phone. If your test fails one of the questions, find it here, change that one input and render the short test again.

Common problems and fixes:

| | Likely cause | Fix |
| --- | --- | --- |
| It doesn't look like you | The source photo or identity description, or the light in your footage | Compare with your real photos and refine the description, or re-record in better light [6] |
| Stiff, robotic movement | A calm, flat recording | Re-record with more energy, hand movement and expression [13] |
| The hands never move | The footage had no hand movement | Prompts can't add gestures the clip never showed, so record again with your hands in frame [8][9] |
| Lip sync is off | The face is too small in the frame | Use a close-up or half-body shot with your face large, then try a shorter test and check the audio [13][6] |
| The voice doesn't sound like you | Audio taken from the footage | Make a standalone voice clone and record it again until it sounds right [13] |
| The voice is flat | A flat clone recording | Add a short, high-energy sample as a Custom Emotion, or test another voice engine [17] |
| It says your name or suburb wrong | The voice engine guessed the pronunciation | Double-click the word, choose Pronunciation and type it as it sounds, then add it to your Brand Glossary for later videos [18][20] |
| Numbers and dates come out garbled | Voice engines stumble on numeric formats | Write them out as words in the script [18] |
| It ends on a silent stare | Blank space or line breaks after your last line | Trim them and check the timeline length before you generate [9] |
| Looks drift from one to the next | Switching base looks | Use the same base look every time, and leave out hats and glasses unless you want them in every look [13] |
| Avatar rejected | The consent clip doesn't match the footage, or it was too dark | Make a new avatar with the same person in both, filmed in good light [16] |

When the same problem survives two changes, stop patching the render. The cause is usually the photo or footage you started from [6], and for a twin that's the time to spend your one redo for the billing cycle [8].

## Tell clients it's your avatar

Your twin says words you never said on camera, so treat every avatar video as AI content and say so. YouTube asks you to disclose content that makes a real person appear to say or do something they didn't do [22], and Meta requires a label on photorealistic video and realistic-sounding audio made with AI [7]. A label also answers the client who'd wonder, before they have to ask.

- Instagram: when you create the Reel, tap the toggle next to Add AI label [7].
- YouTube: in YouTube Studio, under Attributes, select Yes for AI use. Disclosing alone won't limit a video's audience [22].
- TikTok: label realistic AI people and scenes. Unlabelled videos may be removed, restricted or labelled for you [23].
- LinkedIn: say so whenever it isn't obvious from the post that you relied heavily on AI [24]. A line in the caption covers it.

Put the same line at the end of every caption, so clients learn it once. Three to copy:

Caption lines that say it's your avatar:

```
Read by my AI avatar. The answer is mine.

My digital twin, made from a recording of me. I wrote every word.

AI avatar, my words: three clients asked me this on calls this week, so here's my answer.
```

A label covers the face, not the content. LinkedIn says posts that feel generic or lack a clear point of view are less likely to be widely distributed [24], and you're responsible for everything you post there [24]. Read each script aloud before you render, and cut any line you wouldn't say to a client.

## Looks, redos and reuse

Once the twin works, you can make more looks of it with different backgrounds, outfits and angles, up to 500 per avatar [25]. Keep your whole face visible in each one [25]. For client videos, keep one look that matches the room clients see on calls, and save the others for other formats.

Be careful with redos. Re-recording means deleting the avatar from its slot and uploading new consent and footage, and you can only do that once each billing cycle [8]. Deleting an avatar also removes all of its looks [3].

HeyGen lets you use the videos on any platform, business use included [8]. The [AI avatar guide](https://postmosis.com/vault/ai-avatar) covers which jobs an avatar suits beyond the weekly answer.

From here, reuse the avatar and voice for short scripts, explainers about your services and any format you repeat [6]. To keep the same face in other image and video tools, build a reference sheet with the [consistent AI character method](https://postmosis.com/vault/consistent-ai-character).

Postmosis edits the HeyGen render in the browser: captions in 6 styles, B-roll placed for you, and 9:16, 4:5, 1:1 and 16:9 exports in one run. [See the video editor](https://postmosis.com/features/video-editor)

## Your first month of posts

Plan the first month before you make a second video: one avatar video a week, taken from the questions clients ask on calls, and one video you film yourself.

A first month with your twin:

| | Avatar video | Filmed by you |
| --- | --- | --- |
| Week one | The question every new client asks | A short hello from your desk, pinned to your profile  |
| Week two | A mistake you see clients make, and the fix | Your reply to the best comment from week one  |
| Week three | One thing people get wrong about your line of work | A minute from your day, such as a viewing or a client visit  |
| Week four | What happens in a first meeting with you | What you learned this month, in your own words  |

Keep the days fixed, such as Tuesday for the avatar and Friday for the phone video, so the plan survives a busy week. At the end of the month, look at which posts brought comments and calls, and write next month's scripts from those questions.

If a week runs out, the avatar video still goes out, because a typed script fits a busy evening. Let the phone video slip a week rather than the whole plan.

Put both videos on the Postmosis content calendar and get a phone alert when each one is due, or schedule them and Postmosis posts them for you once your accounts are connected. [See scheduling](https://postmosis.com/features/scheduling)

## Common questions

### What is a HeyGen digital twin?

A personal AI avatar built from a short video of you that copies your body language, expressions and delivery [3]. Once it's trained, you type a script and the twin presents it, in your cloned voice if you made one.

### How long does a HeyGen avatar take to make?

The recording and consent clip take a few minutes. After you submit, HeyGen processes a twin in about 10 to 20 minutes for footage that isn't 4K [3]. A photo avatar is ready once HeyGen has validated the look [2].

### Can I make a HeyGen avatar for free?

Yes. The free plan includes one digital twin slot [3] and up to 3 photo avatars [2]. Paid plans can add more twin slots [8], which matters if you want avatars of colleagues too.

### What is Avatar IV in HeyGen?

The engine HeyGen uses for photo-based avatars. It reads the tone, rhythm and emotion in the voice and turns them into facial movement [21]. It's the better pick for cartoon, 3D and non-human characters, while Avatar V is for real people on video looks [4].

### Why is my HeyGen avatar's mouth blurry?

You're looking at the preview. HeyGen only generates the lip sync when you submit the video for rendering, so the mouth is clear in the finished file [8].

### Should I tell clients my video is an AI avatar?

Yes, every time. Instagram requires a label on realistic AI video and voice [7], and YouTube asks you to disclose a real person appearing to say something they didn't [22]. Switch on the label and end the caption with a line such as 'Read by my AI avatar. The answer is mine.'

### Can I make a HeyGen avatar of someone else?

Only with them. The person in the avatar records their own consent video at the same time as the footage, though they can do it from their phone through a QR code [5]. HeyGen doesn't allow using someone's videos without permission [8].

### Can I use HeyGen videos for my business?

Yes. HeyGen says you can use and share the videos on any platform, including for commercial purposes, as long as your inputs don't break copyright [8]. Switch on the AI label wherever the platform asks for one [7].

## Sources

1. [How businesses use AI avatars, Keepy Labs](https://keepylabs.com/ai-avatars/ai-avatars-for-business), read 2026-10-29
2. [How to Get Started with Photo Avatars, HeyGen Help Center](https://help.heygen.com/en/articles/10034438-how-to-get-started-with-photo-avatars), read 2026-10-29
3. [Create your first Digital Twin with Avatar IV, HeyGen Help Center](https://help.heygen.com/en/articles/12089286-create-your-first-digital-twin-video-avatar-with-avatar-iv), read 2026-10-29
4. [Avatar V is now available on HeyGen, HeyGen Help Center](https://help.heygen.com/en/articles/14602974-avatar-v-is-now-available-on-heygen), read 2026-10-29
5. [Recording your Consent Video, HeyGen Help Center](https://help.heygen.com/en/articles/12092609-recording-your-consent-video), read 2026-10-29
6. [Your first AI avatar in HeyGen, Keepy Labs](https://keepylabs.com/ai-avatars/first-ai-avatar-in-heygen), read 2026-10-29
7. [Label AI content on Instagram, Instagram Help Centre](https://help.instagram.com/761121959519495), read 2026-10-29
8. [Digital Twin FAQ, HeyGen Help Center](https://help.heygen.com/en/articles/9380615-digital-twin-faq), read 2026-10-29
9. [What Is a Digital Twin Avatar? How to Create One, HeyGen](https://www.heygen.com/blog/how-to-create-digital-twin-avatar), read 2026-10-29
10. [Federal Trade Commission announces final rule banning fake reviews and testimonials, FTC](https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials), read 2026-10-29
11. [The AI character brief, Keepy Labs](https://keepylabs.com/ai-avatars/the-ai-character-brief), read 2026-10-29
12. [Digital Twin (Video Avatar): Filming Tips, HeyGen Help Center](https://help.heygen.com/en/articles/8389138-digital-twin-video-avatar-filming-tips), read 2026-10-29
13. [How to get the best results with Avatar V in HeyGen, HeyGen Help Center](https://help.heygen.com/en/articles/14602997-how-to-get-the-best-results-with-avatar-v-in-heygen), read 2026-10-29
14. [AI avatars, Synthesia](https://www.synthesia.io/features/avatars), read 2026-10-29
15. [LiveAvatar: Custom LiveAvatar Creation Guide, HeyGen Help Center](https://help.heygen.com/en/articles/9612935-liveavatar-custom-liveavatar-creation-guide), read 2026-10-29
16. [Digital Twin Creation Error Codes, HeyGen Help Center](https://help.heygen.com/en/articles/9824740-digital-twin-creation-error-codes), read 2026-10-29
17. [Improve Your Voice: How to Get the Best and Most Consistent Results, HeyGen Help Center](https://help.heygen.com/en/articles/14010178-improve-your-voice-how-to-get-the-best-and-most-consistent-results), read 2026-10-29
18. [How to Write Scripts in the AI Studio, HeyGen Help Center](https://help.heygen.com/en/articles/11381771-how-to-write-scripts-in-the-ai-studio), read 2026-10-29
19. [Script Tips, HeyGen Help Center](https://help.heygen.com/en/articles/9574152-script-tips), read 2026-10-29
20. [How to Use Brand Glossary, HeyGen Help Center](https://help.heygen.com/en/articles/8830251-how-to-use-brand-glossary), read 2026-10-29
21. [HeyGen Avatar IV Complete Guide, HeyGen Help Center](https://help.heygen.com/en/articles/11269603-heygen-avatar-iv-complete-guide), read 2026-10-29
22. [Disclosing use of GenAI content, YouTube Help](https://support.google.com/youtube/answer/14328491?hl=en), read 2026-10-29
23. [Integrity and Authenticity, TikTok Community Guidelines](https://www.tiktok.com/community-guidelines/en/integrity-authenticity), read 2026-10-29
24. [Best practices for content created with the help of AI, LinkedIn Help](https://www.linkedin.com/help/linkedin/answer/a1481496), read 2026-10-29
25. [Avatar Looks Explained, HeyGen Help Center](https://help.heygen.com/en/articles/9964694-avatar-looks-explained), read 2026-10-29