HeyGen AI Avatar Guide: Photo, Digital Twin and Avatar V Costs

HeyGen gives you four ways to put a presenter on screen: a stock avatar, a Photo Avatar animated from one image, a Digital Twin trained on your own footage, or a character designed from a text prompt. For most business videos, the Digital Twin is the one worth setting up. It needs about 15 seconds of continuous video plus an on-camera consent clip, and it trains on Avatar V in roughly 10 to 20 minutes. The real limit is credits. Avatar V uses 48 credits a minute of output,^[1] so Pro's 1,000 monthly credits cover about 20 minutes of twin video, by our arithmetic.

HeyGen's pricing page lists Creator at $29 a month billed monthly or $24 a month billed yearly ($288 a year), with 600 credits a month, one user and 1080p export.^[2] Credits, not the plan price, decide how much avatar video you get, so read the engine rates in our HeyGen pricing guide before you pick a tier. The Free plan lets you build one twin and three Photo Avatars, but its videos are watermarked, capped at one minute and licensed for personal, non-commercial use only.

Pick the avatar type before you pick a plan

A Photo Avatar runs on Avatar IV by default and can animate cartoons, pets and angled poses. Faces with human-like proportions and a visible mouth lip sync best; beaked, masked or very stylized faces usually do not. Photo Avatars carry no consent check. HeyGen keeps no consent record for them, so getting and keeping the subject's permission is your responsibility.^[3] That matters the day a marketer animates a colleague's headshot without asking. Paid plans place no cap on how many you create, but every video rendered with one draws credits.

Stock avatars are the quickest start, and each ships with several looks: outfits, poses and backgrounds. AI-designed characters suit mascots or fictional hosts, and with no real person shown, no consent step applies. To keep generated looks of yourself consistent, paid plans offer a Personal Model, a dedicated image model of your appearance. Training it costs 60 credits, needs at least 10 photos (30 or more recommended) and takes about 10 to 15 minutes. You can retrain it any time.

Avatar III, IV or V: the engine sets the bill

You choose the engine per video, and it moves cost more than anything else. Avatar V is the highest-fidelity engine and the default for Digital Twins. It needs a video-based look and does not run on public Expressive avatars or some older studio avatars. Avatar IV is the default for Photo Avatars and can animate any single image. Avatar III loops your footage and adds lip sync. It is the cheapest. Creator includes Avatar V with limited usage, so check what your plan shows.

Avatar III costs 4 credits a minute with a video look or 7 with a photo look. Avatar IV costs 16 with a photo look or 31 with a video look. Avatar V costs 48 credits a minute. By our arithmetic, Creator's 600 credits buy about 12 minutes on Avatar V (within its limited Avatar V usage), about 19 minutes on Avatar IV with a video look, about 37 minutes on Avatar IV with a photo look, or 150 minutes on Avatar III with a video look. In AI Studio, Avatar IV charges only for the seconds of avatar generation, not the whole video.

One habit saves real money. Lip sync renders only when you submit; the editor preview plays the audio, not the moving avatar. Every lip-sync check is a render. HeyGen itself suggests Avatar III for testing scripts before finalizing on Avatar V. Two drafts of a two-minute script cost 16 credits on Avatar III with a video look, against 192 credits on Avatar V, by our arithmetic. Cancelled or failed videos are never charged, and layout, caption and timing edits in AI Studio cost nothing.

Building a Digital Twin, step by step

Allow camera and microphone access in your browser, then film one continuous take with your face clearly visible; HeyGen recommends at least 1080p, and a laptop webcam works. Speak during the clip even if you plan to use another voice later. Footage without audio is rejected, because the audio drives lip sync. Avoid shirts with large logos, text or busy patterns, which cause glitches when new looks are generated.

1. Open Avatars, choose New Avatar, then Clone a real person.
2. Record about 15 seconds facing the camera at eye level, scan the QR code to film on your phone, or import footage.
3. Record the consent clip: read HeyGen's text and a four-letter code on camera, in under about 30 seconds.
4. Add a standalone voice clone, which HeyGen calls optional but strongly recommended; our voice cloning guide covers the options.
5. Wait about 10 to 20 minutes for training.

The consent clip must show the same person as the footage.^[4] Screen recordings, AI-generated clips, a misread code or dark, blurry video fail, and the reason appears on the avatar card. For a colleague's twin, that person records consent on their own phone through a QR code. Uploading a pre-recorded consent video is Enterprise-only, and skipping consent needs an Enterprise indemnity agreement. Footage with nudity or minors is rejected, and three failed attempts can bring an avatar ban that only support lifts.

The person depicted can ask HeyGen to remove their likeness at any time. Removing an avatar does not delete videos already made, and it leaves the training footage, consent video and cloned voice in place; each is deleted separately. Note one gap. Professional Voice Clone uploads have no consent step at all. HeyGen's Terms grant it a licence to use your content, including to train its models; its Trust page offers non-Enterprise users an opt-out, and its biometric notice says face data is used for training only with your separate consent. Our HeyGen safety review lays out those layers.

Slots, looks and the once-a-month redo

A slot holds one person with up to 500 looks, photo and video combined. Creator and Pro include one or five slots, depending on region and account creation date; Business includes five, or ten in some regions. Motion comes from the video, the look from a photo and the voice from audio, so record a strong motion clip once and add outfits as new looks. For Avatar V, keep the same base look photo, since switching mid-project causes slight inconsistency across your looks.

HeyGen's pricing page lists an extra Custom Video Avatar slot at $29 a month or $288 a year. Slot pricing can vary by account age and region, and credits, not the slot, still decide how much twin video you render; our pricing guide has the full add-on list. When a plan ends, slots drop to one and extra avatars are locked, not deleted, until you resubscribe. When a slot add-on lapses, that twin expires, but videos already rendered stay untouched.

Each avatar can be redone once a month by deleting it and recording new consent and footage. The reset comes at renewal on monthly plans and on the 1st of each month on annual plans. Deleting an avatar permanently removes all its looks. Choose carefully. If looks drift, HeyGen suggests generating 50, keeping the 10 most consistent and retraining the avatar on those. A face that fills too little of the frame also degrades lip sync, so favor close-up or half-body shots.

Directing motion and putting several avatars on screen

Avatar IV takes free-text motion prompts, while Avatar V offers preset Expression, Gesture and Gaze controls and weighs audio and image expression above your prompt. Custom motion lasts at most 10 seconds before it loops or holds. It cannot move the camera, walk, add props or change scenes, and a twin's gestures are fixed by its training footage. Custom Expressive Motion, on Avatar IV only, adds 40 credits a minute on top of the render. Keep motion prompts short and positive.

AI Studio puts one character in each scene. To show up to three avatars together, Avatar Shots places them in a Seedance 2.0 clip of 5, 10 or 15 seconds. It costs 60 credits per clip at 720p or 150 at 1080p, whatever the length, with no 4K. Shots accept only video-based real-human looks and no uploaded audio. They cannot be edited afterward.

LiveAvatar is a separate product

Real-time, conversational avatars are not in your HeyGen plan. LiveAvatar, called Interactive Avatar until November 2025, runs as its own platform; you sign in with your HeyGen account, but plans, credits, slots and billing are separate, and HeyGen avatars do not carry over. A custom LiveAvatar needs at least two minutes of continuous, chest-up 1080p footage: 15 seconds of silent listening, 90 of talking and 15 of silent idling. Full-body and 4K footage are not supported, processing takes about 24 hours, and a website embed requires coding.

The bottom line

Start with the avatar type, not the plan. A Digital Twin from 15 seconds of footage and a consent clip is the strongest option for a recurring presenter, and Avatar V gives it the best fidelity at 48 credits a minute. That rate is the number to plan around: draft on Avatar III, render finals on Avatar V, and add looks rather than new twins. Animate other people's photos only with their written permission. Budget live, conversational avatars as a second product.