Last Updated: 28 August 2026

Executive Summary

HeyGen turns a fifteen-second recording into a digital twin that can deliver any script you write, in any of 175-plus languages, without filming again. Avatar V, launched June 2026, is the most convincing version yet. What separates HeyGen from most AI video tools is not the output — it is the governance underneath: mandatory verified consent for any real person’s likeness, biometric matching to confirm the consenting person is the one depicted, and SOC 2 Type II certification. The catch is credits. “Unlimited videos” is real, but the good avatar models are metered, and edits consume credits too.

Estimated reading time: 10 minutes

Key Takeaways

• Avatar V, released 5 June 2026, builds a persistent digital identity from a 15-second reference video and applies your motion to any photo of yourself.
• Creating an avatar of a real person requires verified consent, including biometric face-geometry matching that confirms the consenting person is the one depicted.
• “Unlimited videos” applies to standard output. Avatar IV and V cost 20 credits per minute, so a 600-credit Creator plan buys roughly 30 minutes of premium avatar video.
• Users consistently report that edits also consume credits, which is the most common complaint about an otherwise well-regarded product.
• HeyGen holds SOC 2 Type II certification and GDPR and CCPA alignment, with enterprise customer data excluded from model training.
• The Team plan was retired in January 2026 and replaced by Business at $149 a month plus $20 per additional seat.

What Is HeyGen?

HeyGen is an AI video platform built on a single premise: you should not need to film yourself every time you produce a video. Record once, and a digital twin delivers every future script you write.

The product range covers Video Avatar (clone yourself from a recording), Photo Avatar (animate a still image), Video Agent (generate a structured video from a text prompt), Video Translator (translate and lip-sync existing footage into other languages), and AI Studio, a scene-by-scene editor for assembling everything.

The avatar engines are where the real progress has happened. Avatar IV animates a single photo using a diffusion-inspired audio-to-expression system that interprets vocal tone, rhythm and emotion rather than simply syncing lips — and it handles cartoons, mascots and non-human faces well. Avatar V, released 5 June 2026, goes further: it builds a fine-tuned model from a short reference video, capturing your specific gestures and mannerisms, then applies that motion to any photo of you. Change outfit, setting or angle; the movement stays unmistakably yours.

HeyGen’s own guidance is refreshingly specific about which to use. Avatar V performs better for real human avatars and video-based sources; Avatar IV is better for virtual characters, 2D and 3D avatars, and photo-based creation.

2026 also brought LiveAvatar and interactive avatars, HyperFrames, an Android app, Seedance 2.0 integration, B-roll generation via Sora 2 and Veo 3.1, and LMS integration with SCORM export.

Why this matters for your business: Most AI video tools are judged on how convincing the output looks. For avatar tools, that is only half the assessment. The other half is whether the platform can prove the person on screen agreed to be there — and that is where HeyGen has invested most visibly.

Who Is It Best For?

HeyGen is a strong fit for:

• Learning and development teams. SCORM export, LMS integration and interactive video make it a genuine training platform rather than a video generator with training use cases.
• Businesses localising content. Video Translator with lip-sync across 175-plus languages replaces re-recording, re-casting and dubbing entirely.
• Founders and executives scaling presence. Record once, then produce updates, announcements and customer messages without booking studio time repeatedly.
• Marketing teams producing volume. Product explainers, ad variations and social content at a fraction of filming cost.
• Organisations that need governance. Verified consent, biometric verification, SOC 2 Type II and enterprise controls make this approvable where competitors are not.

HeyGen is probably not the right choice for:

• Teams needing cinematic or narrative video, where dedicated generative video tools are stronger.
• Heavy iterators on lower tiers, since edits consume credits and premium avatars burn 20 per minute.
• Anyone wanting to animate a real person without their participation — the consent flow is a deliberate obstacle, not a bug.
• Businesses wanting flat, predictable pricing rather than a credit system.
• Political campaigning or advocacy, which HeyGen’s acceptable use policy prohibits outright.

Key Features

Avatar V. The current flagship. A 15-second reference video creates a persistent digital identity capturing your gestures and expressions, which can then animate any photo of you with multi-angle stability and long-form consistency.

Avatar IV. Animates a single photo using an audio-to-expression engine that reads tone and rhythm to generate matching micro-expressions. Best for cartoons, mascots, animals and stylised characters.

Video Translator. Translates existing footage into 175-plus languages with lip-sync, preserving accent characteristics. The single most commercially valuable feature for any business operating across markets.

Video Agent. Generates a complete structured video from a text prompt, selecting avatar, script structure and B-roll — including footage from integrated models.

LiveAvatar and interactive avatars. Real-time conversational avatars for support, sales and interactive learning rather than pre-rendered clips.

AI Studio. A scene-by-scene editor for assembling avatar segments, B-roll, screen recordings and branding into a finished video.

LMS integration and SCORM export. Direct pipeline into corporate learning systems, which is why HeyGen shows up in L&D procurement rather than only marketing budgets.

API with pay-as-you-go pricing. Programmatic access starting at $5, including a documented consent endpoint for teams building avatar creation into their own products.

Where the credits actually go

UNLIMITED (paid plans) ──► standard video output
   │                        Avatar III · dubbing · stock content
   ▼
METERED (credits) ────────► the models you actually want
   │
   ├─ Avatar III ──────────► 3 credits/min
   │                         600 credits ≈ 200 minutes
   │
   ├─ Avatar IV / V ───────► 20 credits/min
   │                         600 credits ≈ 30 minutes
   │
   └─ Edits and re-renders ► also consume credits
                             ⚠ the most common complaint

[Illustration placeholder: HeyGen credit consumption map]

Ease of Use

HeyGen is among the most approachable tools in this category. Creating a digital twin takes a short recording session — read a provided script to camera, and the engine captures motion, expression and voice. After that, producing a video means typing a script and clicking generate.

Photo Avatar is simpler still: upload an image, paste a script, choose a voice and motion style, render. Most videos complete in five to ten minutes.

The friction is entirely commercial rather than technical. Paid plans advertise unlimited video creation, and that is true for standard output — but the avatar models people actually want are metered. Reviewers report hitting limits unexpectedly, with one noting they exceeded their allocation in two days because every action, including edits, draws down credits.

Future Signal Tip: Work out your monthly minutes at 20 credits per minute for Avatar IV or V, then add at least 30% for re-renders and edits. A team publishing twenty three-minute videos monthly needs roughly 1,200 credits before revisions — which is above the Creator allocation and into Pro territory.

Performance

Avatar quality is genuinely strong and has improved sharply. Avatar V’s identity consistency across angles and long-form delivery is the clearest advance, and Avatar IV’s handling of non-human and stylised faces is better than most competitors attempt.

The translation feature is the standout commercial capability. Lip-synced translation across 175-plus languages, preserving accent characteristics, replaces a workflow that previously involved translators, voice actors and re-editing. For businesses operating across markets, that is a categorical change rather than an incremental saving.

Limitations are honest ones. Custom avatars can show minor inconsistencies at head angles poorly represented in the training footage, and HeyGen’s own guidance notes stock avatars — trained under controlled studio conditions — can look slightly more polished than a self-recorded twin. This is also not a cinematic video tool; it produces people talking to camera extremely well, and everything else adequately.

The Consent Question

This section exists because it is HeyGen’s clearest differentiator, and because avatar tools are where AI’s consent problem is most acute.

Creating a digital twin of a real person requires verified consent. The depicted individual completes a consent flow — typically an interactive webcam recording with a spoken passcode, valid for 24 hours — and HeyGen processes face geometry to confirm the person consenting is the person in the submitted footage. That biometric comparison exists specifically to detect and prevent deepfakes and impersonation, and HeyGen states the verification data is deleted promptly, typically within minutes.

Notably, consent applies only to digital twins. Photo avatars and prompt-generated characters do not require it, because they depict no real identifiable person — a sensible line rather than blanket friction.

Enterprise accounts can be whitelisted to submit pre-recorded consent videos rather than running the live flow, which matters for organisations creating avatars at scale.

Beyond consent, HeyGen operates a two-stage moderation pipeline — automated machine-learning review followed by human moderators — covering fraud, harassment, child safety, misinformation and IP infringement. It prohibits political campaigning, deceptive, hateful and sexually explicit content, and honours removal requests from depicted individuals.

⚠️ One nuance worth reading carefully. HeyGen’s marketing states that user data is not used for model training. Its biometric privacy notice is more specific: for non-enterprise users who have granted consent, videos and associated face geometry may be used to train and refine its models, with consent withdrawable by email. Enterprise customers are excluded from this. If your organisation cannot permit likeness data in a training set, that distinction matters — check which tier you are on.

Integrations

HeyGen’s integrations are unusually well aimed at where corporate video actually lives. LMS integration with SCORM export puts it directly into learning management systems, which is why it competes for training budgets rather than only marketing spend.

The API offers pay-as-you-go access from $5, with documented endpoints including a consent flow for teams building avatar creation into their own products — a genuinely thoughtful piece of developer design.

On the model side, HeyGen has integrated third-party generative video for B-roll, including Sora 2 and Veo 3.1, plus Seedance 2.0. One practical note: OpenAI is retiring the Sora API in September 2026, so any B-roll workflow depending on that specific model is worth reviewing.

Mobile apps cover both major platforms, with Android added in May 2026.

Pricing & Value

HeyGen restructured its plans in January 2026, retiring the Team tier and replacing it with Business.

Two things determine your real cost. First, which avatar model you use: Avatar III runs 3 credits per minute, Avatar IV and V run 20. The same 600 credits buy either 200 minutes or 30. Second, iteration — edits and re-renders draw from the same pool, which is the source of most user frustration.

Top-up credits are available at $0.05 each, but only on Business. Creator and Pro users who run out must upgrade rather than buy a pack, which is a meaningful structural constraint.

Reported figures vary between sources, particularly for Creator credits and the Pro tier, which appears to scale across several price points up toward $4,300 a month for very high volume. Verify at checkout.

Is it good value? For video translation alone, almost certainly — the alternative is translators and voice actors. For talking-head content at volume, $29 to $149 compares well against filming. For teams that iterate heavily, model the credit burn honestly first, because the headline “unlimited” will mislead you.

What would you spend filming and dubbing the same content? For most businesses that comparison, rather than the competitor comparison, settles it.

Strengths

• Best-in-class consent and verification infrastructure, including biometric matching to prevent impersonation.
• Avatar V delivers convincing digital twins from a 15-second reference video.
• Video translation with lip-sync across 175-plus languages replaces an entire production workflow.
• SOC 2 Type II, GDPR and CCPA alignment, with enterprise data excluded from training.
• Genuine L&D credentials through SCORM export and LMS integration.

Limitations

• “Unlimited videos” excludes the premium avatar models most users want.
• Edits and re-renders consume credits, the most consistent user complaint.
• Top-up credits are available only on Business; lower tiers must upgrade instead.
• Non-enterprise users who consent may have likeness data used for model training.
• Custom avatars can be less polished than studio-trained stock avatars.
• Not suited to cinematic or narrative video work.

Alternatives

Synthesia — the closest competitor, with a longer enterprise track record and comparable governance, generally at higher cost.

Kling or Veo — better for cinematic and narrative video, but without avatar consistency, consent infrastructure or translation workflow.

ElevenLabs — the stronger choice if the requirement is voice and dubbing rather than on-screen presence.

Filming with a phone — worth naming honestly. For a small volume of high-trust content, actually recording yourself remains cheaper and more authentic.

The Future Signal

HeyGen has made a bet that most AI video companies have not: that the constraint on synthetic humans will be permission rather than realism.

The evidence is in what it built. Biometric verification that the consenting person matches the footage. A consent endpoint in the public API. Tiered consent levels for enterprises operating at scale. Removal rights for depicted individuals. None of that improves output quality, and all of it adds friction — which is exactly why it signals where the company thinks the market is going.

That looks well judged. Regulation around synthetic likeness is tightening, several jurisdictions are legislating on digital replicas, and the reputational cost of an unauthorised avatar is severe. A platform that can demonstrate consent has an advantage that better rendering will not erase.

The tension is commercial rather than ethical. Consent infrastructure is expensive to build and impossible to market as excitingly as a new avatar model, while competitors with looser controls can move faster and price lower. Whether governance-first proves to be a durable moat or an expensive handicap is the question worth watching over the next two years.

Final Verdict

Overall Rating: 8.4 / 10

HeyGen is the strongest choice available for avatar-led business video, combining genuinely convincing digital twins with translation capability that replaces whole production workflows. Its consent and verification infrastructure is the best in the category and the main reason it clears procurement where rivals do not.

It loses points on cost clarity. “Unlimited videos” is technically accurate and practically misleading, since premium avatars are metered at 20 credits per minute and edits draw from the same pool. Top-ups being Business-only compounds that.

Future Signal recommends HeyGen for L&D teams, businesses localising content across markets, and executives scaling presence without filming. Model your credit burn at real settings before choosing a tier.

Before you subscribe — a short checklist

• Calculate monthly minutes at 20 credits per minute for Avatar IV or V
• Add at least 30% headroom for edits and re-renders — they consume credits
• Check whether your volume needs Business, since only it allows top-ups
• Use Avatar V for real people, Avatar IV for stylised or non-human avatars
• Confirm the consent flow works for your subjects before promising delivery
• If likeness data cannot enter a training set, check your tier’s terms
• Review any B-roll workflow using Sora, which is being retired in September
• Compare total cost against what filming and dubbing would actually cost

Frequently Asked Questions

What is the difference between Avatar IV and Avatar V? Avatar V, released June 2026, builds a persistent digital identity from a 15-second reference video and works best for real human avatars. Avatar IV animates from a single photo and handles virtual, 2D, 3D and non-human characters better. Both cost 20 credits per minute.

Does “unlimited videos” really mean unlimited? For standard output, yes. Premium features — including Avatar IV and V, video translation with lip-sync, and AI-generated looks — draw on a separate credit pool. Edits and re-renders also consume credits, which is the most common source of user frustration.

How many videos do the credits actually buy? It depends entirely on the model. Avatar III costs 3 credits per minute; Avatar IV and V cost 20. A 600-credit Creator plan therefore buys roughly 200 minutes of Avatar III or about 30 minutes of Avatar IV or V.

Can I make an avatar of someone else? Only with their verified consent. The depicted person must complete HeyGen’s consent flow, and the platform uses biometric face-geometry matching to confirm the consenting individual is the one in the footage. Photo avatars and prompt-generated characters do not require consent, as they depict no real identifiable person.

Is HeyGen secure enough for enterprise use? It holds SOC 2 Type II certification with GDPR and CCPA alignment, encryption in transit and at rest, and role-based access controls. Enterprise adds SSO, SCIM provisioning and multi-workspace administration. Enterprise customer data is excluded from model training.

Does HeyGen train on my videos? Enterprise customers are excluded. For other users, HeyGen’s biometric privacy notice states that where consent has been granted, videos and associated face geometry may be used to train and refine its models, with consent withdrawable by email. Check your tier and settings if that matters to you.

How good is the video translation? It is arguably the platform’s strongest commercial feature — translation with lip-sync across 175-plus languages that preserves accent characteristics, replacing translators, voice actors and re-editing for businesses operating across markets.

What happened to the Team plan? HeyGen retired it in January 2026 and replaced it with Business, starting at $149 a month plus $20 per additional seat, which adds shared workspaces, brand kit management, SCORM export and LMS integration.

Conclusion

HeyGen’s strength is that it solved two problems rather than one. The obvious problem was making a synthetic person look convincing, and Avatar V does that as well as anything available. The less obvious problem was proving that person agreed to appear — and HeyGen built biometric verification, tiered consent, API-level consent handling and removal rights to address it.

That governance is why HeyGen clears procurement reviews that competitors fail, and it is the main reason to choose it for any business use involving a real person’s likeness.

Its weakness is pricing honesty. “Unlimited videos” is true of standard output and untrue of the models anyone actually wants, and the fact that edits consume credits turns iteration into a budget decision. Neither is disqualifying, but both mean you should model your real credit burn before subscribing rather than trusting the headline.

For training teams, multilingual businesses and anyone who needs to be on camera more often than they can realistically film, HeyGen is a straightforward recommendation — provided you buy the tier your actual usage requires rather than the one the marketing implies.