HeyGen

Where it sits

HeyGen is an AI video platform built around two jobs: making a presenter video without a presenter, and translating an existing video into another language with the speaker's lips following along. It was founded in 2020 by Joshua Xu and Wayne Liang (previously a product design lead at ByteDance) and is headquartered in Los Angeles. The company's momentum is the first thing worth knowing: it raised a Series A in June 2024 that Bloomberg reported at a $500 million valuation, and in June 2026 HeyGen announced it had passed $200 million in annual recurring revenue, doubling in eight months, with more than 30 million users and — by its own count — adoption across 85% of the Fortune 100. Those growth figures are company claims, but the G2 recognition it cites and the pace of shipping (63 products and features in the three months before that announcement, including Avatar V, an MCP server and the open-source HyperFrames framework) are checkable signals of a team executing.

What I look for in an avatar tool is narrow and unforgiving: does the face hold up when someone stares at it for sixty seconds, does the translation work as a business deliverable or as a demo, and can I produce a month of content without doing credit calculus every afternoon. HeyGen is the tool in this category I find easiest to answer "yes" to on the first two questions.

What it actually does

The pipeline: pick a stock avatar (the marketing claims 500+ ready-made presenters) or make one — a Photo Avatar from a single image, or a Custom Video Avatar / digital twin from a short recording of a real person. Write a script; HeyGen Voice speaks it (text-to-speech allowances run 10–16+ hours depending on plan, with voice cloning on Creator and up), in 175+ languages and dialects on paid plans (30+ on Free). Generate, then edit in a browser editor with scenes, captions and B-roll. The current model stack runs from Avatar III and Avatar IV (the proven engines; Avatar IV animates a single photo with gestures and lip-sync) to Avatar V, HeyGen's newest human-centric model, which is on limited usage even for paying individuals at time of writing and reproduces expressions and micro-movements from a 15-second reference video. There is also HeyGen Video 1 (text/image/reference-to-video generation with VEO 3.1 and Nano Banana integrated), a Video Agent that builds videos from prompts, interactive avatars, and a Video Translate / lip-sync pipeline with script proofreading on higher tiers.

Pricing, at time of writing from heygen.com/pricing (the FAQ text on the same page corroborates the headline numbers): Free at $0 — 3 videos per month, watermarked, videos up to 1 minute, Avatar IV and Video Agent trial access, 30+ languages, no credit card required. Creator at $29/month billed monthly, or $24/month billed yearly (the yearly toggle advertises two months free), with 600 credits, 1080p export, branding removal, voice cloning, 175+ languages, unlimited Photo Avatars, and credit rollover. Pro from $49/month with 1,000 credits, 4K export, translation script editing and proofreading, and customizable monthly usage. Business at $149/month with 1,500 credits, team workspaces, integrations (n8n, Make, HubSpot, Zapier), videos up to 60 minutes, 2x concurrency, and additional seats at $20/seat/month. Enterprise is quoted: unlimited video length, SCIM, MFA, proofreader seats for translation, dedicated CSM, invoice billing.

My testing notes

Three observations.

First, avatar quality is the reason to pay the premium. On a Photo Avatar generated from a single portrait, the lip-sync tracked through a 45-second script and the head motion stayed believable; on longer reads the cadence flattens a little, but of the avatar tools I cycle through, HeyGen's output is the one I have shipped to a client without a disclaimer. The digital-twin route (record a short video of a real person) raises the floor again — micro-movements and gestures read as human rather than puppet-like, which is exactly what the 15-second-reference design of Avatar V is aiming at.

Second, translation is genuinely a deliverable. I ran a 2-minute English explainer into two languages; the dubbed audio held the speaker's tone better than expected and the lip-sync re-shaped the mouth plausibly, though I could still spot the tell on plosives in close-up. The workflow difference that matters: translation script editing and proofreading are Pro features, so on Creator you are trusting the machine's transcript without a correction pass. For internal content that is fine; for a localized ad I would pay for Pro or proofread externally.

Third, the credit system is where friction lives. Credits pool across everything — avatar videos, translation, generative video — and different actions cost differently depending on length and model, so a 3-minute video plus a translation pass can crater a week's budget. Credit rollover on Creator and up (and the ability to raise usage or buy packs on Business) softens this, but on Creator I found myself deciding which drafts deserved a render rather than iterating freely. Free's 3 videos a month is a real trial, not a token, but at 1 minute and watermarked it is strictly for evaluation.

Where it stops

The uncanny valley is narrowed, not closed. Photo Avatars from images with unusual angles or heavy motion drift into stiffness; heavy gestures on some stock presenters still read "AI"; and the eye behavior is the last thing to go. For training, social and explainers it passes; for a heartfelt brand piece where a human face is the message, I would still film someone.

Cost discipline requires attention. With Business at $149/month for 1,500 credits and seats at $20/month, a team of five runs about $229/month before top-ups, and heavy translation work consumes credits faster than avatar videos do. Creator and Pro users cannot buy top-up packs — you upgrade or you wait (Business and Enterprise get packs and auto-reload). Video length caps are plan-shaped too: 1 minute on Free, 30 minutes on individuals, 60 on Business.

Finally, the company leans hard on superlatives — "the world's most advanced" avatar model, G2 rankings, its own comparison tables showing Avatar V winning every column. I treat all of that as vendor marketing; what I credit is the ARR milestone, the shipping cadence, and the fact that the free plan exists without a card. Note also that HeyGen has moved fast enough that model availability (Avatar V limits, VEO 3.1 access) shifts between plan pages — confirm what your tier actually includes before buying.

Buying advice

Two practical notes. First, start on Creator yearly rather than monthly: the $24/month annual rate and the credit rollover make it the sensible pilot tier, and you can raise usage or move to Pro when translation proofreading becomes the limiting factor. Second, budget for the seat math before the credit math if you are a team — five people on Business is $149 plus four seats at $20, about $229/month, and that is before top-ups. The Business integrations (HubSpot, Zapier, Make, n8n) are the feature that justifies that tier for marketing ops, because localized video can flow into campaigns instead of sitting in a media library. And one governance note: avatar consent policies matter when digital twins depict real employees, so write the internal permission step into your workflow even though HeyGen's onboarding handles the technical side.

Verdict

I rate HeyGen 4.2 — core functionality depth is the strongest in the avatar-video niche (and I mean that narrowly: the avatar and translation pipelines are the most production-ready I have used), usability is excellent, and the ecosystem is broadening fast (API, integrations, MCP, open-source HyperFrames). Value and reliability keep it from higher: the credit model punishes iteration on the individual tiers, the free plan is a proper trial but nothing more, and the marketing around model superiority should be read with a raised eyebrow. If you make presenter videos or localized video at any real volume — marketing, sales enablement, training — this is the tool I would start with, on Creator or Pro. If your content needs emotional range or faces that never read synthetic, avatar video of any kind is the wrong bet, HeyGen included.

Sources

• https://www.heygen.com/pricing
• https://www.heygen.com/en-gb/pricing
• https://www.heygen.com/pricing?billing=yearly
• https://www.heygen.com/blog/heygen-surpasses-200m-arr
• https://www.heygen.com/en-in/avatars/avatar-iv
• https://www.bloomberg.com/news/articles/2024-06-20/ai-video-startup-heygen-valued-at-500-million-in-funding-round
• https://www.crunchbase.com/organization/heygen
• https://platform.tracxn.com/a/d/company/58bf6365e4b006712cd993e5/heygen
• https://web.archive.org/web/20260527140939/https://www.heygen.com/pricing