Generative media · USA (Los Angeles; team with roots in China)
HeyGen
An avatar and video-translation platform says it reached hundreds of millions in annual revenue on little capital by letting its output advertise itself and keeping every transaction profitable; retention data is still missing.
Verdict: Close to a real reading
Updated: 2026-10-06
01
Profile
- What it does
- AI avatar and video generation: video translation and dubbing, talking avatars, API and MCP. The founder's three words: "create, localize, personalize" [4]
- Founded
- 2020; first as Surreal, then Movio, then HeyGen (secondary data) [6]. App launch in September 2022 (secondary data) [6]
- Founders
- Joshua Xu (former Snap engineer) and Wayne Liang (secondary data) [6]
- Pricing model
- Credit model: Free $0 (3 videos, 1 minute), Creator $29/month (600 credits), Pro $49/month (1,000 credits), Business $149/month + $20/seat (1,500 credits), Enterprise on request. "Credits are HeyGen's single currency for video and asset creation" [3]
02
Growth timeline
2023-05
$1M ARR in 7 months
"They recently crossed an incredible milestone: $1m ARR in 7 months"
[5]2024-08
$35M annual revenue, more than 40,000 business customers (podcast host's intro)
"serving more than 40,000 businesses"
[4]2025-10
$100M ARR (Upstarts interview: doubled from $100M eight months prior)
[2]2026-06-25
$200M ARR, doubled in 8 months, 30M+ users, 118M+ videos
"People don't want more AI slop."
[1]2026-06-25
$1.5M ARR per employee; $25M spent out of $74M raised (Upstarts interview)
"We've been very constrained."
[2]
03
Growth loop
- 1First video with your own face
- 2Published on your own channel
- 3Viewer: "how was this made?"
- 4New sign-up, 3 free videos
- 5Credit purchase
- Leak: deepfake perception: trust leak
The product's output is the distribution channel itself. Credits pass inference cost to the customer on every transaction; the red arrow shows the category's trust problem.
The drawing shows the shape of the loop; each node is explained in the list below. The red arrow marks value leaking out of the loop.
04
Mechanism
- Acquisition channels
- Not paid ads but circulating output. Translated and dubbed videos are shared on social media; the official text calls it "a product that grows largely through word of mouth" [1]. The main revenue base is small businesses, real-estate agents, trainers and course creators [2]. The claim of more than 40,000 business customers in mid-2024 is in the podcast host's intro [4].
- Loop
- A user makes their own avatar or translation video, publishes it on their own channel, a viewer asks "how was this made", a new user arrives. The enterprise layer (team workspace, SCIM, MFA) rides on top of the self-serve base [3].
- Activation
- The first moment of value is the first video with your own face and voice. The free plan gives 3 videos and 1 minute, and 1 custom video avatar is free [3].
- Retention evidence
- No retention evidence. Churn, cohorts and credit usage rates are not disclosed; whether the 30M users are registered or active is unclear [1]. Credit rollover on monthly plans is a retention mechanic [3].
- Role of pricing
- Credits plus model selection; the founder says they keep every transaction profitable [2]. This structure passes inference cost straight to the customer and closes the negative-margin risk (opinion). Price tiers run from $29 to $149 [3].
- Distribution moves
- Launch cadence: 63 products and features in the past three months, per the official blog [1]. Avatar V, the Seedance partnership and the real-estate vertical are in the 2026 blog archive [7]. API and MCP as a partner channel; the open-source HyperFrames project reached 21,600 GitHub stars in its first month (company claim) [1].
05
Real reading
The five questions of the Growth Claim Check, applied to this company's public claims.
- 01
What was measured, and by whom?
yes
ARR is on the company's official blog; no independent audit.
- 02
Perception or outcome?
yes
Revenue and cash-flow claims, not perception metrics.
- 03
Is there a stability or quality metric?
unclear
Retention, churn and credit usage are not disclosed; the definition of "30M users" is unclear.
- 04
Is there a counterfactual or cohort?
no
No data on which channel's users stay.
- 05
Nominal or real?
unclear
Credit-based ARR includes prepayments and enterprise commitments; the break-even claim strengthens the real side.
Verdict
Close to a real reading
Revenue, cash flow and capital efficiency are claimed together; that trio alone is rare. What is missing is proof of retention.
Test your own claim with the same five questions: Growth Claim Check →
06
Shadow side
- Deepfake misuse: security firms and media listed HeyGen among tools used in fraud and misleading videos (secondary data) [6].
- "AI slop" is the founder's own phrase; the category is fighting that perception [1].
- Competitor Synthesia says it reached 85% of the Fortune 100 in the same period; the same claim appears at two companies at once (secondary data) [9].
07
What it teaches a founder in Türkiye
This section is opinion; a reading, not a source.
- 01
(opinion) When the output is the distribution, what matters is not an ad budget but the design of a "shareable first output". That is why the free plan has three videos.
- 02
(opinion) With dollar-priced inference and lira revenue, the "profit on every transaction" rule is vital: tie credit prices to model cost, never sell an unlimited plan.
- 03
(opinion) Growing on little capital is the only accessible path in Turkey: first a profitable small segment (trainers, real-estate agents), then enterprise.
- 04
(opinion) Sixty-three features in three months is possible with a small team; but without measurement you never learn which feature carries revenue. HeyGen does not disclose it.
A question for you: remove this company's logo and drop the loop onto your own product. Which node stays empty?
08
Sources
Numbers are the references in the text. An 'opened' source was read by the researcher; 'secondary' ones are verified against their primary source before publication.
- [1]HeyGen Surpasses $200M ARR
HeyGen official blog · 2026-06-25 · opened
“People don't want more AI slop. They want to communicate with trust, clarity, and presence in every language and every format where their audience lives. · a product that grows largely through word of mouth”
- [2]Exclusive: HeyGen's AI Video Reaches $200M In ARR, Without The Burn
Upstarts Media (Alex Konrad) · 2026-06-25 · opened
“We've been very constrained. · doubled from $100M eight months prior”
- [3]HeyGen pricing
HeyGen · accessed 2026-10-05 · opened
“Credits are HeyGen's single currency for video and asset creation”
- [4]AI Avatars & the Future of Video with HeyGen CEO Joshua Xu
Cognitive Revolution (podcast) · 2024-08-02 · opened
“we have about 40 and 50, somewhere around that”
- [5]How I Built This In Public: Joshua Xu
Build in Public Hub · 2023-05-05 · opened
“They recently crossed an incredible milestone: $1m ARR in 7 months”
- [6]HeyGen
Wikipedia · 2026-06-10 revision · secondary
Note: secondary data: name changes, Series B, deepfake reports
- [7]HeyGen blog archive
HeyGen · accessed 2026-10-05 · opened
“Avatar V (2026-04-08), Seedance 2.0 (2026-04-07), Real Estate (2026-08-11)”
- [8]Joshua Xu, Co-Founder & CEO at HeyGen – Interview Series
Unite.ai · 2024-10-08 · opened
Note: customer examples: McDonald's, Weather Channel, UNDP
- [9]
09
Open questions
- Primary address of the official $100M ARR announcement; the official Series B blog post (404).
- The co-founder's name (Liang or Zheng) should be verified against a primary source.
- No retention or churn data at all; the definition of "30M users".
- The definition behind "85% of the Fortune 100": paying customers or any account.