A trainer at a Shenzhen exporter once recorded one Mandarin onboarding video and hoped it would do. This year the same team shipped the same presenter — cloned, lip-synced, and fluent — in 40 languages from a single script. HeyGen's digital human (数字人) pipeline turned one recording into a multilingual library.
Key takeaways
- HeyGen, founded in 2020 (originally Movio/Surreal) by Joshua Xu and Wayne Liang, builds AI avatar and dubbing video tools; its team has deep China roots though it is now Los Angeles-headquartered.
- The company was named G2's #1 Fastest Growing Product in 2025 and supports 175+ languages and dialects with voice cloning and lip-sync.
- Privately reported figures put ARR around US$100M in 2025 (≈ US$95M in September per secondary trackers), with ~US$65.6M raised including a US$60M Benchmark-led Series A in 2024.
- HeyGen says it passed 31 million registered users and 100,000+ paying business customers in 2025 — company-reported, not audited.
- Its instant-avatar feature turns a short selfie clip into a reusable digital twin, a fit for cross-border sellers localizing at scale.
What the product actually does
HeyGen is not a text-to-video model in the cinematic sense; it is a presenter engine. You write a script, pick or clone an avatar, and the platform renders a talking-head video with synced voice and lips. The standout feature is the video translator: drop in existing footage and it dubs and lip-syncs it into another language while preserving the original voice's tone. For an exporter, that means one shoot becomes a dozen market-ready clips without re-filming or hiring local voice actors.
Why China-rooted matters here
Although HeyGen is headquartered in Los Angeles, its founders are Chinese and its earliest and heaviest user base includes Chinese cross-border e-commerce sellers who need to localize product and training videos fast. That makes it a useful lens on China's AIGC (生成式内容) export: the innovation is global in distribution but China-shaped in demand, built for sellers who live or die by listing videos across marketplaces.
The numbers — and their caveats
The ARR and funding figures are not from audited filings; they come from startup trackers and secondary reporting. We use them as directional, not authoritative. The "31 million users / 100,000 business customers" claim is HeyGen's own and should be read as a company milestone statement. What is verifiable is the public recognition: G2's 2025 fastest-growing ranking is an externally awarded position, and the 175-language capability is a stated product spec.
Competition and moat
HeyGen sits among Synthesia, D-ID, Colossyan, and Hour One in the avatar-video space, plus text-to-video entrants. Its differentiation is avatar realism, fast localization, and a self-serve model that serves both solo creators and enterprises. Public case studies on its site cite names like OpenAI, HubSpot, Ogilvy, trivago, and Würth Group — brand references from the company, useful as signal but not independent endorsements.
Built for the export seller
The China angle shows up in product priorities. HeyGen's instant-avatar flow — record a 30-second selfie clip, get a reusable digital twin — is tailor-made for small exporters who cannot afford a film crew in every target market. A single Mandarin or English presenter becomes the face of localized listings across Southeast Asia, Europe, and Latin America. That use case, more than enterprise demos, is what drove the steep user curve the company reports, and it explains why a US-HQ tool keeps showing up in Chinese cross-border playbooks.
The deeper shift: faces as a service
HeyGen points to a quieter change in creative labor: the presenter is becoming a reusable asset. Instead of booking a person per recording, teams mint a digital twin (数字人) once and summon it on demand in any language. That collapses the cost of "being present" in many markets at once. The open question is trust — viewers may not know they are watching a synthetic spokesperson, which is exactly why visible and metadata labeling (see our labeling piece) matters as the technology spreads.
Honest limitations
HeyGen is US-headquartered, so it is not a domestic Chinese product in the regulatory sense, even with China-rooted founders and a China-heavy user base; readers outside China should weigh that. All financial figures (ARR ~US$100M, ~US$65.6M raised, US$60M Series A) are privately reported estimates from trackers such as NeuronFeed and IA Hunt, not audited statements; treat them as indicative. User and customer counts are HeyGen's own claims. We did not test output quality or language coverage ourselves, and avatar-video realism varies by avatar and source footage. As with all generative video, the China labeling rules mean any AI-produced (生成式内容) clip published domestically should carry the required labels.
What readers can do now
- Cross-border sellers should prototype one product video through HeyGen's translator and measure time saved versus hiring local voice talent per market.
- Trainers can build a cloned presenter once and localize onboarding into the languages your workforce actually uses.
- Before scaling, check each market's disclosure rules — including China's labeling standard — so dubbed AI presenters are clearly marked.