General · EdgepediaSeptember 17, 20267 min read

HeyGen

HeyGen is an American AI video generation platform that creates videos of photorealistic avatars speaking a user's script, with tools for custom avatar creation, voice cloning, and video translation into more than 175 languages and dialects.¹ Founded in Shenzhen in 2020 under the name Surreal and later rebranded from Movio, the company is headquartered in Los Angeles and competes with Synthesia in the AI avatar video market.²^•³

Key fact	Detail
Founded	2020 in Shenzhen as Surreal; rebranded Movio, then HeyGen; headquarters moved to Los Angeles in 2022²^•⁴
Founders	Joshua Xu, former Snap engineer, and Wayne Liang, former ByteDance product designer⁵
Series A	$60 million led by Benchmark, June 2024, at a $500 million valuation⁶
Revenue	ARR of $1 million in March 2023, $10 million in October 2023, $18 million by end of November 2023; surpassed $200 million in June 2026⁵^•⁷
Scale	More than 30 million users, customers in 196 countries, adoption across 85% of the Fortune 100 (company-reported, June 2026)⁷
Translation	175+ languages and dialects; Speed and Precision lip-sync engines; Precision costs 10 credits per minute⁸^•⁹
Pricing	Credit-based plans: Creator at $29/month for 600 credits; Pro from $49/month for 1,000 credits, scaling to 100,000⁹

History

Joshua Xu and Wayne Liang met at school and both completed master's studies at Tongji University in Shanghai and Carnegie Mellon University. After graduating they moved to the US West Coast; Xu joined Snap, where he worked on Snapchat Ads and the AI camera, and Liang worked as a product designer at ByteDance, TikTok's parent company.⁵^•¹⁰ Xu left Snap at the end of 2020 and founded the company, initially called Surreal, in Shenzhen, with early backing from Chinese venture firms including Sequoia China and ZhenFund.²^•⁵

The company released its AI video creation app in September 2022 and relocated its headquarters to Los Angeles in December 2022. As a spokesperson-video tool it operated under the name Movio before rebranding as HeyGen; at the rebrand the company reported more than 2 million interactive videos generated and a team of about 30 people.⁴^•¹¹

Growth and funding. Annual recurring revenue reached $1 million in March 2023, $10 million in October 2023, and $18 million by the end of November 2023.⁵ In November 2023 the company raised $5.6 million led by Conviction, the venture firm founded by Sarah Guo, at a $75 million valuation, alongside the launch of a product that generated an AI avatar from smartphone photos in about five minutes.⁵ Video translation drove a viral moment in September 2023, and a translation of a speech by Argentina's president at Davos drew further attention in January 2024.⁴

In June 2024 HeyGen raised a $60 million Series A led by Benchmark, joined by Thrive Capital, Bond, and Conviction, valuing the company at $500 million.⁶^•¹² Months before the round, the company had dissolved its mainland Chinese operation amid heightened US-China tensions, and its Chinese investors, including IDG Capital, Baidu Ventures, HongShan, and ZhenFund, sold their stakes to US counterparts.²^•⁴ The company reported surpassing $200 million in ARR in June 2026, double its level of eight months earlier, with more than 30 million users and customers in 196 countries.⁷

Products

Avatars. The core product generates talking-head videos from a script using AI avatars. Users can create a Personal Avatar, a digital version of themselves built from recorded footage, or use a library of stock avatars; the Creator plan includes 700+ stock video avatars and up to 40 voice clones.¹⁰^•⁹ The company's Avatar V model, described in a technical report, conditions generation on a full reference video rather than a single image, using a Sparse Reference Attention mechanism to capture both static identity features such as facial geometry and dynamic patterns such as talking rhythm and gesture, and produces 1080p avatar videos of unlimited duration.¹³

Video translation. HeyGen's translation tool re-voices an uploaded video in a target language while preserving the original speaker's voice and optionally syncing lip movements. Two engines are offered: Speed, suited to mostly front-facing single-speaker footage, and Precision, at 10 credits per minute, which handles side-profile views, facial occlusions, camera angle changes, and multi-person conversations with higher lip-sync quality.⁸ The tool translates spoken audio, and optionally lip movement and captions, but not text baked into the video image; the company notes that AI translation cannot guarantee 100% accuracy.⁸

Video Agent and API. Video Agent, introduced in beta in 2025, handles an entire production workflow from a single prompt, including scripting, visuals, voiceovers, avatars, and editing.¹⁴

Pricing

HeyGen moved to credit-based plans in which credits serve as the single usage currency across video generation and asset-creation features. The Creator plan costs $29 per month for 600 credits, videos up to 30 minutes, and 1080p export; the Pro plan starts at $49 per month for 1,000 credits and scales to tiers of up to 100,000 credits per month. A Free plan offers up to 3 videos per month with limited trial access to premium features.⁹^•¹⁶ Legacy unlimited plans remain active as long as subscriptions stay in good standing, but users who switch to credit-based plans cannot switch back.⁹

Consent, privacy, and safety

Creating a video-based Digital Twin requires a consent video, read from provided text, kept under 30 seconds, and recorded by the person depicted; avatars created for someone else require that person's own consent video, submitted at the same time as the avatar footage.¹⁷ HeyGen's biometric privacy notice states that the company collects face geometry and, where applicable, voiceprint data, uses it for avatar creation and for identity verification to prevent deepfake creation or impersonation, and may, with consent and except for enterprise customers, use videos and face geometry to train its models.¹⁸ Enterprise customers' data is excluded from AI model training by default, and non-enterprise customers may opt out.¹⁹

The company's ethics policy, effective November 13, 2023, prohibits content promoting violence, hate speech, discrimination, sexual explicitness, illegal activity, depiction of children, spam, fraud, or intellectual-property infringement, and requires demonstrated consent of the depicted person for custom avatars. HeyGen is a member of the Content Authenticity Initiative and lists compliance with GDPR, SOC 2 Type II, CCPA, the EU AI Act, and the DPF.²⁰ In 2024 the company's CEO stated that HeyGen does not allow political or election-specific content on its platform.²¹

Reception and research

HeyGen's avatar realism drew broad attention in 2023, when a reporter's AI-generated video double, created from a two-minute webcam recording plus a consent video, was described as convincing by friends and family, though the system mispronounced some names.²² Commentators noted that the same capability raises concerns about scams, fraud, and political misinformation.²²

In research on synthetic training videos, a 2024 study found that when employees perceived a trainer as synthetic, the avatar triggered a mild uncanny-valley effect and slightly lowered scores for perceived effectiveness and brand impression, though actual knowledge transfer was not reduced.²³ A 2024 comparative analysis of HeyGen, Synthesia, Colossyan, and HourOne using the Kano model evaluated avatar diversity, natural appearance, lip synchronization, and realistic emotion as key feature groups for medical education use.²⁴ HeyGen publishes its own research, including the Avatar V technical report, and in 2026 released an open-source real-time avatar stack and a Code2Video benchmark for evaluating AI-generated motion graphics.¹³^•²⁵

References

Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › AI companies, people, and products › AI startups and application companies › Generative media companies

Initially written Sep 17, 2026 · Reviewed: Oct 15, 2026 · Edited: Oct 11, 2026; Oct 15, 2026 · Last review: Oct 15, 2026

Notice something wrong?