HeyGen turns a script into a video of a presenter reading it. The presenter is an AI avatar, either a stock character or a digital twin of a real person made from a short recording, and the finished video can be translated with the lip movements matched to the new language.
Joshua Xu, a former Snap engineer, co-founded the company and runs it as CEO. In 2024 it raised a $60 million Series A led by Benchmark, which valued it at over $500 million. HeyGen counts more than 40 million users, 85% of the Fortune 100 among them, and in June 2026 it passed $200 million in annual recurring revenue.
A presenter for every script
The business uses fall into a few jobs. Training teams turn a policy document into a short video and regenerate it when the policy changes. Sales teams send prospects a video that greets them by name. Marketing teams take one video into several countries, since HeyGen’s video translation covers 175+ languages and dialects.
In the web app, Batch Mode renders many single-scene videos in one go, each script with its own avatar and voice.
A digital twin needs proof that the person in it agreed. Usually that person records a short consent statement on camera through HeyGen’s hosted consent page, and Enterprise accounts can upload a pre-recorded consent video instead. Photo avatars have no consent step, and HeyGen’s docs leave getting and recording the subject’s agreement to the account holder.
One template, a video per customer
Personalized video at scale runs on Studio templates. A template is a video designed once in HeyGen Studio, with the parts that change marked as named variables: a line of the script, the avatar, the voice, a background image. Each request fills those variables with one customer’s details and renders a separate video.
A variable’s name in Studio is the key the API fills, so a slot called first_name takes whatever arrives under that name, from a CRM field or a spreadsheet row.
Videos from HubSpot workflows
HeyGen’s HubSpot app adds a workflow action called “Generate a HeyGen personalized video for a contact.” Installing it creates two contact properties, and connecting it takes a HeyGen API token. In the workflow, the action points at a template and maps contact properties onto its script variables. The video’s language can follow the contact’s language preference or country.
When the video is ready, its link is written to the contact record. A marketing email later in the same workflow can show an animated GIF preview from the heygen_video_gif property, linked to the video itself. A post-demo follow-up goes out with the prospect’s own name spoken by the presenter, and nobody had to record anything.
Training modules for the LMS
For training, any HeyGen video can be downloaded as a SCORM package, the format learning management systems import, in SCORM 1.2 or SCORM 2004 (4th Edition). A slider sets how much a learner has to watch before the LMS marks the module complete, 80% for example. Choosing “Host video on HeyGen” keeps the video on HeyGen’s servers, so a corrected and regenerated video updates the course without uploading a new package.
Automation tools and AI assistants
In Zapier, a new lead can create a video from a template, and a finished video can trigger whatever happens next. The verified HeyGen node in n8n renders a video from a script and an avatar, then passes the video URL to the next step.
Inside Canva, the HeyGen AI Avatars app adds a presenter to a Canva design and exports the result as an MP4. Voices from an ElevenLabs account can be imported with an ElevenLabs API key and used in HeyGen’s AI Studio.
For live conversation, LiveAvatar streams a real-time avatar and can sit on top of an existing LiveKit setup or an ElevenLabs agent. HeyGen also lists Zoom among its integrations, for interactive avatars in meetings.
A ChatGPT plugin and a Claude connector make videos from a chat, through HeyGen’s MCP server. They draw on the web account’s subscription credits, a route HeyGen’s docs describe as sized for trial rather than scale.
Rendering through the HeyGen API
The API authenticates with an API key and bills to HeyGen’s API plans, separately from web app subscriptions. Video jobs run in the background, and the number running at once is capped by plan: 10 on pay-as-you-go and 20 on Enterprise. An avatar video’s script can run to 5,000 characters, and output defaults to 1080p, landscape or portrait.
An integration can also register a webhook, so it hears about finished videos without checking. HeyGen posts an event when an avatar video or a translation succeeds or fails, and each event carries a signature to prove HeyGen sent it.
Where the script comes from
HeyGen’s templates and webhooks open plenty of ways to fit video into an existing process. A new lead can get a personal welcome straight from the CRM, a policy change can regenerate its training module in the LMS, and a finished video can go into the next email in the sequence.
A personalized video is only as accurate as the CRM fields feeding its template, and a misspelled first name gets spoken on camera to the customer, so tidy customer data comes before AI. For a workflow built in Zapier or n8n, the signs it needs something sturdier are worth knowing, and the time it saves per video is quick to work out.