TL;DR
HeyGen launched HeyGen Voice on October 9, its first in-house voice model — and it debuted at number one on Artificial Analysis' independent leaderboard for AI voice systems. The model is free inside HeyGen's platform and API; a $99/month Professional Voice Clone add-on trains on 30 minutes to three hours of your own speech for the closest match to your real voice. The idea is business-quality video narration without re-recording takes — and HeyGen requires the voice owner's explicit consent before a clone can be made.
What HeyGen shipped
HeyGen Voice is HeyGen's first voice model built entirely in-house. It sits inside the same stack as the company's Avatar V model — identity and voice now come from one place instead of a patchwork of vendors. The launch follows the recent release of HeyGen Video, part of the company's push to own the whole pipeline from script to finished video.
The headline claim is checkable: at launch, HeyGen Voice debuted at number one on Artificial Analysis' leaderboard, the independent benchmark site that ranks AI models across categories. Third-party, public, and anyone can go look at the board — that is the part of the announcement that does not depend on taking HeyGen's word for it.
The free model and the $99 clone
HeyGen Voice is free to use inside the HeyGen platform and its API. The paid tier is HeyGen Professional Voice Clone, a $99/month add-on trained on 30 minutes to three hours of the owner's speech to produce the closest representation of their real voice. Training a clone requires the voice owner's explicit consent, with built-in safeguards over how the voice and likeness can be used.
The model is built for delivery, not just reading text aloud — emphasis, emotion, and pacing are meant to carry over so the output sounds like the person actually delivering the line. As HeyGen CTO Rong Yan put it, the goal is a voice “one that always sounds like you at your best.”
Why now: the unposted video problem
HeyGen's target is the business owner who records a video and never posts it. The company surveyed more than 1,000 small business owners in its own user base and found 71.6% had recorded a business video and then decided not to post it — blaming production quality or discomfort on camera. Voice cloning plus a personal avatar removes both excuses: record once, fix the flubs in text, regenerate.
The company numbers around it: more than 40 million users across 196 countries, more than 118 million videos created on the platform, adoption across 85% of the Fortune 100, and annual recurring revenue above $200 million, doubled in eight months. Those figures are company-supplied, not third-party verified — but the direction is clear. AI-generated business video is moving from novelty to infrastructure.