The Pulse
HeyGen Says Its New Voice Model Debuted No. 1
HeyGen launched its in-house HeyGen Voice model on October 9, saying it debuted at No. 1 on Artificial Analysis. The company offers the model free through its platform and API, with a separate $99-per-month voice-cloning add-on.
AI.info Team · Written with AI ·
We built HeyGen Voice to deliver the best possible AI voice that technology can deliver; one that always sounds like you at your best.
Rong Yan, CTO of HeyGen
HeyGen adds voice to its Avatar V stack
HeyGen announced HeyGen Voice on October 9, describing it as a new in-house voice model built into the same stack as its Avatar V system. The company says the model aims to preserve a speaker’s tone, pacing and expression, adding emphasis and emotion to generated speech. HeyGen presents the integration as a way to keep a user’s identity and voice in one system rather than combining products from different vendors.
The launch expands HeyGen’s video tools with a voice offering for creators and businesses. The company says it has more than 40 million users across 196 countries and has been used to create more than 118 million videos. Those figures, like the product-performance descriptions, come from HeyGen’s announcement.
The No. 1 claim lacks a score in the release
HeyGen says the model debuted at No. 1 on Artificial Analysis’ leaderboard, calling the ranking independent evidence of its quality. The announcement does not give an Elo score, a vote count, the models it beat or the specific Artificial Analysis voice arena in which it placed first.
Artificial Analysis describes its text-to-speech rankings as results from blind comparisons in which listeners choose between two speech samples generated from the same text. Its published methodology separates a Provider Voice Arena, which uses models’ own voices, from a Controlled Voice Arena, which standardizes voices to compare models more directly. Because HeyGen’s release does not specify which arena supports its claim, readers cannot tell from the announcement alone exactly what the No. 1 result measures.
The distinction matters: a preference ranking is not, by itself, a measure of every feature a buyer might weigh. Artificial Analysis describes separate measures for speed, price and pronunciation, while HeyGen’s release centers on natural-sounding delivery and likeness to the user’s voice. The company’s announcement does not provide results for those other measures.
Free access, with cloning at $99 a month
HeyGen Voice is available free through the company’s platform and API, according to the release. For customers seeking a closer match to their own voice, HeyGen offers Professional Voice Clone as a $99-per-month add-on. The company says that option trains on 30 minutes to three hours of the owner’s speech and requires explicit consent.
HeyGen also says its avatar products include safeguards intended to keep people in control of how their voice and likeness appear. The release does not detail those safeguards or explain how consent is collected for every use case, so the announcement establishes the company’s stated policy, not an independent assessment of its enforcement.
A product launch and a benchmark claim
HeyGen’s pitch combines two distinct offers: a free voice model integrated into its video product and a paid clone trained on a user’s recordings. Its No. 1 ranking could help distinguish the launch in a crowded text-to-speech market, but the missing score, vote count and arena designation leave the headline comparison difficult to assess. For now, the clearest specifics for buyers are access and price: HeyGen Voice is free on the platform and API, while Professional Voice Clone costs $99 a month.