HeyGen Ships Professional Voice Clone API for Avatars

HeyGen makes AI-generated avatar video, where a photo or short clip of a person is turned into a talking video reciting arbitrary text. Voice has been the weaker half of that: even a convincing avatar face often played back a generic text-to-speech voice. HeyGen's new Professional Voice Clone is a dedicated model trained on about twenty minutes of a person's own recorded speech, now in preview through HeyGen's API.

HeyGen's post says a clip alternating between the real person and the clone is hard to tell apart, but no video or audio file accompanied the post, so that comparison cannot be checked directly here. What is confirmed is that a trainable, person-specific voice model is now available to developers building avatar or dubbing products, not the degree of realism it achieves in practice.