HeyGen Voice is the company's first native text-to-speech model and currently ranks first on the Artificial Analysis Controlled Voice TTS Arena with an Elo of 1,201. It has outperformed Alibaba's Qwen-Audio-3.1-TTS-Plus and ElevenLabs' Eleven v4 Turbo in blind tests, particularly in UK English and assistant roles. While it generates 40 characters per second and ranks 10th for overall pronunciation robustness at 83.1%, it placed second for preserving exact text sequences.
