HeyGen Voice Tops Blind-Test TTS Leaderboard, API at Half Price
HeyGen released its first speech synthesis model, HeyGen Voice, which topped Artificial Analysis's Controlled Voice TTS blind-test arena with an Elo of 1201, beating Alibaba's Qwen-Audio-3.1-TTS-Plus (1182) and mainstream models like ElevenLabs' Eleven v4. The model is now live on the HeyGen platform and API.
Confirmed
- Artificial Analysis's Controlled Voice arena compares voice-cloning models using the same 8 cloned voices (4 American, 4 British); HeyGen Voice ranked first with an Elo of 1201.
- Generation speed is 40 characters per second; standard API pricing is $30/million characters, below Eleven v4 Turbo ($40) and Eleven v4 ($80).
- HeyGen officially and @aftahiai both confirmed a promotion: API users get half price until October 31, i.e., $15/million characters.
Why it matters
- HeyGen started in video generation/digital avatars; topping a third-party blind-test leaderboard with its first TTS model puts it directly into the speech synthesis turf held by Qwen and ElevenLabs.
- Pricing and speed data show a clear cost advantage ($30 vs. Eleven v4's $80/million characters), and with the limited-time half-off deal, developer migration from existing solutions could accelerate.
2026-10-10 ~ 2026-10-10 · 7 related posts
Primary sources
- [source] HeyGen Voice tops Artificial Analysis TTS Arena with Elo 1201, beats Qwen and ElevenLabs — ArtificialAnlys · 2026-10-10
- HeyGen Voice takes #1 on Artificial Analysis controlled-voice TTS leaderboard, beating Qwen and ElevenLabs — ArtificialAnlys · 2026-10-10
- [source] HeyGen Voice: 40 chars/sec at $30/1M, cheaper than ElevenLabs' two tiers — ArtificialAnlys · 2026-10-10
- HeyGen Voice tops Artificial Analysis voice cloning leaderboard, undercutting Eleven v4 on price — ArtificialAnlys · 2026-10-10
- [source] HeyGen announces #1 TTS model, offers API at 50% off through 10/31 — HeyGen · 2026-10-10