THU, OCTOBER 22, 2026
Independent · In‑Depth · Practitioner‑Tested
✎ Voice & Audio

HeyGen Just Took First Place And Cut The Price By 87%

HeyGen Voice Turbo went live on 21 October at $4.00 per million characters and took the top spot on Artificial Analysis' Controlled Voice leaderboard with 1,202 Elo, beating HeyGen's own previous model at $30 and Eleven v4 at $80.

By AIToolsRecap October 22, 2026 4 min read 40 views
Home › Articles › Voice & Audio › HeyGen Voice Turbo Tops TTS Leaderboard at $4 P...
THE SHORT VERSION

● #1 at 1,202 Elo — HeyGen Voice Turbo, $4.00 per million characters
● #5 at 1,156 Elo — Eleven v4, $80.00 per million characters
● Model ID — heygen-voice-1-turbo, live on the API 21 October
● Caveat — prices are Artificial Analysis conversions, and ranks 1-3 are grouped

Artificial Analysis' Controlled Voice Arena as of 22 October 2026 reads: HeyGen Voice Turbo #1 at 1,202 Elo, $4.00 per million characters; HeyGen Voice #2 at 1,195, $30.00; Qwen-Audio-3.1-TTS-Plus #3 at 1,184, $19.30; Eleven v4 Turbo #4 at 1,156, $40.00; Eleven v4 #5 at 1,156, $80.00; Sonic 3.6 #6 at 1,151, $49.00; Inworld Realtime TTS-2 #7 at 1,145, $20.80. Read the first and fifth rows together — Turbo scores 46 Elo above Eleven v4 and costs one twentieth as much. Against HeyGen's own previous model it is 7 Elo higher at 87% less.

Two caveats most coverage is dropping, and both matter before you switch vendor. The prices are derived, not published: Artificial Analysis converts HeyGen's credit pricing into a per-million-character figure, and HeyGen does not advertise $4.00 as a rate. And ranks 1 to 3 are statistically grouped, so the ordering between Turbo at 1,202 and Qwen-Audio at 1,184 is not established — Turbo's confidence interval is 1,183 to 1,221 across 1,685 samples. The defensible claim is not that HeyGen is objectively the best TTS model. It is that Voice Turbo is indistinguishable from the top of the board while costing a fraction of everything near it, which is the stronger fact for a buying decision anyway.

The model is live on HeyGen's API as heygen-voice-1-turbo. If you run high-volume narration — audiobooks, course content, bulk voiceover — the gap between $4 and $80 per million characters stops being a line item and becomes the budget. If you are on Eleven v4 for quality reasons, run your own A/B on your actual scripts rather than trusting the board. And check first-token latency separately: these are cloned-voice quality results, not a latency ranking, so do not move a live voice agent on this table alone.

Sources

Artificial Analysis Controlled Voice leaderboard · OrcaRouter analysis

Tags
AI NewsVoice AIElevenLabs2026
⚑

Spot an inaccuracy?

We verify facts before publishing and correct errors promptly. If something in this article is wrong or outdated, let us know.

Report an error →