WoodstitchHuman evals for voice AI

The human benchmark for speech AI.

Woodstitch runs blinded listening studies on text-to-speech models with paid human raters. Every study states its method, its numbers and what it does not show.

Live · updated 20 Nov 2026, 20:30 UTC

TTS Voice Leaderboard

Different jobs, different leading voices. 75 stock voices from 19 providers, ranked by blind paired listening.

See the leaderboards →

Phone support

1RuthAmazon Polly1197
2MarinOpenAI1166
3GordonMurf1159
4KelseyInworld1159
5KatieCartesia1145

Nonfiction narration

1MarinOpenAI1169
2KatieCartesia1169
3DanielleAmazon Polly1163
4CelesteRime1143
5JonElevenLabs1130

Elo within each use case (1000 = average voice). Preliminary: 1,093 judgments so far; intervals on the full boards.

Research

All studies →
18 Nov 2026

Which Fish Audio voice is best for audiobooks?

76%“Old” first of 32 voices

998 judgments · 72 listeners

11 Nov 2026

Which ElevenLabs voice is best for audiobooks?

66%Marie first of 32 voices

1,122 judgments · 125 listeners

6 Nov 2026

Which TTS model is best for audiobooks?

56%ElevenLabs v4 win rate; Gemini and Fish tied with it

481 judgments · 78 listeners

5 Nov 2026

Which AI model writes the best piano music?

66%Gemini 4 Argon win rate, Llama 4 last

62 listeners · interim

6 Nov 2026

Can listeners tell AI narration from human?

48.8%accuracy: no better than a coin flip

496 judgments · 75 listeners

For AI teams

Testing a release candidate? Run blinded human evaluations on your own models, voices and scripts. Private by default — paid work never changes a public result.

Evaluate your model

For listeners

Get paid to compare short audio clips and pick the one that sounds better. Work whenever you want.

Become a rater

How every study is run — listeners, payment, blinding, statistics, publication rules — is in the platform methodology.