WoodstitchHuman evals for voice AI

Marin

OpenAI · voice id marin

female · rendered November 2026 at default settings

Results by use case

Use caseRank / fieldRank rangeJudgments
Phone supportgpt-4o-mini-tts7 / 421–3325View ranking →
Nonfiction narrationgpt-4o-mini-tts2 / 421–3016View ranking →

The rank range is the set of ranks consistent with the listening data at 95%. Each row is a separate evaluation; ranks are not comparable across rows.

Evaluated samples

Phone support: Nonfiction narration:

Why this voice

Phone support: docs: "for best quality use marin or cedar" Nonfiction narration: docs: "for best quality use marin or cedar"

Entries are immutable: a newer model or changed defaults becomes a new dated entry. To dispute a render, nominate the voice for the next edition.

Share: Marin

Share image
Suggested post text
Embed
<a href="https://woodstitch.com/text-to-speech/voices/openai--marin"><img src="https://woodstitch.com/text-to-speech/cards/voice--openai--marin.png" alt="OpenAI Marin: Woodstitch TTS Voice Leaderboard results" width="480"></a>
Independent blind listening evaluation. Paid, vetted US listeners.
Methodology · Evaluated audio · Judgment data · Contact