The AI Voice Production Agent
Built by the people who measure it. Onepin orchestrates today’s best voice models, validates every output, so you can ship audio that’s ready to publish.
We’re building the future of AI voice
across all human languages.
Measured, not guessed
We built Podonos, the infrastructure used to score speech AI. We define how good is measured.
Models aren't the bottleneck
We route to the best model for the job. Your quality never rides on one vendor.
Voice for every language
Built for the world's languages, not English-only.
Sweat the last 5%
Every generation is checked before it ships. No surprises in production.
Soohyun Bae (SB)
Founder & CEO
Note from the founder
During the pandemic I tried to turn the books I loved into audio I could actually listen to. The technology was supposed to be ready. It wasn't. Voices mispronounced names, flattened the emotion, and fell apart the moment a sentence got complicated. In English it was frustrating. In my own language, and in most of the world's languages, it was far worse.
The surprising part is that the models were never the problem. Today's voice models are extraordinary. The gap is in everything around them: picking the right one, guiding it, and checking that what comes out is genuinely good enough to publish. That layer barely existed, and the moment to build it is now.
So we did. We started Podonos to measure voice quality at scale, and Onepin is what we built on top of it: the layer that turns great models into voice you can ship, in any language.
Production-quality voice for everyone with something to say, in the language they say it in.
Built with expertise from
Ready to publish?
Turn any script into production-quality voice,
in any language, in minutes.