Late 2025. We were prototyping a voice agent for a customer-support workflow. The model worked. The conversation worked. Then we priced a month of audio at the incumbent APIs, multiplied by a hundred concurrent users, and the number stopped us cold. Lifelike speech, the kind a real customer actually wants to hear, cost more than the entire rest of the stack put together.
That math does not only fail a scrappy startup. It fails an established product team rolling voice out across a million users, a public-service app trying to reach citizens in their own language, and a research group trying to ship anything at all. So the lab was founded on one question: what would it take to run a competitive speech model on a normal CPU, in a normal cloud region, at a price nobody flinches at?
Three months of benchmarking, distillation, and infrastructure work later, the first instrument was serving. feather: a 117-million-parameter flow-matching model, eight synthesis steps, no GPU anywhere in the path, and the first word out in under 100 ms. 0.19¢ per 1,000 characters, which is $1.90 for a million.
Then came starling, the expressive flagship: 107 languages, zero-shot voice cloning, $9.00 per million characters, and the same rule as feather. It has to run where your data already lives.
We do not chase leaderboards. Squeezing out the last decimal of "sounds human" is astronomy for its own sake; frontier models crossed that threshold already. We chart the sky so products can navigate it, and every claim on this site is measured on the hardware that serves you, not on a slide.