We train speech models to help AI understand what people say, how they say it, and what they mean—directly from the audio.

Try the demo · Explore our models · API docs · Research

Speech understanding, from the audio

Words and speakers. Transcribe speech and follow who said what, with time-coded segments.
Emotion and delivery. Add context from tone, rhythm, emphasis, and vocal expression.
One API. Bring transcripts and acoustic context into the products you build.

Open research

• Orukeet — our open multilingual speech recognition model, with code, weights, evaluation results, and a technical report.
• oruk-bench — published speech-emotion evaluation results, with methods and limitations documented alongside the scores.
• Research highlights — experiments, model releases, and interactive explanations from the team.

Built by researchers from Stanford, Berkeley, and Cambridge. Meet the team →

Website · GitHub · Brand assets · Contact