Oruk

Active

Speech foundation models that understand people: words, tone of voice, emphasis, intent, and emotion.

What it does

As AI tools become mainstream and robots enter the workforce, conversation is replacing the text box as the only interface fast enough to keep up with critical demands. The voice layer behind every agent, coding session, robot command, and phone call needs the full context: words, tone of voice, emphasis, intent, and emotion. When someone says "I'm fine," they could be reassuring you or asking you to leave them alone. The words are the same. The meaning is in the voice. We are speech, AI, and linguistics researchers from Stanford, Cambridge, and Berkeley, building Oruk to train speech foundation models that actually understand people. We've already set new records across speech understanding benchmarks, deployed to hundreds of thousands of users, and assembled the world's largest speech perception dataset.

Its site says now · captured 2026-11-07

oruk — Speech understanding API

Speech understanding models for transcription and vocal delivery in recorded audio or live speech. Language support and outputs vary by model.

How it has moved

  • 2026-11-07 · pitch rewritten

    Was: oruk — Speech API for transcripts, emotion, and style - Oruk builds speech models and an API for English transcription, emotion scores, and speaking style, with Python and TypeScript SDKs and reproducible research.

    Now: oruk — Speech understanding API - Speech understanding models for transcription and vocal delivery in recorded audio or live speech. Language support and outputs vary by model.