ai-coustics

ai-coustics builds real-time, on-device speech enhancement software designed to turn raw and unpredictable audio into stable input for Voice AI systems. Founded in 2021 and spun out of the Technical University Berlin, the company positions its technology as an audio intelligence layer sitting between the physical world and machine understanding. Its work spans speech enhancement, acoustics, machine learning, voice activity detection, multi-speaker processing and audio reliability telemetry.

The company sells a developer-facing SDK and APIs, which Voice AI companies use to integrate real-time speech enhancement into their voice pipelines. These are reported to process millions of hours of audio each month. Named customers include PolyAI, Synthesia, Perplexity, Retell, LiveKit and AssemblyAI.

Its model suite comprises Quail, a speech enhancement model running on-device at sub-40ms latency; VAD, for voice activity detection; Tyto, an audio insight model; and Rook, for multi-speaker processing. The company serves customers worldwide across voice, speech, audio and conversational AI markets.

ai-coustics is a research-driven organisation made up of researchers and engineers specialising in audio, acoustics and machine learning, with roots in academic research at the Technical University Berlin.