ZUG, SWITZERLAND - Lalal.ai, the developer of an AI-powered audio processing platform, has released Lynx, its first neural network designed exclusively for speech denoising. Lynx is trained to separate speech from background music, crowd noise, mechanical interference, environmental sounds and a full range of acoustic artifacts. The result is a clean voice track that’s ready for further post production without the need for manual cleanup.
The model is six times smaller than Andromeda, the company's flagship cloud stem separation model, which reduces computational load without compromising output quality. The model was trained on a manually-curated dataset. Since a reliable automated pipeline doesn't exists for cleaning the full dynamic range of real-world speech from diverse noise sources, the team spent months hand-selecting, filtering, trimming down and preparing audio tracks covering a range of conditions, from quiet interviews to noisy field recordings.
Lynx is available now through Lalal.ai's Voice Cleaner and Voice & Noise stem via browser, mobile app, desktop app (cloud mode), and the Lalal.ai API. Planned improvements to the Lynx architecture include enhanced separation of choral and group vocals, and improved isolation of speech recorded at a distance from the microphone.