What is over-suppression?
Over-suppression happens when a speech enhancement or noise reduction system removes too much of the audio signal. It's the opposite risk to leaving too much noise in.
How does over-suppression work?
It's usually a side effect of training or tuning an enhancement model purely for noise reduction without constraints on preserving speech structure. The model learns to err on the side of silence, which feels "cleaner" to the ear but loses content machines and listeners need.
How does ai-coustics help over-suppression?
Avoiding over-suppression is a core design principle of our Quail family. Instead of training Quail to scrub every trace of noise, we optimize it to preserve the acoustic cues that matter for ASR and voice agent performance.
