Audio Noise Reducer Online
or drop files here
Audio Noise Reducer
Reduce steady background hiss using a sampled noise profile and spectral attenuation. The selected opening profile should contain only background noise. Short-time Fourier spectral subtraction attenuates matching noise; speech in the profile can damage wanted speech.
- Dedicated audio noise reducer controls
- Local 16-bit PCM WAV export
- Automatic background-rendered preview
Choose a noise-only opening
The first profile interval teaches the processor what background sound to attenuate. Use a recording that begins with room hiss, fan noise or another steady bed rather than a spoken introduction.
Subtract estimated spectral noise
Short Fourier windows separate the signal into frequency bins. Estimated noise power is subtracted proportionally, and a spectral floor retains some energy to avoid completely empty bins.
Balance cleanliness and intelligibility
Start near 40–60% reduction. If consonants sound watery or metallic, lower the amount or raise the floor; aggressive attenuation can remove voice components that overlap the noise spectrum.
Understand changing backgrounds
A profile of a steady fan cannot accurately describe later traffic, keyboard clicks or another person speaking. Those intermittent sounds may remain, and the method is not an AI speech-isolation model.
Technical references: Web Audio API specification