Drop Audio File Here
MP3, WAV, FLAC, OGG
Neural Noise Suppression in Your Browser
This tool uses RNNoise, a recurrent neural network trained on thousands of hours of noisy speech. It removes stationary background noise — fan hum, air conditioning, keyboard clicks, room echo — while preserving natural voice quality. The neural network runs locally via WebAssembly, so your audio never leaves your device.
Clean Up Podcast and Interview Recordings
Background noise is the number one amateur-sounding mistake in podcast and interview recordings. A noisy HVAC system, outdoor traffic, or a buzzing USB mic cable can ruin otherwise great content. Drop your recording here and get a clean version in seconds — no DAW plugins, no subscription.
Fix Zoom and Teams Meeting Recordings
Video call recordings often pick up laptop fan noise, echoes from untreated rooms, and keyboard typing from participants on mute. Extract the audio, run it through the noise remover, and get a clean transcript-ready file.
Instant Processing
No server upload. Neural noise suppression runs locally using WebAssembly — results in seconds.100% Private
Your audio never leaves your device. Nothing is uploaded to any server.Works on Any Device
Desktop, tablet, phone, or Chromebook. Remove noise wherever you have a browser.Frequently Asked Questions
RNNoise is best at stationary noise: fan hum, air conditioning, room tone, white noise, and similar constant background sounds. It also handles keyboard clicks and moderate room echo. It is NOT designed to remove music, other voices, or sudden loud noises like dog barks.
RNNoise is specifically trained to preserve speech. In most cases, the voice sounds natural after processing. Very aggressive noise (like shouting over construction) may cause slight artifacts, but for typical indoor recordings the quality is excellent.
RNNoise is trained specifically for speech. Using it on music will likely produce artifacts. For music noise reduction, a spectral editor like iZotope RX is more appropriate. This tool is designed for voice: podcasts, interviews, meetings, voice memos.
RNNoise processes mono audio internally. Stereo files are downmixed to mono for processing, then expanded back to stereo. The noise removal is applied equally to both channels.
No. The neural network runs entirely in your browser via WebAssembly. You can verify this by opening the browser's Network tab — nothing outbound.