Remove Background Noise from Audio
Strip HVAC hum, fan noise, traffic, keyboard clicks, and mic hiss from any audio recording — entirely in your browser.
Powered by RNNoise · 100% Private · No Account Required
Drop audio files here — noise removed instantly
MP3 · WAV · FLAC · M4A · OGG · OPUS · up to 20 files
110 KB model · loads once · 100% private · no uploads
What This Tool Is Best For
Meeting & Interview Recordings
Remove HVAC noise, AC hum, and background chatter from Zoom, Teams, or in-person recordings before sharing or transcribing.
Podcast & Voice Over Cleanup
Fix home studio recordings with room noise, mic hiss, or electrical hum without expensive DAW plugins or cloud services.
Lecture & Classroom Audio
Clean up recordings captured in noisy lecture halls, classrooms, or conference rooms where background noise masks the speaker.
Video Narration Tracks
Clean the audio track from a screen recording or tutorial video before export. Works on the audio extracted with MP4 to MP3 first.
How It Works
Upload your audio
Drag and drop your MP3, WAV, FLAC, M4A, OGG, or OPUS file onto the tool — or click "Choose files" to browse. Up to 20 files at a time.
Adjust strength (optional)
Open Settings to set the suppression strength. 85% works well for most recordings. Lower values preserve more room sound; higher values give a drier, cleaner result.
Wait for processing
The RNNoise model processes your audio frame by frame entirely in your browser. A 10-minute file typically takes under 15 seconds on a modern device.
Preview and download
Click the Play button next to any result to preview the clean audio. Click Download to save it as a WAV file, or Download All to get a ZIP of all cleaned files.
The Technology Behind Noise Removal
This tool uses RNNoise — a recurrent neural network noise suppressor originally developed by Jean-Marc Valin at Mozilla and Xiph.Org, and now used in production by WebRTC, Firefox, and Jitsi Meet. It was designed specifically for real-time voice processing and runs on audio sampled at 48 kHz.
RNNoise processes audio in 10 ms frames (480 samples each). For each frame, the network computes a gain mask across frequency bands — suppressing noise-dominant bands while preserving speech-dominant ones. Because the model is compact (the WebAssembly binary is only ~110 KB), it runs instantly in your browser without needing a GPU or large model download.
For stereo files, both the left and right channels are processed independently so spatial positioning is preserved. The suppression strength slider controls a wet/dry blend between the RNNoise output and the original signal, letting you tune from subtle noise reduction to aggressive suppression.
Your audio is decoded using the browser's native Web Audio API (decodeAudioData), then resampled to 48 kHz if necessary via OfflineAudioContext. The output is a 16-bit PCM WAV at 48 kHz — a lossless container that every audio editor, video tool, and phone can play.
Frequently Asked Questions
What types of noise does this tool remove?
The tool uses Mozilla's RNNoise model, which is designed for voice recordings. It excels at removing steady-state background noise: HVAC hum, fan noise, air conditioning, road traffic, light rain, electrical hiss, and microphone self-noise. It also handles moderate keyboard clicks and background chatter. It works best on recordings with a continuous noise floor, not sudden sharp sounds.
Does my audio get uploaded to a server?
No. The RNNoise model (a 110 KB WebAssembly binary) loads directly into your browser. Your audio file is decoded, processed, and re-encoded entirely in local memory — nothing is transmitted to any server. This makes the tool safe for sensitive recordings such as meetings, medical consultations, or legal interviews.
What audio formats are supported?
You can drop any format your browser can decode: MP3, WAV, FLAC, M4A, OGG, OPUS, WebM audio, and AAC. The output is always a 16-bit PCM WAV file at 48 kHz — a lossless, universally compatible format. If you need MP3 output, convert the WAV with the free MP4 to MP3 tool afterward.
What does the suppression strength slider do?
RNNoise processes audio in binary frames — each 10 ms frame is either suppressed or passed through. The strength slider controls a wet/dry blend: 100% means you hear only the RNNoise output; lower values mix back the original signal. 80–100% works best for voice cleanup. 40–60% is useful when you want to reduce noise while preserving room acoustics for music or ambient recordings.
Will noise removal affect voice or music quality?
RNNoise is specifically trained for voice frequencies (300 Hz – 4 kHz), so it is very gentle on speech intelligibility. For music, it may color the sound slightly, especially at higher frequencies — use a lower strength setting (40–60%) for music recordings. For pure voice recordings such as podcasts, interviews, or meeting audio, 80–100% strength typically preserves natural voice quality very well.
How fast does it process audio?
Processing speed depends on your device. On a modern laptop, a 10-minute audio file typically takes 5–15 seconds. RNNoise is a small recurrent neural network model (compared to larger Whisper or Demucs models) — no GPU is required, and no large model download is needed. The 110 KB WASM binary loads in under a second.
Can I process multiple files at once?
Yes — drop up to 20 files at once. The tool processes them sequentially in your browser and lets you preview or download each result individually. A "Download All" button packages all clean files into a single ZIP when you have multiple results.