music

Voice Cleaner

Reduce background noise in recorded speech, compare before and after, and save the cleaned audio as WAV in your browser.

View on GitHub

Local processing

About your filesFiles and content selected in this tool are processed in your browser and are not uploaded to a Browser Kitty backend.

Operation demo See the basic workflow in a short video.

About this tool

Voice Cleaner reduces background noise such as air-conditioner hum, fans and hiss around recorded speech. Import an audio or video file or record a new take using the microphone. The GTCRN model runs in the browser, without uploading the source recording to a conversion server.

Select a position in the waveform, try 10 seconds, and compare the original with the cleaned audio before processing the entire file. The saved result is 16 kHz mono WAV. Noise is not guaranteed to disappear, and the voice may sound different after processing, so always listen before exporting.

Good for

Reduce steady fan or air-conditioning noise in meeting or interview recordings to make speech easier to hear
Clean hiss from a phone-recorded explanation or narration and save the result as WAV
Test just 10 seconds and compare original and cleaned audio at the same position before processing everything
Process the voice audio from a compatible MP4 or other video file and extract the result as audio-only WAV

What it can do

Key capabilities available in this tool.

Use GTCRN to reduce steady background noise such as fans, air conditioning and hiss in speech recordings
Import WAV, MP3, M4A, AAC, OGG, WebM, MP4 and other formats that your browser can decode
Record from your microphone or select an existing audio or video file
Choose a waveform position and try a 10-second section before processing the entire recording
Process the full file with progress and cancellation, then compare Original and Cleaned at synchronized positions
Export 16 kHz mono 16-bit PCM WAV; the AI model and runtime are embedded in the HTML

How to use

Choose an audio or video file, or allow microphone access and record a new take.
Tap, click or drag on the waveform to move to the section you want to compare.
Select Try 10 seconds from here and listen to Original and Cleaned.
If the result is useful, choose Clean the whole file. Watch progress or cancel when needed.
After the entire recording has been processed, compare both versions, choose a filename and select Save WAV.

Supported

Input: WAV, MP3, M4A, AAC, OGG, WebM, MP4 and more, subject to audio decoder support in the browser and OS.
Limits: 250 MB per file and up to 60 minutes of audio. Device performance and available memory still affect processing.
Preview: waveform seeking, a processed 10-second sample from the selected position, and original/cleaned comparison at matched playback positions.
Whole-file processing: fixed GTCRN noise reduction with progress and cancellation. No user-adjustable reduction strength.
Output: after processing the entire file, save 16 kHz mono 16-bit PCM WAV. Video input does not produce a video output.
Japanese/English mobile and desktop UI. Microphone recording requires browser permission.

Limitations and notes

The tool primarily targets steady noise around speech. It does not guarantee removal of music, other speakers, heavy reverberation or sudden sounds.
Output is 16 kHz mono 16-bit PCM WAV only. Stereo channels, original sample rate and compressed encoding are not preserved; MP3/M4A export is unavailable.
Even with a video file as input, the tool cannot write video back out. It saves processed audio only.
The 10-second preview is for listening; full-file processing is required before saving. Live-call processing and speaker separation are not supported.
MP3, M4A, MP4 and other input support varies by browser. Files within the 250 MB and 60-minute limits may still exceed device memory or speed.
The microphone requires permission. Use HTTPS or localhost if your browser does not allow recording from a local file:// page.

Frequently asked questions

Can it remove all noise from a voice recording?

No. It aims to reduce steady noises such as fans or hiss around speech, but the voice may also sound different. Test a 10-second sample first.

Can I import MP3 and save MP3?

MP3 input depends on browser support, but output is only 16 kHz mono 16-bit PCM WAV. MP3 and M4A export are not supported.

Can I clean speech in an MP4 and save the video?

If your browser can decode audio from the MP4, you can process it, but the output is audio-only WAV, not a new video.

Can I export the 10-second test?

The 10-second sample is for previewing. Save WAV becomes available only after processing the entire file.

Can I clean a recording on a phone?

A supported mobile browser can import a recording or capture one with microphone permission. Long files may hit device memory or performance limits.

Are my recordings sent to a server?

No. The hosted page is downloaded initially, but audio files, microphone recordings, GTCRN noise reduction and WAV creation are processed in your browser.

Offline version and source code

The standalone HTML embeds the GTCRN model, WebAssembly and processing runtime. A supported browser can process files with no runtime CDN or external AI API; CSP includes connect-src 'none'. Microphone access from file:// varies, so use the HTTPS version if needed.

View source on GitHub