Add a recording
Drop, choose or paste an audio file or a video: a voice memo, podcast, interview, lecture or clip.
Remove hiss, hum, fans, traffic and keyboard clicks from voice recordings, podcasts and videos. Listen to the original and the cleaned version side by side, then download audio or the video with its picture.
Open Audio Noise Remover →Drop, choose or paste an audio file or a video: a voice memo, podcast, interview, lecture or clip.
Click Remove noise. The AI cleans the speech on your device and shows the waveforms before and after.
Switch between Original and Cleaned while it plays, set the strength, then download WAV, MP3 or the video.
Use Manlin to reflect, plan and take your next step.

Yes. Clean as many recordings as you like with no account, no watermark and no daily limit.
No. The AI runs in your browser and your recordings never leave your device. The first time, your browser downloads the AI model (about 23 MB) once and keeps it for next time.
Steady and changing background sounds under speech: hiss, mains hum, fans and air conditioning, traffic, rain, keyboard and mouse clicks, and room noise. Other voices talking at the same time are harder to separate and are only partly reduced.
Yes. Drop an MP4, MOV, WebM or MKV file. The picture is copied exactly as it was, and only the sound is replaced with the cleaned version. You can also download just the audio.
It sets how far the noise is turned down, from 6 dB to maximum. Lower values mix a little of the original back in for a more natural sound; the change is instant, without cleaning the file again.
It makes the result as loud as typical podcasts and voice-overs (−16 LUFS, measured the EBU R128 way) and keeps peaks below clipping, so you don't have to adjust the volume by hand.
On a typical laptop, a minute of mono audio takes about 5 to 15 seconds. Stereo recordings take about twice as long. You can keep using the page while it works.
Recordings up to 30 minutes. Stereo recordings longer than 15 minutes are cleaned as mono to keep memory use reasonable. Split longer files and clean them one by one.
The AI is trained to keep speech natural, but very noisy recordings can come out slightly muffled. Use the Original/Cleaned switch to compare, and lower the strength if the voice suffers.