You have an interview, a lecture recording or a voice memo with a fan, an air conditioner, a fridge or street noise behind the speech. You want the words intelligible without learning a DAW.
The short path
- Open the noise reducer and drop the recording in, or record straight into the page.
- Pick a reduction level and listen to the preview.
- Download the cleaned file as WAV, MP3 or AAC.
Up to five minutes per pass, nothing uploaded, no account. For a longer episode, split it into sections in any editor and run them one at a time.
What it fixes, and what it cannot
The tool is built for steady background noise — the kind that is present the whole time and does not carry information:
- Laptop and projector fans, air conditioning, ventilation hum
- Mains hum and electrical buzz
- Broadband hiss from a cheap microphone or a preamp turned up too far
- Wind and traffic rumble underneath speech
What it will not fix:
- Another person talking. Speech is not noise to a model trained on speech.
- A single loud event — a door slam, a cough, a chair scrape. Cut those out by hand; that is faster and cleaner than any filter.
- Clipping. If the recording was already distorted at the input, the information is gone and nothing downstream restores it.
- A room. Heavy reverb is not additive noise, it is the room's response convolved with the voice. Reducing noise does not remove a room.
Getting a better result
- Do this before editing, not after. Clean the raw take, then cut, level and mix. Cleaning a finished mix means the processing also sees your music bed and your other speaker.
- Start with the mildest setting that works. Aggressive reduction leaves speech sounding thin and slightly underwater — the classic over-processed podcast sound. Listeners notice it more than they notice a little hiss.
- Check the quiet parts. Artefacts are audible in the gaps between sentences, not while someone is talking.
- Keep the original. There is no undo after you close the tab, because nothing is stored.
Turn the cleaned recording into text
After cleaning, open AI Speech to Text to create interview notes or subtitles. Each file can be up to 5 minutes long. Review the transcript and export TXT, SRT, VTT or JSON.
Frequently asked questions
How do I remove background noise from a podcast recording?
Run the raw take through a noise reducer before you edit. Open the noise reducer, drop the file in, choose a reduction level, and download the cleaned audio. It runs in your browser, so the recording is not uploaded and no account is needed.
Will it remove a second person talking in the background?
No. The model separates speech from steady background noise, and a second voice is speech. Cut those sections manually.
Why does heavy noise reduction make voices sound strange?
Because speech and broadband noise overlap in frequency. Removing more noise necessarily removes some of the voice with it, which thins the sound and can add a watery quality. Use the mildest setting that gets the recording intelligible.
Can I clean an hour-long episode?
Not in one pass — each pass takes up to five minutes of audio, because the processing runs on your own device. Split the episode and run the sections in turn.