How to remove background noise from audio

Why does my audio sound bad?

Put on headphones, play the recording and name what you hear before you reach for a plug-in. Most of these problems are cheaper to solve with distance and the room than with software.

How do you reduce echo in a room before you record?

Echo is your voice bouncing off hard surfaces and reaching the microphone a moment later. Soft and uneven surfaces break it up, so start with the room you choose and what is in it:

• pick the smallest room with the most furniture in it;
• use whatever soaks up sound: a rug, curtains, a sofa, cushions, full bookshelves;
• hang a duvet or a couple of blankets behind and beside the speaker;
• stay out of empty rooms, kitchens and bathrooms;
• sit away from bare walls that face each other.

Then try the clap test. Clap once, sharply, and listen. If the sound rings on for a moment, the room still echoes. A short, dull clap means it is ready.

The other fix is distance. Halve the gap between your mouth and the microphone and your voice becomes much louder compared with the room, so the same echo is far less audible.

Two different jobs often get mixed up here. Acoustic treatment controls reflections inside the room, with soft furnishings, blankets or absorbing panels. Soundproofing stops sound travelling through walls, windows and doors, and that is building work: mass and sealed gaps. Foam panels reduce echo, but they will not keep out the neighbours or the street.

If the room cannot be fixed, record somewhere that is set up for it. In Riga you can record your podcast at Baltic Podcast Studio, where each speaker has their own professional microphone and the sound is mastered.

How do you set up the microphone and the levels?

Run through this before every recording, even a short one:

1. Place the mic at the right distance: about 5 to 10 cm from the mouth for a dynamic mic, 15 to 30 cm for a condenser, which also needs a quiet room. Not sure whether yours is dynamic or condenser? The equipment guide explains the difference.
2. Aim it at your mouth from slightly to one side, so the air from P and B sounds goes past the capsule.
3. Turn off automatic gain and any voice enhancement setting in phone and call apps.
4. Set the gain while speaking as loudly as you will during the recording, laughter and emphasis included. Aim for peaks around -12 dBFS.
5. Record at 24-bit, so quiet passages can be raised cleanly later.
6. Make a 30-second test recording and play it back on headphones.
7. Record 10 to 20 seconds of silence in the room. This room tone gives noise reduction a sample to learn from.

On a phone, hold it close to your mouth or plug in a clip-on mic. Switch to aeroplane mode and record in the quietest corner you can find.

Narration has a few rules of its own: see recording a voice-over at home, or record a voice-over in a studio when the job calls for it.

How do you remove background noise from a recording?

Noise reduction works on steady sound: hiss, a fan, air conditioning, the hum of a fridge. The tool learns what the noise sounds like from a passage where nobody speaks, ideally your room tone, and subtracts it from the whole track. Apply it gently. Pushed too far, it leaves the voice watery or robotic.

The order of the steps matters:

1. Cut the loud distractions first: bangs, coughs, a phone buzzing on the table.
2. Add a high-pass filter around 80 to 100 Hz to take out rumble.
3. Remove hum if there is any. Mains power in Latvia and the rest of Europe runs at 50 Hz, so set the de-hum to 50 Hz and its multiples (100 and 150 Hz), not the 60 Hz you will see in American guides.
4. Apply noise reduction, lightly.
5. Use EQ for clarity.
6. Compress to even out the loud and quiet passages.
7. Normalise the loudness as the very last step.

To clean up the sound of a video, detach or export the audio, clean it and put it back. The steps are the same in the editors most people already have, whether that is Audacity for sound or DaVinci Resolve, Adobe Premiere Pro or CapCut for video. If there is video too, plan the sound for a filmed podcast before the shoot, and you will have far less to clean.

Keep the original file. Every cleanup is a compromise, and you may want to go back and try lighter settings. If you send your recording for editing, send the untouched original, not a copy you have already cleaned. The FAQ explains how long editing takes and how many revisions are included.

What can AI audio cleanup fix, and where does it fail?

AI speech enhancers separate a voice from steady noise and some room echo, often more cleanly than classic noise reduction. They also make mistakes of their own:

• the voice can sound artificial and over-processed;
• consonants and breaths get swallowed;
• music and laughter start to wobble;
• when two people talk at once, the tool can lose track of who is speaking.

No tool can recover detail that was never recorded. A voice far from the mic, heavy clipping or speech under loud music will stay that way. When you do use an AI enhancer:

• compare the result with the original on headphones;
• if the tool allows it, blend some of the original back in, which keeps the voice natural;
• listen closely to names and to the quiet ends of sentences, where words go missing.

Why does my audio sound worse on YouTube or a phone?

Platforms adjust the playback loudness. A very quiet upload sounds weak next to everything else, and a heavily compressed one sounds flat once the platform turns it down. Around -16 LUFS is a common target for spoken-word podcasts, and video platforms tend to bring louder content down to a reference level, often near -14 LUFS. These references change, so check the current figure before you rely on it.

Check the mix in mono as well. The voice should sit in the centre, and it should not thin out or disappear when a phone plays the track through a single speaker.

Export from your editor at a good quality setting. Each time a file is compressed again, for example when you download it, edit it and upload it once more, it loses a little more quality.

Phone speakers reproduce almost no bass. If the voice only sounds good on big headphones, it has too much low end and too little clarity, and a phone will make that obvious.