This is the lesson worth reading before you spend anything. AudioPod's noise reduction does one thing well and several things not at all — and crucially, the things it cannot do do not fail. There is no error, no warning and no badge: the job finishes, the credits go, and you are left listening to a recording with the same problem it had before, minus a little of the voice.
Steady noise and everything else
Almost everything about whether a cleanup will work comes down to one distinction, and the tool never mentions it.
- Kind of noise
- What to expect
- Steady background noise — mains hum, tape hiss, a fan, air conditioning, traffic drone, a computer
- Noise that is there the whole time and barely changes. This is what noise reduction is for and what it does best, because a constant sound is easy to tell apart from a voice that starts and stops.
- One-off and intermittent noise — a door slamming, a cough, a chair scraping, a notification, music behind the speech
- Noise that happens once, or comes and goes, or moves the way speech moves. This is the hard case: something short and loud is more likely to leave a hole in the audio than to disappear cleanly, and music behind a voice is a separation problem rather than a noise problem.
The distinction that predicts the result.
Tip
A quick test: if you can hum the noise, it is steady, and this is the right tool. If you would describe it as "and then someone…", it is not.
The four it cannot touch
These are not edge cases. They are the four most common reasons a recording sounds bad, and all four are outside what removing background noise can do.
- What is wrong
- Why cleanup does not help
- The recording is distorted because it was recorded too loud.
- Clipping is not noise added on top of the voice — it is part of the voice that was never captured. There is nothing underneath to uncover. Re-record with the input level lower. If you cannot, accept the distortion: a cleanup will remove some hiss around it and leave the distortion exactly where it was.
- It sounds like it was recorded in an empty room, or a bathroom.
- Reverb is the voice itself arriving late off the walls. It moves when the voice moves and stops when the voice stops, so it does not look like background noise at all. Record somewhere softer — a room with a rug, curtains and furniture, or a closet full of clothes — and get closer to the microphone. That change is worth more than any amount of processing.
- The voice is thin, boxy or muffled to begin with.
- Noise reduction takes things away. It cannot put back a top end that was never recorded, and taking more away makes a dull recording duller. Fix it at the source: a better microphone, or the same microphone closer to the speaker and out of the path of a fan. Then clean up what is left.
- The background is louder than the person talking.
- There has to be more voice than noise for the difference to be usable. When the speech is underneath, what comes back is quiet, patchy speech with the noise mostly gone — which is rarely what anyone wanted. Re-record if you can. If it is a recording you cannot repeat, run it and judge the result against the original honestly; sometimes the noisy version is the more listenable one.
Recordings a cleanup will not rescue, and what to do instead.
Careful
The voice is thin, boxy or muffled to begin with. Noise reduction takes things away. It cannot put back a top end that was never recorded, and taking more away makes a dull recording duller. This is the one that catches people out, because a thin recording and a noisy recording sound equally "bad" until you ask which one you are actually hearing.
- What you want
- Where it lives
- You want to pull two people, or a song's instruments, apart.
- Noise reduction returns one file. It has no concept of who is speaking and no concept of an instrument, so a second voice is not noise to it. Those are different tools: speaker separation for a conversation, the stem splitter for a piece of music.
- You want it to sound louder, fuller or more produced.
- This tool removes; it does not level, compress, EQ or master. A cleaned file is usually a little quieter than what went in, because something was taken out of it. Do the cleanup here first, then take the result into the editor for levelling — the completed job has a button that hands it straight over.
Not noise problems at all.
The second one has a path built for it. Every completed job carries an "function(){throw Error("Attempted to call OPEN_IN_STUDIO_LABEL() from the server but OPEN_IN_STUDIO_LABEL is on the client. It's not possible to invoke a client function from the server, it can only be rendered as a Component or passed to props of a Client Component.")}" button, which takes the cleaned audio straight into the editor instead of making you download and re-upload it. So the natural order is cleanup here first, then levelling and editing there — and the "function(){throw Error("Attempted to call OPEN_IN_STUDIO_LABEL() from the server but OPEN_IN_STUDIO_LABEL is on the client. It's not possible to invoke a client function from the server, it can only be rendered as a Component or passed to props of a Client Component.")}" button is what saves you a download and a re-upload in between.
When to record it again
Say it plainly, because a tool's own guide rarely does: if you can record it again, and the problem is the room, the microphone or the level, record it again. Ten minutes and a softer room beat any amount of processing, and you end up with a recording that has nothing taken out of it.
- Move somewhere with soft things in it — a room with a rug, curtains and furniture, or a closet full of clothes. Hard empty rooms are what put reverb in a recording, and reverb is the one problem on this list that no processing removes.
- Get closer to the microphone. A voice recorded close is louder relative to everything else in the room, which is the same thing a cleanup is trying to achieve — except it costs nothing and removes none of the voice.
- Turn the input level down until nothing distorts. Distortion is not noise sitting on top of the voice; it is part of the voice that was never captured, and there is nothing underneath it to recover.
- Turn off the fan, the air conditioning and the fridge. The steady noise this tool handles best is also the steady noise that is easiest not to record in the first place.
When it is worth running
None of this makes the tool a bad one — it makes it a specific one. It earns its credits on a recording you cannot make again, where a constant sound sits under speech you can already hear clearly.
- Recording
- Why it works
- An interview recorded in the wrong room
- A conversation with an air conditioner running through it. The hum is constant, the voices are not, and removing it is the single change that makes the recording publishable.
- A remote call with line noise
- Hiss and electrical buzz picked up on the way. It sits underneath the whole call at the same level, which is exactly the case that cleans up best.
- Field and phone recordings
- Traffic, wind rumble or a room's background collected by a phone in a pocket. Worth running before you decide whether the recording is usable at all.
- Prepping audio for something else
- A cleaner recording is a better input to everything downstream — an edit, a transcript, a voice you are about to work with. Running the cleanup first is usually cheaper than fixing the consequences later.
The cases people bring here that it genuinely solves.
Note
Prepping audio for something else is the use people underrate. A cleaner recording is a better input to everything downstream — an edit, a transcript, a voice you are about to work with. Running the cleanup first is usually cheaper than fixing the consequences later.
That is the honest shape of Noise Reduction: one recording in, one cleaner recording out, brilliant at a hum and helpless against a bathroom. Knowing which of those you have before you upload is the whole skill, and it is worth more than any setting in the menu.