Cleaning up audio for a podcast
Cleaning up audio for a podcast means doing three small jobs once the episode edit is locked: trimming the dead air at the start and end, levelling the whole file to a publishing loudness of about -16 LUFS, and adding short fades in and out. Noise reduction, EQ and balancing the speakers against each other belong earlier, in a full audio editor.
Most of what people call audio cleanup is really three small jobs at three different moments in the week. Knowing which moment each one belongs to saves more time than any tool does, because the expensive mistake is doing them in the wrong order.
The shape of the week
A published episode passes through roughly the same five stages whatever the show: record, assemble, edit for content, level for publication, upload. The two stages people conflate are the middle ones, and they have different rules. Editing for content is a judgement call made while listening. Levelling for publication is a measurement made once, at the end, on the file that is about to leave your machine.
Conflating them is what produces the familiar mess: an episode levelled halfway through, then edited further, then levelled again on top of itself, arriving somewhere nobody chose.
Before you record
The decisions worth making in advance are the ones no tool can undo. Microphone distance, input gain with headroom to spare, and a quiet room are the entire difference between a file that needs thirty seconds of work and one that cannot be rescued at all. If you take one habit from this page, take recording a few seconds of silence at the top of every session: it tells you where your noise floor sits, and that number decides what is possible afterwards.
Remote guests are the recurring problem, because you control none of it. Ask for a local recording rather than the call audio, and expect their file to arrive at a different level from yours. That mismatch is a levelling job, not a mixing job, and it is the one case where it is worth touching individual tracks before you assemble anything. Why your audio is too quiet covers what to tell a guest whose file arrives faint.
While you edit
This is the stage where you are listening for content: the false start, the tangent that went nowhere, the question you asked twice. It is also where you decide the balance between speakers, so that your guest is not a step behind you for forty minutes.
What does not belong here is the publication level. Setting it now means re-doing it later, because every cut you make afterwards changes what the measurement sees. Leave it. The episode is not the episode until the last cut is made.
If you film the episode as well, the video edit is its own job with its own tool, and the dead air is the part worth automating. CutNoodle for podcasters covers cutting the silences out of a filmed episode; nothing on this page applies to that.
When the episode is locked
Now the file stops changing, and now the three small jobs are worth doing, in this order:
- Trim the ends. The countdown before you started and the shuffle after you finished are both still there. They are also both measured as part of your episode, so cutting them first means the level that follows is calculated on the programme rather than on the programme plus two minutes of room tone.
- Level the whole thing. One measurement, one offset, applied to the locked mix. The target is not a matter of taste: how loud should a podcast be is the reference for the number and for what the platforms do with what you give them.
- Fade in and out. A half-second at the top stops the episode starting on a click, and a second or two at the end stops it ending on one. Small, and the difference between an episode that sounds published and one that sounds exported.
All three are one pass in the Simple Audio Editor, and normalize audio online walks through the controls if you want to see them before you open a file. The whole thing runs in the browser tab, which matters more than usual here: an unreleased episode with an embargoed guest never goes to anybody else's server, and a ninety minute recording is ready the moment you pick it because there is no upload.
What this does not replace
Be realistic about the boundary. The Simple Audio Editor cuts, removes pauses, sets the volume of each clip, levels and fades. It has no noise reduction, no equaliser, no compressor beyond the levelling itself, no de-esser, and no multitrack mixing. If your recording needs any of those, it needs a proper editor, and no amount of levelling will stand in for one.
What it covers is the last mile: the part that is the same every week, that nobody enjoys, and that listeners notice immediately when it is skipped. Doing it as a fixed routine at the end of the week is worth more than doing it well occasionally.
Common questions
Should I level each podcast guest separately or the finished episode?
Both, in that order, and for different reasons. Balancing speakers against each other is an editing decision made per track. Levelling the finished mix to a publishing target of about -16 LUFS is a measurement made once, on the file you are about to upload.
Do I need to normalize my podcast if the hosting platform normalizes it anyway?
It is worth doing. Platforms adjust playback toward their own target, but they do it to whatever you uploaded, noise included, and not all of them raise a quiet file. Arriving at the target means the episode sounds the same everywhere instead of depending on who is playing it.
Can I clean up individual podcast tracks before mixing?
You can, and it is a reasonable way to get two remote guests into the same neighbourhood before you start editing. In CutNoodle's Simple Audio Editor, normalizing is a single level offset per file, not a mix. A passage that is too quiet or too loud can be split into its own clip and given its own volume, but anything beyond that belongs in a full audio editor.
Is my podcast audio uploaded when I clean it up in the browser?
No. CutNoodle's Simple Audio Editor opens the episode from your disk and processes it inside the browser tab, so an unreleased episode never reaches anyone else's server. It is free, needs no account, and opens MP3, WAV, M4A, AAC, OGG and FLAC files up to 2 GB.
How do I remove silence from the video version of a podcast?
A filmed episode needs a video tool. CutNoodle's silence remover finds the dead air in a video by loudness, shows the cuts on an editable timeline and exports an MP4, all in the browser without uploading the file.
CutNoodle is free and runs in your browser. Open the Simple Audio Editor and pick a file.