I record podcast interviews in rooms that were never built as studios. A heater clicks on mid-sentence, a laptop fan ramps up under the desk, and the guest's keyboard lands right on the consonants. Full audio cleanup can fix a lot of that, but it costs money and time, and some tools flatten a voice if you commit the whole file before you have heard the tradeoff.
The habit I use now is a short preview before I pay for a full pass.
What I listen for first
I do not start by processing the entire episode. I pull three short clips, about ten seconds each:
- A quiet pause where only the room is audible.
- A sentence with sharp s and t sounds.
- A moment where two people laugh or talk over each other.
Those three clips tell me more than a waveform overview. The pause shows the noise floor. The sentence shows whether reduction starts eating the voice. The overlap shows whether transients get smeared.
Preview, then decide
I run only those clips through a noise-reduction preview, then A/B them against the raw audio at the same loudness. I am listening for four failures:
- The room tone drops, but the voice picks up a watery tail.
- Breaths get chopped into clicks instead of staying soft.
- Noise pumps back in the moment the speaker pauses.
- The two speakers become harder to tell apart.
If any of those show up, I lower the amount of reduction or I leave that section alone. A slightly noisy interview is easier to live with than a voice that sounds processed on headphones.
Why the preview comes before the purchase
Paid cleanup is usually priced per file, per minute, or as a subscription. A one-minute preview answers the only question that matters before I spend: is this recording worth a full pass, or should I re-record the intro and clean only the interview?
I keep the raw file. The preview is a decision, not the master. When the preview sounds right, I apply the same setting once to the whole episode, then do a normal edit: trim, level, and a light limiter. I do not stack a second noise reducer on top "just in case." Two reducers in a row are how a take turns metallic.
Where I run the short listen
When I want to hear the tradeoff before I commit to a paid cleanup, I use EraseNoise: https://erasenoise.com/ . I upload a short excerpt, compare before and after on the same headphones, and only then decide whether the rest of the episode deserves the same treatment. The tool does not replace judgment. It lets me hear the compromise on a small sample while the session is still fresh.
Checklist
- Export a 30 to 60 second excerpt, not the full multitrack.
- Match playback volume so "louder" does not masquerade as "cleaner."
- Check one pause, one sharp consonant sentence, and one overlap.
- Write down the setting that passed, and apply it once to the full file.
- Archive the unprocessed original.
That is the whole method. Preview noise reduction on a slice you can judge in a minute, then pay for full audio cleanup only when the voice still sounds like the person you interviewed.