Podcast Audio Quality: Recording to Final Export
Good podcast audio starts before the repair button. Follow the recording through editing, cleanup, loudness and export, with a check at each stage.
By the fix.fm team · Published · Updated
Start with a short recording test
Before recording an hour, record thirty seconds that includes normal speech, a louder sentence, a breath and a few seconds of quiet. Listen through headphones. The quiet reveals fans and hum; the louder sentence reveals overload; a pause after speech reveals how much the room keeps ringing. Save that test as a reference for the session.
Move the microphone closer to the speaker while keeping breath blasts off its front. The useful distance depends on the microphone, room and speaking style, so compare two positions instead of treating a single distance as a rule. A pop filter helps with breath blasts. Curtains, a rug and soft furnishings can reduce some reflections, but they do not turn a noisy room into a soundproof studio.
Record each remote speaker on a separate track when possible. That lets you adjust one person's level without changing everyone else. If your recorder offers a lossless format, keep that original. A repair tool can reduce unwanted sound; it cannot reconstruct an unrecorded microphone or guarantee recovery of a badly overloaded take.
Edit the conversation before its final level check
Remove false starts, private material and unwanted introductions first. Keep enough natural breathing room that the edit still sounds like a person speaking. When cutting at a nonzero sample, a short fade can soften a boundary click. It will not disguise a chopped syllable.
The audio trimmer keeps one continuous range, which is useful for a short clip or for trimming an introduction. A full episode with many cuts belongs in a timeline editor. Make a separate editing copy and keep the original takes, even if the first edit seems finished.
Match the repair to the sound
Do not apply every effect simply because it exists. Use the audio repair finder to identify the symptom. A steady hiss, electrical hum, gust of wind and room tail are different problems. Treating them all as noise can remove more of the voice than necessary.
| What you hear | Start here | What to check afterward |
|---|---|---|
| A constant hiss or fan | Background noise repair | Are consonants still clear? |
| A pitched buzz under speech | Hum repair | Are low voices still natural? |
| Low rumbles that surge | Wind repair | Is the voice becoming too thin? |
| Speech hanging in the room | Echo repair | Are words less smeared without a choppy tail? |
| Crunch on loud syllables | Distortion repair | Are those syllables clearer, or does a retake sound better? |
Our fix chain handles channel imbalance and clipping before noise, then room tail and tonal balance, with loudness last. This is the order implemented by our tool, not a claim that every editor must use an identical chain. Audacity's noise-reduction guidance similarly puts notch filtering before denoising and compression afterward. A simple uniform gain change is different from dynamic compression and can be applied at more than one stage.
Compare before and after at similar listening levels. A louder version can seem better just because it is louder. If a cleaned voice sounds watery, use less processing and consult why noise reduction makes a voice muffled or robotic. The room-tail repair is still beta; a very reverberant recording may need a new take.
Check loudness and peaks separately
Use the LUFS meter on the edited, cleaned file. Integrated loudness describes the whole programme; a peak reading tells you how close the waveform comes to its ceiling. A quiet average with a few loud peaks may need gentle dynamic control rather than a large uniform boost.
As of October 1, 2026, Apple Podcasts recommends overall loudness around -16 LKFS, within 1 dB, and a true peak no higher than -1 dBFS before encoding. Treat destination guidance as a delivery check, not permission to flatten every sentence. Spotify's music playback normalization is a different context, explained in LUFS vs dBFS.
Listen to the quietest speaker and the loudest section, not just the number. The quiet-audio fixer raises a quiet recording toward its speech target with compression and limiting, but it is not a multitrack speaker mixer. If one person is much quieter, adjust their separate track first.
Export a master, then a delivery copy
Keep a lossless master for future edits. Make a compressed delivery copy for the host or listener if that destination requires it. WAV vs MP3 explains what conversion preserves, while mono vs stereo helps you avoid doubling channels without a reason.
The audio converter exports a decoded recording as 16-bit WAV or 192 kbps MP3. It does not add chapters, artwork or publish to your podcast host. Those jobs belong in the host or an editor that supports them. Reopen the delivered file, check its peaks, and listen to both ends. An MP3 encode can alter peaks and add padding even when the editing master was correct.
A final check you can repeat
Keep a small checklist with the episode: correct content and cut points, both channels as intended, natural voice after cleanup, loudness suited to the destination, and a final file that opens. Test a short section on headphones and on the device your listeners commonly use. Small-speaker playback is useful for noticing unclear words, but it cannot reveal every low rumble or stereo problem.
Repeat the same checks each episode. A reliable workflow and a few reference recordings are more useful than chasing a new processing preset every week.