How to level the podcast audio? Ask Claude

Transcript

This is Voice Notes to the World, and I’m John-Paul Flintoff.

I want to make a note to myself, as much as anything, about how I have achieved something with the podcast, and why, using Fable, which is the current top level of Claude.

I typed the following into Claude:

I have started making a podcast using an iPhone 17e and my Apple M1 Mac, sometimes using a Blue Snowball plug-in mic. I’ve also ordered a plug-in lavalier mic with fluffy cover for recording voice notes into my phone.

My podcast is called Voice Notes to the World, and it includes real actual voice notes sent by me to others, often, but not only, using WhatsApp. Otherwise, I mostly record with the iPhone voice memo function, which comes out as an M4A file. On the Mac, I record with the Blue Snowball into Audacity 4, and I export MP3 files.

I save the audio files to my desktop to upload to my own website, which has a feed that I’ve successfully shared to Spotify and, not yet successfully, to Apple Podcasts. You previously wrote me a bit of code that creates transcripts using Whisper with one click, and I upload those to the posts on my site that go out with each episode.

My point: I’ve now made a lot of episodes and the audio quality is awfully variable. I don’t know yet how the lavalier mics will help, but I want to know if there’s a simple routine that I can follow to adjust every recording to a basic, decent level. Most importantly, is it something you can help with? And can it be at least somewhat automated?

What I didn’t tell Claude, because it didn’t seem relevant, but I can say to you, dear listener, is that I got the idea for those lavalier mics from my daughter, who’d bought a rather fetching pink version with pink fluffy bits. But that’s irrelevant to Claude, so I didn’t mention it.

Also, I’m recording with the Blue Snowball into Audacity 4 right now, as I speak.

Claude then did a whole load of stuff, which I’m not going to bother repeating. It kept on saying “I’m doing this, I’m doing that”, commentating on its own work. And then it announced:

The action is built and tested. Here is what you have now.

The routine: select any recording in Finder, right-click, choose Quick Actions, then Level for Podcast. A notification says it has started and another says it is done, telling you what loudness the original was. The new file appears beside the original with Leveled added to its name. The original is never touched and the action never overwrites, so running it twice just adds the time to the second name. Then upload the Leveled file and transcribe it with the Whisper action as before.

What it did to your four desktop files: one recording was 11 decibels quieter than the standard, which is the sort of gap that a listener hears as “turn it up, now turn it down”. All four now sit at the same level, with peaks held safely below clipping. Have a listen to the four leveled files on your desktop.

The noise reduction is set gently on purpose, so a bit of room should remain. If it sounds too processed or not clear enough, that’s one number to change at the top of the script.

For recordings in Audacity, export WAV files rather than MP3 from now on. The action accepts MP3 too, so nothing breaks if you forget.

One thing to check yourself: the first time you right-click an audio file, confirm Level for Podcast appears in the Quick Actions submenu. It should, since it was registered the same way as the Whisper action, but macOS occasionally needs a log out before new actions show up.

And then Claude told me where all the files are stored.

I think this may be one of the most boring podcast episodes I’ve ever recorded. But one of the points of doing it in the first place was to make notes about work in progress, and the kind of boring in-betweeny bits.

So I’m really chuffed, not only to have done this, but to be speaking out loud a reminder to myself about how I’m to proceed in future.

Anyway, I’m John-Paul Flintoff. This has been Voice Notes to the World. Thank you very much for listening.

jp at this domain name.