Self-authoring podcasts (for free!)

This repository contains a set of self-written tools that I use when authoring my podcasts, as well as notes on the technical side of my process. I primarily keep this repository for my own sake – I don't particularly expect that the tools will be user-friendly enough to be a perfect option for most people, or that the processes will be applicable to everyone's skillsets and needs – but perhaps they could help someone in some way, so why not make it public! Especially worth noting is that this entire stack – including hosting – is entirely free, which ain't too shabby in this day and age.

At the time of writing, the tools and processes found here are used to make Obscuriosities. If you, dear reader, would like any advice on podcast authoring or otherwise have any questions, please do hit me up, for example via DM @obskyr on Twitter.

Tools

encode.bat: Encode a lossless audio file to the formats used in my podcasts. (You may have different needs, but the parameters found in this file may still serve as a base, and the comments within elucidate why I made some of the decisions I made!)
chapters-to-ffmpeg-metadata.py: Convert a label file exported from Audacity to an ffmetadata.txt, so that chapter marks can be added with FFmpeg.
crossfade-video.py: String together any number of video clips with a cross-fade effect with FFmpeg.

Workflow

Audacity

More or less the same procedure should be usable in most any other audio editing software as well. I have been considering migrating to Reaper…

Edit speech unprocessed; add music and what have you on separate tracks.
Copy the raw speech to a new audio track; mute the original one. (This ensures that your original recording will be available if ever you need to remaster it.)
Noise reduction.
Compression, step 1: Drop peaks. (I've been using TDR Kotelnikov for this.)
Equalization. (I've been using TDR Nova for this.)
Compression, step 2: Smooth out. (TDR Kotelnikov here as well.)
Loudness normalization to -18 LUFS (it really should be -16, by Apple Podcasts's standards, but I just couldn't compress well enough to make that work without clipping – worth investigating further).
Limiter: Hard limit; -1 dB.
File → Export Other → Export Labels… to export chapters. (If you have other label tracks, delete those temporarily while doing this.)
Export audio to FLAC.

Tips

To move a clip and all clips after it, select the clip (by clicking the bar on top of it) and press Shift + K to select to the end. (Make sure sync lock is turned on if you wish to move other tracks with it.) Shift + End does the same but also jumps the view to the end, and is thus impractical.
Use Punch Copy/Paste to insert room tone instead of silence. Always keeping a bit of room tone in a separate track may be helpful as well.
If using “Truncate silence” to speed up editing, don't go too hard on it, or you'll cut off breaths unnaturally! A threshold of -40 dB with a minimum duration of 3 seconds, truncating to 1 second, worked for my particular recordings.

Encoding

Drag-and-drop your label file (containing chapter marks) onto chapters-to-ffmpeg-metadata.py.
Drag-and-drop your FLAC file onto encode.bat.

Publishing

The above steps are applicable whatever you may be doing with the files afterward. Here's how I personally publish and host my podcast, however.

Self-publish the podcast using Jekyll to generate both the website and the RSS feeds. It's not by any means a plug-and-play solution – more suitable for people who are comfortable writing web front-end code (and perhaps even Ruby, though that's me going beyond the call of duty for minimal gains) – but it gives you full control, is entirely free, and makes you feel like a hearty DIY dad! The source is available on GitHub and is libre software, so feel free to fork and redesign it (you can drop in any Jekyll theme, really) to suit your own needs.
Host the website on GitHub Pages. Also free! The config for that can be found here. GitHub frowns upon serving too large files to too many people, however, so while you can host your audio files there (and I did, for a while), I now instead…
Host the audio files via a public bucket in Cloudflare R2, which is free up to 10 GiB of storage and a large number of “operations” (a download takes… some number of operations). You get a lot of downloads – and about 18 hour-long episodes – for free, and beyond that it's all prorated, so there's no sudden jump in price. You should, however, probably have your own domain if you choose to use this option – without one, R2 has rate limits and no caching.