How an Episode Gets Made

Every episode of this show is written from my own thinking and spoken in a clone of my own voice. Here's the whole process, start to finish.

1. I talk into a voice memo

I open Apple Voice Memos and just start talking about the topic. No outline, no script, no second takes. I stumble over words, forget things and circle back to them, and let background noise happen. The only goal is to get everything I'm thinking out loud.

Voice Memos transcribes it automatically, so I have text the moment I stop recording.

2. Claude turns the transcript into a script

I paste the transcript into a Claude project I set up for the show. The project holds the guardrails: the episode format, the intro I use every time, and a running "do not use these words" document that keeps out the phrasing that makes writing sound like it came from a machine.

What comes back is usually about 95% of a finished script.

3. ElevenLabs reads it in my voice

The voice is a clone of mine. I built it by recording about two hours of myself talking about random things across several clips. No reading from a screen, just speaking, so the model learned how I actually sound.

For each episode I duplicate the previous ElevenLabs Studio project, which means the theme music is already in place, and paste in the new script. I use the Eleven Flash v2.5 model with my voice.

Then I listen to the whole thing. Sometimes a phrase lands with an odd cadence, and I fix it by editing the text rather than re-recording anything. Occasionally there's a glitch that needs a regeneration. I nudge the intro music forward or back a few beats and the episode is done.

I export two files: an MP3 of the audio and a VTT transcript with timestamps.

4. One command publishes it

The rest runs through a pipeline I built in Claude Code. I drop the MP3 and the VTT into their folders and run a single command. It asks me for the episode number, the title, the publish date, and whether the episode is explicit.

From there it handles everything: uploading the audio, writing the show notes from the transcript, placing the transcript, building the episode page on this site, and updating the RSS feed. Podcast apps pick it up right away.

The theme song

Generated with Suno. I described the kind of podcast theme I wanted and the style, and it produced a clip I've used ever since.

How long it takes

Idea to live episode, under thirty minutes.

The voice memo takes about five minutes. The script comes back almost instantly. The ElevenLabs pass takes around ten minutes, and most of that is just listening to a seven-minute episode straight through to catch anything off.

About the voice

I have a monotone voice to begin with, so the clone works well. ElevenLabs offers models with more animation and character, and I deliberately picked the one that sounds most like me instead.

I'd rather you know it's AI. I could spend another pass polishing each episode to the point where nobody could tell, and the tools are good enough now to do that, including full multi-voice conversations. I'd rather trade that polish for speed and get episodes out the door.