Smart for Work
Making a Podcast with AI
This is Part A of Field Notes 26W45-EP06. People ask how Field Notes itself gets made. Fair question. So this half is about the principles I use — not a catalogue of every plugin and version. Part B — Floor Plan → Simple 3D covers how a floor plan becomes a simple 3D you can walk in your head.
Packaging vs judgment
Think of it as a reusable line, not a one-off chat. Text-to-speech voices, a music bed, subtitles and show notes — those are the packaging layer. The ideas, the cuts, and the final yes still belong to a person.
Tools change every few months. If I listed every name and build number here, the note would age in a week — and it still wouldn't tell you when to publish. What travels is the habit: draft with help, listen with ears, then wait for a human go. Without that last step, it isn't a Field Notes episode. It's just audio that happened to exist.
Voice
Start with voice. Field Notes uses clear text-to-speech hosts — disclosed as AI assistants. We're not pretending this was a quiet studio take by a person who then forgot to say so.
Clarity matters more than sounding "more human." If the line is hard to follow, rewrite it. Don't bury the meaning under a prettier fake.
Music
Music is the same idea in another key. Use what you're allowed to use — and credit it. Field Notes sticks with a Free To Use standing bed for intro and background. The links live in the notes; we don't dump every URL on air.
Credits aren't decoration. They're how you stay honest when someone asks "whose track is that?"
Subtitles and notes
Subtitles and show notes help people skim — titles, timestamps, the one-line takeaway. They don't replace listening. If the only way to get the point is reading the page, the audio still needs work.
So packaging serves the ear first. The page is the spare map, not the whole journey.
The edit loop and human GO
Then the edit loop. AI can draft an outline, a script, even a first listen. What it must not do is ship itself. I still review the outline, listen, and only then say go.
That go is the boundary. Draft early. Publish late. If something feels off in the headphones, it doesn't go out — even if the waveform looks tidy.
If I had to put Part A in one line: principles over a tool list. Packaging can be AI — voice, music, notes — and the judgment stays human. No human go, no Field Notes episode.
Continue with Part B — Floor Plan → Simple 3D. More notes are at kpoleung.com/notes/. Podcast (English · EP06): Field Notes with Kpo (Spotify) · Podcast (粵語 · EP06-CT): 稍後上架 / coming soon. 粵語版 EP06-CT（Spotify）獨立集。Comments or topic requests welcome at m@kpoleung.com.