My New Podcast Generating Workflow

In Accessibility First in Podcasts, I wrote that since my podcasts are scripted, I don’t have to work hard to get a transcript.

But I found that writing a script was both hard and resulted in a podcast that sounded written. Even if I memorize and perform it well, it didn’t sound like my spoken “voice”.

So, I decided to

  1. Start with a rough outline
  2. Make a recording of a lot of extemporaneous speaking on the subject
  3. Make a transcription of the recording using Whisper from OpenAI
  4. Edit the transcription into a coherent story, but try to preserve the phrasing
  5. Practice it
  6. Make a recording that basically follows the script. It’s ok to make mistakes, rephrase, or veer off.
  7. Edit the recording to remove mistakes and reduce overly long pauses
  8. Listen to the recording and fix the script so it’s now a transcript.

The key thing is step #4 which helps me make a script that sounds like me talking (not writing).