The Idea
Most podcasters publish an episode and never turn it into content again. But a 45-minute episode is a goldmine of search-friendly material if you can transcribe it cleanly and restructure it into something Google can read.
PodcastBrain does exactly that. You drop an MP3 in a folder. A few minutes later there's a fully structured, SEO-optimized blog post in WordPress, with timestamped chapter markers linking back to the audio.
Pipeline
- Watch a folder for new audio files (uploaded by the host's RSS feed).
- Transcribe with OpenAI Whisper, with speaker diarization.
- Summarize each segment with GPT-4 and identify natural chapter breaks.
- Generate an SEO title, meta description, and FAQ block from the transcript.
- Push the final markdown to WordPress via the REST API.
- Notify the host on Slack with a draft preview link.
The Stack
What Worked
The single best decision was treating the transcript as the source of truth and never throwing it away. Every downstream artifact, summary, title, FAQ, references back to specific timestamps in the transcript, so when a host pushes back on a generated post, we can show them exactly which 30-second clip the claim came from.