How I Automated My Podcast Post-Production Without Losing Sleep

Description and use

Automating podcast post‑production with Cloudflare Workers and Llama 3 8B Instruct reduced episode prep time to seconds, delivering structured show notes and titles.

Problem

The author spent about 40 minutes per episode writing show notes, titles, and metadata, which was more time than creating content.

Outcome

Show notes and titles are generated in a few seconds, with structured markdown that requires minimal reformatting, saving the author from manual transcription work.

Workflow

• 1. Capture raw transcript from recording
• 2. Send transcript to Cloudflare Worker via JSON request
• 3. Worker calls Cloudflare AI inference with Llama 3 8B Instruct model and a strict system prompt to generate structured show notes (summary, key takeaways, timestamps, resources)
• 4. Worker returns JSON response
• 5. Use generated markdown directly in podcast hosting platform
• 6. Separate prompt for title generation with tone variable to produce numbered list of titles

Evidence

Writing good show notes is not just about summarizing what was said. It is about structuring the content for search engines, highlighting key takeaways, and providing timestamps so listeners can navigate the audio easily.

Source

I decided to build PodCrisp using Cloudflare Workers, leveraging their AI inference capabilities at the edge.

Source

The results have been genuinely helpful for my own workflow. What works exceptionally well is the consistency of the structured output.

Source