Building an Edge AI Pipeline to Automate Podcast Post-Production

Description and use

An Edge‑AI pipeline was built to automate podcast post‑production, reducing manual note‑taking from 45 min to 10 s and improving title quality through prompt engineering.

Problem

Podcast post‑production consumes nearly three hours per episode, draining creative energy and causing burnout.

Outcome

Structured show‑notes generation cut manual effort from 45 min to ~10 s generation, while title generation improved quality through prompt constraints.

Workflow

• 1. Record episode and upload audio
• 2. Transcribe audio (external tool)
• 3. Send transcript to Cloudflare Worker
• 4. Worker calls Llama 3 8B with structured system prompt to generate show notes, timestamps, resources, and SEO keywords
• 5. Receive JSON response and paste into CMS
• 6. Use separate title generator endpoint with prompt engineering for compelling titles
• 7. Review and edit brief outputs

Evidence

Transcribing the audio, writing structured show notes, brainstorming clickable (but honest) titles, and pulling SEO keywords was taking me nearly three hours per episode.

Source

I decided to build PodCrisp, a suite of AI podcast tools. For the architecture, I chose Cloudflare Workers.

Source

By explicitly defining the rules in the system prompt—like enforcing a 5-10 word limit and banning overpromising clickbait—the output quality improves dramatically.

Source

What worked exceptionally well was the structured show notes generator. It completely changed my workflow. What used to take 45 minutes of re-listening to the audio and typing out bullet points now takes about 10 seconds of generation and 5 minutes of light editing.

Source

I also added SEO keyword suggestions and an episode script outline generator to the suite, using similar structured prompting techniques to ensure the output is immediately actionable.

Source