AI automation pipeline for Podcast

Working closely with Pamela's team, we were tasked with transitioning hundreds of radio episodes of her shows to a podcast format.

In collaboration with Pam’s team, we designed and deployed a fully automated local workflow to transform raw podcast recordings into publish-ready episodes on Transistor — complete with titles, descriptions, and summaries. The goal was to eliminate manual steps, reduce turnaround time, and keep all data processing fully local for privacy and reliability. The solution was built around a self-hosted n8n instance that orchestrates every stage of the process. When a new audio file is dropped into the system, OpenAI Whisper transcribes it locally with high accuracy. The resulting transcript is then processed through Ollama and OpenAI language models to automatically generate editorial assets: concise summaries, SEO-friendly titles, and formatted show descriptions. To complete the workflow, we developed a custom n8n node for Transistor’s API, enabling direct upload of each episode and its metadata without leaving the local environment. This integration ensures consistency across all published episodes and allows for easy scalability as the show’s production grows. The collaboration between our team and Pam’s editorial group was key — their insights into narrative flow and audience tone helped refine the AI prompts and metadata templates, ensuring the automated output matched the brand’s voice. The result is a production pipeline that is efficient, maintainable, and human-aligned: what once took hours of manual work can now be accomplished in minutes, with full editorial control retained.