Process YouTube Transcripts: Apify, OpenAI & Pinecone

This n8n workflow serves as a powerful backend for indexing YouTube video transcripts into a Pinecone vector database. It starts by retrieving video records from Airtable, including URLs and metadata. Next, it triggers an Apify actor to scrape transcripts with timestamps, updates the Airtable records with the raw transcript data, and then intelligently processes the transcripts by converting timestamps to seconds and chunking them into meaningful segments enriched with video details and direct links.

The workflow culminates in generating high-quality embeddings using OpenAI's models for each chunk and upserting them into Pinecone, enabling efficient semantic retrieval by a separate agent. This setup is ideal for building RAG (Retrieval-Augmented Generation) systems, content analysis tools, or AI-powered video search engines.

Benefits include significant time savings on manual transcript handling, scalable processing for large video libraries, and seamless integration of cutting-edge AI for vector search. Use cases span content creators organizing libraries, researchers analyzing video talks, marketers extracting insights from webinars, or developers prototyping multimodal AI applications.