Introduction
The idea was to create an AI-generated podcast show where the news narrator could present daily news updates, invite guests to share their horror stories, and provide updates about the Olympics and sports. The next plan is to generate podcasts specifically for kids, where they can learn new things.
Pipeline
- Scrape news from a base website using an in-house web scraper
- Convert news titles into queryable text and gather the top 5 results on each topic from different sources
- Summarize collected articles using LLaMA 3 (local) and GPT-4 API
- Generate a complete narration script with AI, including structured segments and transitions
- Convert the script to MP3 using a text-to-speech model with AI-generated intro and outro
- Assemble the final podcast episode and prepare for release
Key Technologies
- LLaMA 3 running locally for script summarization and generation
- GPT-4 API for higher-fidelity script generation and quality control
- In-house web scraper with multi-source aggregation for comprehensive news coverage
- Text-to-speech model with voice cloning for a consistent, realistic narrator voice
- Automated MP3 assembly pipeline for complete episode production