AI Podcast Generator: Produce a Show from a Script, Open Source

An AI podcast generator that turns a script into a finished episode: TTS voices, music beds, and editing with open-source tools.

💡 What You Will Learn

An AI podcast generator that turns a script into a finished episode: TTS voices, music beds, and editing with open-source tools.

Starting a podcast costs time and money - or it costs one evening with open-source tools. The quality gap to studio recording is closing fast.

The Pipeline

1) Write the script (an LLM via Ollama, 177,874 stars, drafts it). 2) Generate voices with Coqui TTS (45,862 stars) or Piper (11,275 stars) - multiple voices for host/guest. 3) Add a music bed with Audiocraft (23,540 stars). 4) Mix and master with ffmpeg (63,025 stars). A 20-minute episode renders in about 30 minutes on a mid-range machine.

What Works and What Does Not

Explainer and news-style shows work well - scripted, single-voice, minimal improvisation. Unscripted banter does not: TTS cannot do overlapping natural conversation yet. For long-form, generate per-segment and stitch - shorter clips sound more natural.

FAQ

Q: Do I need to disclose AI voices?
A: Best practice - and many platforms now require it. Label synthetic content clearly in the show notes.

Q: Can I clone my own voice for the show?
A: Yes - Coqui XTTS clones from 30 seconds of your speech; keep the consent trail if guests are involved.