Give it a topic and get a full AI-generated podcast with two hosts having a natural conversation. Uses an LLM to write the script and Lightning TTS to voice each host.
- Generate a two-host podcast script from any topic using GPT
- Two distinct voices (male + female) with natural conversation flow
- Concatenates all segments into a single WAV file
- Saves the script as a text file alongside the audio
- Customizable host names, voices, and conversation length
Base dependencies are installed via the root
requirements.txt. See the main README for setup.
Requires both API keys in your .env:
SMALLEST_API_KEY=your-smallest-key
OPENAI_API_KEY=your-openai-key
uv run generate.py "The future of AI in healthcare"With custom options:
uv run generate.py "Space exploration in 2030" --exchanges 8 --host1-voice magnus --host2-voice sophia| Parameter | Description | Default |
|---|---|---|
--exchanges |
Number of back-and-forth exchanges | 6 |
--host1-name |
Name of host 1 | Alex |
--host2-name |
Name of host 2 | Sarah |
--host1-voice |
TTS voice for host 1 | magnus |
--host2-voice |
TTS voice for host 2 | sophia |
--model |
TTS model | lightning-v3.1 |
--llm |
LLM model for script generation | gpt-4o-mini |
podcast_output/
├── podcast.wav # Full podcast audio
├── script.txt # Generated script
├── segment_001_Alex.wav # Individual segments
├── segment_002_Sarah.wav
└── ...
- Audiobook Generator — Convert long text into a narrated audiobook
- Voices — Pick different voices for your hosts