How skulux/voicetral works
Voicetral is a voice-driven chatbot: it listens for speech, transcribes it, gets an Ollama response, synthesizes that response with Bark, and plays the audio. The main loop also loads and saves conversation history. Most orchestration and audio processing live in main.py; short_term_memory.py provides JSON-backed persistence. The README documents the end-to-end workflow, while the sampled main.py excerpt does not show every call between stages.
Main entry points: main.py Main loop · config.ini Runtime settings · short_term_memory.py Conversation history
Components 12
Conversation
- Main loop
main.py - Conversation history
short_term_memory.py - Ollama model
Audio I/O
- Speech capture
main.py - Speech transcription
main.py - Bark synthesis
main.py - Audio resampling
main.py - Audio playback
main.py
Configuration
- Runtime settings
config.ini
Other components
- Speaker
- Microphone
- Audio output
Connections 12
- Speaker → Microphone: speaks into
readme.md - Microphone → Speech capture: provides audio
main.py - Main loop → Speech capture: listens via
main.py - Speech capture → Speech transcription: transcribes audio
main.py - Speech transcription → Main loop: returns text
readme.md - Main loop → Ollama model: requests response
readme.md - Main loop → Conversation history: loads and saves
main.py - Runtime settings → Main loop: configures
readme.md - Main loop → Bark synthesis: synthesizes response
readme.md - Bark synthesis → Audio resampling: writes audio
main.py - Audio resampling → Audio playback: provides resampled audio
main.py - Audio playback → Audio output: plays audio
main.py