How whenkey2007/true-tone works
TrueTone detects AI-generated speech from live microphone or system audio, WAV uploads, and replayed files. Its principal workflow captures audio, prepares it for speech gating and model/feature scoring, stabilizes scores over time, then presents live results and alerts in the dashboard. The graph preserves the documented capture, analysis, and evaluation capabilities; unsampled feature and preprocessing connections are based on README descriptions, while code-level edges cite the excerpts or SOURCE INDEX relationships available here.
Main entry points: orchestrator.py Threaded Orchestrator · app.py Application Modes · detector.py AI Detector · live_pipeline.py Standalone Live Pipeline
Components 16
Audio Sources
- Microphone Capture
audio/mic_capture.py - System Loopback
audio/system_capture.py - WAV Replay
audio/wav_loader.py - Audio Loading and Chunks
audio/wav_utils.py - Sliding Audio Buffer
audio/buffer.py
Signal Analysis
- Audio Preprocessing
processing/preprocessor.py - Speech Gates
audio/vad.py - DSP Features
processing/features.py - AI Detector
inference/detector.py
Pipeline Runtime
- Application Modes
app.py - Threaded Orchestrator
pipeline/orchestrator.py - Standalone Live Pipeline
live_pipeline.py - Temporal Aggregation
pipeline/temporal.py
User Experience
- Dashboard and Alerts
ui/dashboard.py
Evaluation
- Streaming Evaluation
tools/evaluate_streaming.py
Other components
- User
Connections 31
- User → Application Modes: selects mode
README.md - Application Modes → Microphone Capture: configures capture
app.py - Application Modes → System Loopback: configures capture
app.py - Application Modes → WAV Replay: configures replay
app.py - Application Modes → AI Detector: creates detector
app.py - Application Modes → Speech Gates: creates gate
app.py - Application Modes → Threaded Orchestrator: starts pipeline
app.py - Microphone Capture → Threaded Orchestrator: supplies chunks
- System Loopback → Threaded Orchestrator: supplies chunks
- WAV Replay → Threaded Orchestrator: supplies chunks
- WAV Replay → Audio Loading and Chunks: loads and chunks
audio/wav_loader.py - Threaded Orchestrator → Audio Preprocessing: prepares audio
README.md - Threaded Orchestrator → Speech Gates: gates speech
pipeline/orchestrator.py - Threaded Orchestrator → AI Detector: requests scores
pipeline/orchestrator.py - AI Detector → DSP Features: extracts features
inference/detector.py - AI Detector → Audio Loading and Chunks: uses audio helpers
inference/detector.py - Threaded Orchestrator → Temporal Aggregation: stabilizes scores
pipeline/orchestrator.py - Dashboard and Alerts → Threaded Orchestrator: reads state
ui/dashboard.py - Dashboard and Alerts → WAV Replay: offers file replay
ui/dashboard.py - Dashboard and Alerts → Microphone Capture: offers live capture
ui/dashboard.py - Dashboard and Alerts → System Loopback: offers loopback
ui/dashboard.py - Threaded Orchestrator → Dashboard and Alerts: provides scores
README.md - Standalone Live Pipeline → Microphone Capture: starts capture
live_pipeline.py - Standalone Live Pipeline → System Loopback: supports capture
live_pipeline.py - Standalone Live Pipeline → Speech Gates: gates speech
live_pipeline.py - Standalone Live Pipeline → AI Detector: requests scores
live_pipeline.py - Standalone Live Pipeline → Temporal Aggregation: aggregates scores
live_pipeline.py - Streaming Evaluation → Speech Gates: gates speech
tools/evaluate_streaming.py - Streaming Evaluation → AI Detector: requests scores
tools/evaluate_streaming.py - Streaming Evaluation → Temporal Aggregation: evaluates aggregation
tools/evaluate_streaming.py - Streaming Evaluation → Audio Loading and Chunks: loads audio
tools/evaluate_streaming.py