Debugging

Diagnostic tools and benchmarks: they answer "what does this stage actually output right now" and do not describe system design.

• ASR Speech Recognition — debugging entry point and benchmarks for the sherpa-onnx recognizer
• TTS Audio Quality Evaluation — synthesize → decode → three-axis scoring, for comparing models and tuning speaking rate

The semantics of VAD and turn detection live in Server / VAD and Turn Detection.

• TTS Audio Quality Evaluation
• ASR Speech Recognition