vad

type Analyzer interface {
	// SetSampleRate configures the audio sample rate. Must be called before
	// Analyze; implementations may clamp/validate as needed.
	SetSampleRate(sampleRate int)
	// SetParams updates the VAD parameters. Zero-values pick sensible defaults.
	SetParams(params Params)
	// Params returns the current VAD parameters.
	Params() Params
	// Analyze consumes audio for this stream, updates internal state, and
	// returns the current state, last confidence, and last smoothed volume.
	//
	// Audio is expected to be 16-bit PCM mono, matching audio.Frame/Data.
	Analyze(buf []byte) (State, float64, float64, error)
}

type AnalyzerDetector struct {
	Analyzer Analyzer
}

type Detector interface {
	IsSpeech(f audio.Frame) (bool, error)
	SetSampleRate(sampleRate int)
}

type EnergyAnalyzerBackend struct {
	Threshold float64
}

type EnergyDetector struct {
	Threshold float64
	// contains filtered or unexported fields
}

type Params struct {
	// Confidence is the minimum voice confidence (0..1) required to treat audio
	// as speech.
	Confidence float64
	// StartSecs is how long speech must be continuously detected before we move
	// from Quiet->Starting->Speaking.
	StartSecs float64
	// StopSecs is how long silence must be observed before we move from
	// Speaking->Stopping->Quiet.
	StopSecs float64
	// MinVolume is the minimum smoothed volume (0..1) required to treat audio
	// as speech. This is a second gate in addition to Confidence.
	MinVolume float64
	// Threshold is the RMS energy threshold for energy-based VAD (e.g. 0.02).
	// Used by EnergyAnalyzerBackend; zero means use default.
	Threshold float64
}

type State int