go-whisper-api/whisper/vad.go

package whisper

import (
    "fmt"
    "math"

    "go-whisper-api/config"

    pkg "github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
)

func ApplyVAD(ctx pkg.Context, vad config.VAD) {
    if !vad.Enabled {
        return
    }
    vad = vad.WithDefaults()
    ctx.SetVAD(true)
    ctx.SetVADModelPath(vad.Model)
    ctx.SetVADThreshold(float32(vad.Threshold))
    ctx.SetVADMinSpeechMs(vad.MinSpeechMs)
    ctx.SetVADMinSilenceMs(vad.MinSilenceMs)
    if vad.MaxSpeechSec > 0 {
        ctx.SetVADMaxSpeechSec(float32(vad.MaxSpeechSec))
    } else {
        ctx.SetVADMaxSpeechSec(float32(math.MaxFloat32))
    }
    ctx.SetVADSpeechPadMs(vad.SpeechPadMs)
    ctx.SetVADSamplesOverlap(float32(vad.SamplesOverlap))
}

func prepareVAD(vad *config.VAD, modelsDir string) error {
    if vad == nil || !vad.Enabled {
        return nil
    }
    if modelsDir != "" {
        vad.Model = vad.ResolveModelPath(modelsDir)
    }
    *vad = vad.WithDefaults()
    if err := vad.Validate(); err != nil {
        return fmt.Errorf("vad: %w", err)
    }
    return nil
}