feat: loudnorm

examples/ffmpeg/loudnorm/main.go

@@ -0,0 +1,75 @@
+package main
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/fxkt-tech/liv/ffmpeg"
+	"github.com/fxkt-tech/liv/ffmpeg/codec"
+	"github.com/fxkt-tech/liv/ffmpeg/filter"
+	"github.com/fxkt-tech/liv/ffmpeg/input"
+	"github.com/fxkt-tech/liv/ffmpeg/output"
+	"github.com/fxkt-tech/liv/internal/encoding/json"
+)
+
+// ffmpeg -i voice.wav -filter_complex "loudnorm=I=-16:TP=-1:LRA=11:print_format=json" -f null -
+func main() {
+	var (
+		ctx = context.Background()
+
+		// inputs
+		iMain = input.WithSimple("voice.wav")
+
+		// filters
+		fLoudnorm = filter.Loudnorm(-12, 7, -1.5).Use(iMain.A())
+	)
+
+	ln, err := ffmpeg.New(
+		ffmpeg.WithDebug(true),
+		ffmpeg.WithDry(true),
+		ffmpeg.WithLogLevel(""),
+	).AddInput(
+		iMain,
+	).AddFilter(
+		fLoudnorm,
+	).AddOutput(
+		output.New(
+			output.Map(fLoudnorm),
+			output.Format("null"),
+			output.File("-"),
+		),
+	).ExtractLoudnorm(ctx)
+	if err != nil {
+		fmt.Println(err)
+	}
+	fmt.Println(json.ToString(ln))
+
+	var (
+		// filters
+		fLoudnorm2 = filter.LoudnormDoublePass(
+			-12, 7, -1.5,
+			ln.InputI, ln.InputLRA, ln.InputTP, ln.InputThresh,
+		).Use(iMain.A())
+
+		// output
+		oOnly = output.New(
+			output.Map(fLoudnorm2),
+			output.AudioCodec(codec.PCMS16LE),
+			output.File("voice_ln.wav"),
+		)
+	)
+
+	err = ffmpeg.New(
+		ffmpeg.WithDebug(true),
+		ffmpeg.WithDry(true),
+	).AddInput(
+		iMain,
+	).AddFilter(
+		fLoudnorm2,
+	).AddOutput(
+		oOnly,
+	).Run(ctx)
+	if err != nil {
+		fmt.Println(err)
+	}
+}

ffmpeg/ffmpeg.go

@@ -5,11 +5,13 @@ import (
 	"errors"
 	"fmt"
 	"os/exec"
+	"regexp"
 	"strings"
 
 	"github.com/fxkt-tech/liv/ffmpeg/filter"
 	"github.com/fxkt-tech/liv/ffmpeg/input"
 	"github.com/fxkt-tech/liv/ffmpeg/output"
+	"github.com/fxkt-tech/liv/internal/encoding/json"
 	"github.com/fxkt-tech/liv/internal/sugar"
 )
 
@@ -90,3 +92,35 @@ func (ff *FFmpeg) Run(ctx context.Context) (err error) {
 	}
 	return
 }
+
+type LoudnormParms struct {
+	InputI            float32 `json:"input_i,string"`
+	InputTP           float32 `json:"input_tp,string"`
+	InputLRA          float32 `json:"input_lra,string"`
+	InputThresh       float32 `json:"input_thresh,string"`
+	OutputI           float32 `json:"output_i,string"`
+	OutputTP          float32 `json:"output_tp,string"`
+	OutputLRA         float32 `json:"output_lra,string"`
+	OutputThresh      float32 `json:"output_thresh,string"`
+	NormalizationType string  `json:"normalization_type,string"`
+	TargetOffset      float32 `json:"target_offset,string"`
+}
+
+func (ff *FFmpeg) ExtractLoudnorm(ctx context.Context) (*LoudnormParms, error) {
+	if ff.debug {
+		ff.DryRun()
+	}
+	cc := exec.CommandContext(ctx, ff.bin, ff.Params()...)
+	retbytes, err := cc.CombinedOutput()
+	retstr := string(retbytes)
+	if err != nil {
+		return nil, errors.New(retstr)
+	}
+
+	reg := regexp.MustCompile("(?s)({.*})")
+	matches := reg.FindStringSubmatch(retstr)
+	if len(matches) < 2 {
+		return nil, errors.New(retstr)
+	}
+	return json.ToV[*LoudnormParms]([]byte(matches[1])), nil
+}

ffmpeg/filter/audio.go

@@ -50,10 +50,20 @@ func AFade(t string, st, d float32) *SingleFilter {
 	}
 }
 
-func Loudnorm(i, tp int32) *SingleFilter {
+func Loudnorm(i, lra, tp float32) *SingleFilter {
 	return &SingleFilter{
 		name:    naming.Default.Gen(),
-		content: fmt.Sprintf("loudnorm=I=%d:TP=%d", i, tp),
+		content: fmt.Sprintf("loudnorm=i=%f:lra=%f:tp=%f:print_format=json", i, lra, tp),
+	}
+}
+
+func LoudnormDoublePass(i, lra, tp float32, mi, mlra, mtp, mthres float32) *SingleFilter {
+	return &SingleFilter{
+		name: naming.Default.Gen(),
+		content: fmt.Sprintf(
+			"loudnorm=i=%f:lra=%f:tp=%f:measured_i=%f:measured_lra=%f:measured_tp=%f:measured_thresh=%f:print_format=json",
+			i, lra, tp, mi, mlra, mtp, mthres,
+		),
 	}
 }