Merge pull request #13 from schollz/aubio

Aubio

src/aubio/aubio.go

@@ -0,0 +1,68 @@
+package aubio
+
+import (
+	"fmt"
+	"os/exec"
+	"strconv"
+	"strings"
+
+	"github.com/schollz/logger"
+	"github.com/schollz/teoperator/src/ffmpeg"
+	"github.com/schollz/teoperator/src/models"
+)
+
+// SplitOnSilence splits any audio file based on its silence
+func SplitOnSilence(fname string, silenceDB int, silenceMinimumSeconds float64, correction float64) (segments []models.AudioSegment, err error) {
+	cmd := fmt.Sprintf("-s -30 %s", fname)
+	logger.Debug(cmd)
+	out, err := exec.Command("aubioonset", strings.Fields(cmd)...).CombinedOutput()
+	if err != nil {
+		return
+	}
+	logger.Debugf("aubio output: %s", out)
+
+	numSamples, sampleRate, err := ffmpeg.NumSamples(fname)
+	if err != nil {
+		return
+	}
+	duration := float64(numSamples) / float64(sampleRate)
+
+	var segment models.AudioSegment
+	segment.Start = 0
+	segment.Filename = fname
+	for _, line := range strings.Split(string(out), "\n") {
+		seconds, err := strconv.ParseFloat(strings.TrimSpace(line), 64)
+		if err != nil {
+			logger.Error(line, err)
+			continue
+		}
+		if seconds == 0 {
+			continue
+		}
+		segment.End = seconds
+		segment.Duration = segment.End - segment.Start
+		segments = append(segments, segment)
+		segment.Start = segment.End
+	}
+	if segment.Start < duration {
+		segment.End = duration
+		segment.Duration = segment.End - segment.Start
+		segments = append(segments, segment)
+	}
+	logger.Debugf("segments: %+v", segments)
+
+	newSegments := make([]models.AudioSegment, len(segments))
+	i := 0
+	for _, segment := range segments {
+		if segment.Duration > 0.1 {
+			newSegments[i] = segment
+			i++
+		}
+	}
+	if i == 0 {
+		err = fmt.Errorf("could not find any segments")
+		return
+	}
+	newSegments = newSegments[:i]
+	return newSegments, nil
+}

src/audiosegment/audiosegment.go

@@ -14,6 +14,7 @@ import (
 	"strings"
 
 	"github.com/schollz/logger"
+	"github.com/schollz/teoperator/src/aubio"
 	"github.com/schollz/teoperator/src/ffmpeg"
 	"github.com/schollz/teoperator/src/models"
 	"github.com/schollz/teoperator/src/op1"
@@ -89,11 +90,16 @@ func SplitEqual(fname string, secondsMax float64, secondsOverlap float64, splice
 				}
 
 				if splices == 0 {
-					r.segments, r.err = ffmpeg.SplitOnSilence(fnameTrunc, -22, 0.2, -0.2)
-					if r.err != nil {
-						logger.Error(r.err)
-						results <- r
-						continue
+					logger.Debug("-- splitting on silence --")
+					r.segments, r.err = aubio.SplitOnSilence(fnameTrunc, -22, 0.2, -0.2)
+					if r.err != nil || len(r.segments) > 20 {
+						logger.Debug("-- splitting on silence w/ ffmpeg --")
+						r.segments, r.err = ffmpeg.SplitOnSilence(fnameTrunc, -22, 0.2, -0.2)
+						if r.err != nil {
+							logger.Error(r.err)
+							results <- r
+							continue
+						}
 					}
 				} else {
 					r.segments = make([]models.AudioSegment, splices)

src/convert/convert.go

@@ -71,7 +71,7 @@ func ToDrumSplice(fname string, slices int) (err error) {
 		}
 	} else {
 		var totalSamples int64
-		totalSamples, err = ffmpeg.NumSamples(fname2)
+		totalSamples, _, err = ffmpeg.NumSamples(fname2)
 		if err != nil {
 			return
 		}
@@ -112,7 +112,7 @@ func ToDrum(fnames []string, slices int) (err error) {
 			return
 		}
 		_, fnames2[i] = filepath.Split(fname2)
-		sampleEnd[i], err = ffmpeg.NumSamples(fname2)
+		sampleEnd[i], _, err = ffmpeg.NumSamples(fname2)
 		if err != nil {
 			return
 		}

src/ffmpeg/ffmpeg.go

@@ -28,12 +28,17 @@ func IsInstalled() bool {
 	return true
 }
 
-func NumSamples(fname string) (numSamples int64, err error) {
+func NumSamples(fname string) (numSamples int64, sampleRate int64, err error) {
 	file, err := os.Open(fname)
 	if err != nil {
 		return
 	}
 	reader := wav.NewReader(file)
+	format, err := reader.Format()
+	if err != nil {
+		return
+	}
+	sampleRate = int64(format.SampleRate)
 
 	defer file.Close()
 
@@ -109,7 +114,7 @@ func Normalize(fname string, fnameout string) (err error) {
 	if err != nil {
 		return
 	}
-	logger.Debugf("ffmpeg output: %s", out)
+	logger.Tracef("ffmpeg output: %s", out)
 	index := bytes.LastIndex(out, []byte("{"))
 	var n Normalization
 	err = json.Unmarshal(out[index:], &n)
@@ -137,7 +142,7 @@ func Normalize(fname string, fnameout string) (err error) {
 	if err != nil {
 		return
 	}
-	logger.Debugf("ffmpeg output: %s", out)
+	logger.Tracef("ffmpeg output: %s", out)
 
 	return
 
@@ -151,7 +156,7 @@ func SplitOnSilence(fname string, silenceDB int, silenceMinimumSeconds float64,
 	if err != nil {
 		return
 	}
-	logger.Debugf("ffmpeg output: %s", out)
+	logger.Tracef("ffmpeg output: %s", out)
 	// if !strings.Contains(string(out), "silence_end") {
 	// 	err = fmt.Errorf("could not find silence")
 	// 	logger.Error(err)
@@ -232,7 +237,7 @@ func RemoveSilence(fnameIn, fnameOut string) (err error) {
 	if err != nil {
 		return
 	}
-	logger.Debugf("ffmpeg output: %s", out)
+	logger.Tracef("ffmpeg output: %s", out)
 
 	return
 }