Merge pull request #142 from bcc-code/feat/silence-detection

Feat/silence detection

activities/normalize.go

@@ -83,13 +83,25 @@ type NormalizeAudioParams struct {
 
 type NormalizeAudioResult struct {
 	FilePath       paths.Path
+	IsSilent       bool
 	InputAnalysis  *common.AnalyzeEBUR128Result
 	OutputAnalysis *common.AnalyzeEBUR128Result
 }
 
 func NormalizeAudioActivity(ctx context.Context, params NormalizeAudioParams) (*NormalizeAudioResult, error) {
 	out := &NormalizeAudioResult{}
 
+	silent, err := transcode.AudioIsSilent(params.FilePath)
+	if err != nil {
+		return nil, err
+	}
+	if silent {
+		return &NormalizeAudioResult{
+			FilePath: params.FilePath,
+			IsSilent: true,
+		}, nil
+	}
+
 	r128Result, err := AnalyzeEBUR128Activity(ctx, AnalyzeEBUR128Params{
 		FilePath:       params.FilePath,
 		TargetLoudness: params.TargetLUFS,

cmd/trigger_ui/main.go

@@ -190,6 +190,7 @@ func (s *TriggerServer) triggerHandlerPOST(c *gin.Context) {
 		VXID:          vxID,
 		WithFiles:     c.PostForm("withFiles") == "on",
 		WithChapters:  c.PostForm("withChapters") == "on",
+		IgnoreSilence: c.PostForm("ignoreSilence") == "on",
 		WatermarkPath: watermarkPath,
 		AudioSource:   audioSource,
 		Destinations:  c.PostFormArray("destinations[]"),

cmd/trigger_ui/templates/vx-export.gohtml

@@ -115,6 +115,11 @@
             <input class="ml-2 h-4 w-4 my-auto" type="checkbox" name="withChapters"
                    id="withChapters">
         </div>
+        <div class="flex">
+            <label for="ignoreSilence" class="my-auto">Ignore silence</label>
+            <input class="ml-2 h-4 w-4 my-auto" type="checkbox" name="ignoreSilence"
+                   id="ignoreSilence">
+        </div>
 
         <input id="submit"
                class="cursor-pointer rounded-md bg-[#6A64F1] py-3 px-8 text-center text-base font-semibold text-white outline-none"

services/transcode/audio.go

@@ -1,9 +1,12 @@
 package transcode
 
 import (
+	"bytes"
 	"fmt"
 	"os"
+	"os/exec"
 	"path/filepath"
+	"regexp"
 	"strconv"
 	"strings"
 
@@ -13,6 +16,77 @@ import (
 	"github.com/bcc-code/bcc-media-flows/services/ffmpeg"
 )
 
+type SilencePeriod struct {
+	Start float64 `json:"start"`
+	End   float64 `json:"end"`
+}
+
+func AudioGetSilencePeriods(path paths.Path, threshold float64) ([]SilencePeriod, error) {
+	params := []string{
+		"-loglevel", "info",
+		"-hide_banner",
+		"-i", path.Local(),
+		"-af", fmt.Sprintf("silencedetect=noise=-90dB:d=%f", threshold),
+		"-f", "null",
+		"-",
+	}
+
+	cmd := exec.Command("ffmpeg", params...)
+
+	var stderr bytes.Buffer
+	cmd.Stderr = &stderr
+
+	// Run the command
+	err := cmd.Run()
+	if err != nil {
+		fmt.Println(fmt.Sprint(err) + ": " + stderr.String())
+		return nil, err
+	}
+
+	result := stderr.String()
+
+	var silencePeriods []SilencePeriod
+	r := regexp.MustCompile(`silence_(start|end): ([0-9.]+)`)
+
+	var start float64
+	for _, line := range strings.Split(result, "\n") {
+		matches := r.FindStringSubmatch(line)
+		if len(matches) == 3 {
+			if matches[1] == "start" {
+				start, _ = strconv.ParseFloat(matches[2], 64)
+			} else if matches[1] == "end" {
+				end, _ := strconv.ParseFloat(matches[2], 64)
+				silencePeriods = append(silencePeriods, SilencePeriod{Start: start, End: end})
+			}
+		}
+	}
+
+	return silencePeriods, nil
+}
+
+func AudioIsSilent(path paths.Path) (bool, error) {
+	info, err := ffmpeg.GetStreamInfo(path.Local())
+	if err != nil {
+		return false, err
+	}
+
+	silencePeriods, err := AudioGetSilencePeriods(path, 10)
+	if err != nil {
+		return false, err
+	}
+
+	var dur float64
+	for _, p := range silencePeriods {
+		dur += p.End - p.Start
+	}
+
+	if int(info.TotalSeconds) == int(dur) {
+		return true, nil
+	}
+
+	return false, nil
+}
+
 func AudioAac(input common.AudioInput, cb ffmpeg.ProgressCallback) (*common.AudioResult, error) {
 	params := []string{
 		"-progress", "pipe:1",

services/transcode/audio_test.go

@@ -27,3 +27,8 @@ func Test_AudioSplit(t *testing.T) {
 
 	spew.Dump(files)
 }
+
+func Test_AudioSilence(t *testing.T) {
+	_, err := AudioIsSilent(paths.MustParse("/private/temp/workflows/5d2ea767-6b71-44c6-a207-005d7522326c/FKTB_20210415_2000_SEQ-slv.wav"))
+	assert.Nil(t, err)
+}

workflows/export/vx_export.go

@@ -40,6 +40,7 @@ type VXExportParams struct {
 	AudioSource   string
 	Languages     []string
 	Subclip       string
+	IgnoreSilence bool
 }
 
 type VXExportResult struct {

workflows/export/vx_export_vod.go

@@ -41,7 +41,7 @@ func VXExportToVOD(ctx workflow.Context, params VXExportChildWorkflowParams) (*V
 		}
 	}
 
-	audioFiles, err := prepareAudioFiles(ctx, params.MergeResult, params.TempDir, true)
+	audioFiles, err := prepareAudioFiles(ctx, params.MergeResult, params.TempDir, true, params.ParentParams.IgnoreSilence)
 	if err != nil {
 		return nil, err
 	}
@@ -142,7 +142,7 @@ type vxExportVodService struct {
 	errs                   []error
 }
 
-func prepareAudioFiles(ctx workflow.Context, mergeResult MergeExportDataResult, tempDir paths.Path, normalizeAudio bool) (map[string]paths.Path, error) {
+func prepareAudioFiles(ctx workflow.Context, mergeResult MergeExportDataResult, tempDir paths.Path, normalizeAudio, ignoreSilence bool) (map[string]paths.Path, error) {
 	prepareFilesSelector := workflow.NewSelector(ctx)
 
 	if normalizeAudio {
@@ -172,7 +172,14 @@ func prepareAudioFiles(ctx workflow.Context, mergeResult MergeExportDataResult,
 				return nil, fmt.Errorf("failed to normalize audio for language %s: %w", lang, err)
 			}
 
-			mergeResult.AudioFiles[lang] = normalizedRes.FilePath
+			if normalizedRes.IsSilent {
+				if ignoreSilence {
+					continue
+				}
+				return nil, fmt.Errorf("audio for language %s is silent", lang)
+			} else {
+				mergeResult.AudioFiles[lang] = normalizedRes.FilePath
+			}
 		}
 	}