Merge pull request #352 from bcc-code/feat/mp3-cbr-support

Implement CBR for MP3

common/transcode.go

@@ -25,6 +25,9 @@ type AudioInput struct {
 	Path            paths.Path
 	Bitrate         string
 	DestinationPath paths.Path
+
+	// Not all codecs may support constant bitrate
+	ForceCBR bool
 }
 
 type DetectSilenceInput struct {

services/ffmpeg/probe.go

@@ -136,6 +136,7 @@ func GetStreamInfo(path string) (StreamInfo, error) {
 	if err != nil {
 		return StreamInfo{}, err
 	}
+
 	return ProbeResultToInfo(info), nil
 }
 

services/transcode/audio.go

@@ -3,14 +3,15 @@ package transcode
 import (
 	"bytes"
 	"fmt"
-	"github.com/bcc-code/bcc-media-flows/utils"
 	"os"
 	"os/exec"
 	"path/filepath"
 	"regexp"
 	"strconv"
 	"strings"
 
+	"github.com/bcc-code/bcc-media-flows/utils"
+
 	"github.com/bcc-code/bcc-media-flows/paths"
 
 	"github.com/bcc-code/bcc-media-flows/common"
@@ -284,9 +285,15 @@ func AudioMP3(input common.AudioInput, cb ffmpeg.ProgressCallback) (*common.Audi
 		"-hide_banner",
 		"-i", input.Path.Local(),
 		"-c:a", "libmp3lame",
-		"-q:a", fmt.Sprint(getQfactorFromBitrate(input.Bitrate)),
 	}
 
+	if input.ForceCBR {
+		params = append(params, "-b:a", input.Bitrate)
+	} else {
+		params = append(params, "-q:a", fmt.Sprint(getQfactorFromBitrate(input.Bitrate)))
+	}
+
+
 	outputFilePath := filepath.Join(input.DestinationPath.Local(), input.Path.Base())
 	outputFilePath = fmt.Sprintf("%s-%s.mp3", outputFilePath[:len(outputFilePath)-len(filepath.Ext(outputFilePath))], input.Bitrate)
 

services/transcode/audio_test.go

@@ -1,17 +1,18 @@
 package transcode
 
 import (
-	"github.com/bcc-code/bcc-media-flows/services/ffmpeg"
-	"github.com/bcc-code/bcc-media-flows/utils/testutils"
 	"os"
 	"testing"
 
+	"github.com/bcc-code/bcc-media-flows/services/ffmpeg"
+	"github.com/bcc-code/bcc-media-flows/utils/testutils"
+
 	"github.com/bcc-code/bcc-media-flows/common"
 	"github.com/bcc-code/bcc-media-flows/paths"
 	"github.com/stretchr/testify/assert"
 )
 
-func Test_Audio(t *testing.T) {
+func Test_AACEncode(t *testing.T) {
 	tempDstPath := paths.MustParse("./testdata/test" + t.Name() + ".wav")
 	err := GenerateToneFile(1000, 5, 48000, "01:00:00:00", tempDstPath)
 	assert.NoError(t, err)
@@ -35,6 +36,58 @@ func Test_Audio(t *testing.T) {
 	assert.Equal(t, 1, len(info.AudioStreams))
 }
 
+func Test_MP3Encode_VBR(t *testing.T) {
+	tempDstPath := paths.MustParse("./testdata/test" + t.Name() + ".wav")
+	err := GenerateToneFile(1000, 5, 48000, "01:00:00:00", tempDstPath)
+	assert.NoError(t, err)
+
+	p, stop := printProgress()
+	defer close(stop)
+	res, err := AudioMP3(common.AudioInput{
+		Path:            tempDstPath,
+		DestinationPath: tempDstPath.Dir(),
+		Bitrate:         "128k",
+		ForceCBR:        false,
+	}, p)
+	assert.Nil(t, err)
+	assert.NotNil(t, res)
+
+	info, err := ffmpeg.GetStreamInfo(res.OutputPath.Local())
+	assert.NoError(t, err)
+
+	assert.True(t, info.HasAudio)
+	assert.False(t, info.HasVideo)
+	assert.InDelta(t, 5, 0.2, info.TotalSeconds)
+	assert.Equal(t, 1, len(info.AudioStreams))
+
+}
+
+func Test_MP3Encode_CBR(t *testing.T) {
+	tempDstPath := paths.MustParse("./testdata/test" + t.Name() + ".wav")
+	err := GenerateToneFile(1000, 5, 48000, "01:00:00:00", tempDstPath)
+	assert.NoError(t, err)
+
+	p, stop := printProgress()
+	defer close(stop)
+	res, err := AudioMP3(common.AudioInput{
+		Path:            tempDstPath,
+		DestinationPath: tempDstPath.Dir(),
+		Bitrate:         "128k",
+		ForceCBR:        true,
+	}, p)
+	assert.Nil(t, err)
+	assert.NotNil(t, res)
+
+	info, err := ffmpeg.GetStreamInfo(res.OutputPath.Local())
+	assert.NoError(t, err)
+
+	assert.True(t, info.HasAudio)
+	assert.False(t, info.HasVideo)
+	assert.InDelta(t, 5, 0.2, info.TotalSeconds)
+	assert.Equal(t, 1, len(info.AudioStreams))
+	assert.Equal(t, "128000", info.AudioStreams[0].BitRate)
+}
+
 func Test_AudioSplit_Stereo(t *testing.T) {
 	tempDstPath := paths.MustParse("./testdata/test" + t.Name() + ".wav")
 	testutils.GenerateMultichannelAudioFile(tempDstPath, 2, 10)

workflows/export/vx_export_bmm.go

@@ -3,13 +3,14 @@ package export
 import (
 	"crypto/sha1"
 	"fmt"
-	bccmflows "github.com/bcc-code/bcc-media-flows"
-	"github.com/bcc-code/bcc-media-flows/services/telegram"
 	"path"
 	"strconv"
 	"strings"
 	"time"
 
+	bccmflows "github.com/bcc-code/bcc-media-flows"
+	"github.com/bcc-code/bcc-media-flows/services/telegram"
+
 	pcommon "github.com/bcc-code/bcc-media-platform/backend/common"
 
 	platform_activities "github.com/bcc-code/bcc-media-flows/activities/platform"
@@ -122,6 +123,7 @@ func VXExportToBMM(ctx workflow.Context, params VXExportChildWorkflowParams) (*V
 				Path:            audio.FilePath,
 				DestinationPath: params.OutputDir,
 				Bitrate:         bitrate,
+				ForceCBR:        true,
 			})
 			encodings = append(encodings, f.Future)
 		}