Fix audio transcoding

Clement Champetier · Clement Champetier · commit 8a338d60ba2a · 2015-02-04T11:03:09.000+01:00
* AudioTransform: dstFrame.resize takes number of channels.
* AudioGenerator: need to specify nbSamples at each decodeNextFrame.
diff --git a/src/AvTranscoder/decoder/AudioGenerator.cpp b/src/AvTranscoder/decoder/AudioGenerator.cpp
@@ -27,6 +27,9 @@ bool AudioGenerator::decodeNextFrame( Frame& frameBuffer )
 	// Generate silent
 	if( ! _inputFrame )
 	{
+		AudioFrame& audioBuffer = static_cast<AudioFrame&>( frameBuffer );
+		audioBuffer.setNbSamples( _frameDesc.getSampleRate() / _frameDesc.getFps() );
+
 		// Generate the silent only once
 		if( ! _silent )
 		{
@@ -35,10 +38,9 @@ bool AudioGenerator::decodeNextFrame( Frame& frameBuffer )
 			// input of convert
 			AudioFrame intermediateBuffer( _frameDesc );
 			intermediateBuffer.assign( _frameDesc.getDataSize(), fillChar );
+			intermediateBuffer.setNbSamples( audioBuffer.getNbSamples() );
 
 			// output of convert
-			AudioFrame& audioBuffer = static_cast<AudioFrame&>( frameBuffer );
-			audioBuffer.setNbSamples( 1.0 * _frameDesc.getSampleRate() / _frameDesc.getFps() );
 			_silent = new AudioFrame( audioBuffer.desc() );
 
 			// convert and store the silence
diff --git a/src/AvTranscoder/transform/AudioTransform.cpp b/src/AvTranscoder/transform/AudioTransform.cpp
@@ -39,16 +39,15 @@ AudioTransform::AudioTransform()
 
 bool AudioTransform::init( const Frame& srcFrame, const Frame& dstFrame )
 {
-	const AudioFrame& src = static_cast<const AudioFrame&>( srcFrame );
-	const AudioFrame& dst = static_cast<const AudioFrame&>( dstFrame );
-
 	_audioConvertContext = AllocResampleContext();
-
 	if( !_audioConvertContext )
 	{
 		throw std::runtime_error( "unable to create audio convert context" );
 	}
-	
+
+	const AudioFrame& src = static_cast<const AudioFrame&>( srcFrame );
+	const AudioFrame& dst = static_cast<const AudioFrame&>( dstFrame );
+
 	av_opt_set_int(  _audioConvertContext, "in_channel_layout",  av_get_default_channel_layout( src.desc().getChannels() ), 0 );
 	av_opt_set_int(  _audioConvertContext, "out_channel_layout", av_get_default_channel_layout( dst.desc().getChannels() ), 0 );
 	av_opt_set_int(  _audioConvertContext, "in_sample_rate",     src.desc().getSampleRate(), 0 );
@@ -65,15 +64,19 @@ bool AudioTransform::init( const Frame& srcFrame, const Frame& dstFrame )
 	return true;
 }
 
-bool AudioTransform::initFrames( const Frame& srcFrame, Frame& dstFrame )
+void AudioTransform::initFrames( const Frame& srcFrame, Frame& dstFrame )
 {
 	const AudioFrame& src = static_cast<const AudioFrame&>( srcFrame );
-	const AudioFrame& dst = static_cast<const AudioFrame&>( dstFrame );
+	AudioFrame& dst = static_cast<AudioFrame&>( dstFrame );
 
+	// resize buffer of output frame
 	int dstSampleSize = av_get_bytes_per_sample( dst.desc().getSampleFormat() );
-	dstFrame.resize( src.getNbSamples() * dstSampleSize );
+	dstFrame.resize( src.getNbSamples() * src.desc().getChannels() * dstSampleSize );
+
+	// set nbSamples of output frame
+	dst.setNbSamples( src.getNbSamples() );
+
 	_previousProcessedAudioFrameSize = srcFrame.getSize();
-	return true;
 }
 
 void AudioTransform::convert( const Frame& srcFrame, Frame& dstFrame )
@@ -98,12 +101,6 @@ void AudioTransform::convert( const Frame& srcFrame, Frame& dstFrame )
 	{
 		throw std::runtime_error( "unable to convert audio samples" );
 	}
-
-	size_t nbOutputSamples = nbOutputSamplesPerChannel * static_cast<const AudioFrame&>( dstFrame ).desc().getChannels();
-	
-	if( dstFrame.getSize() != nbOutputSamples )
-		dstFrame.resize( nbOutputSamples );
-	static_cast<AudioFrame&>( dstFrame ).setNbSamples( static_cast<const AudioFrame&>( srcFrame ).getNbSamples() );
 }
 
 }
diff --git a/src/AvTranscoder/transform/AudioTransform.hpp b/src/AvTranscoder/transform/AudioTransform.hpp
@@ -26,7 +26,9 @@ class AvExport AudioTransform : public ITransform
 
 private:
 	bool init( const Frame& srcFrame, const Frame& dstFrame );
-	bool initFrames( const Frame& srcFrame, Frame& dstFrame );
+
+	/// Resize output buffer if source has a different size from the last process
+	void initFrames( const Frame& srcFrame, Frame& dstFrame );
 	
 	ResampleContext* _audioConvertContext;
 	

Original file line number	Diff line number	Diff line change
`@@ -39,16 +39,15 @@ AudioTransform::AudioTransform()`
`39`	`39`
`40`	`40`	`bool AudioTransform::init( const Frame& srcFrame, const Frame& dstFrame )`
`41`	`41`	`{`
`42`		`- const AudioFrame& src = static_cast<const AudioFrame&>( srcFrame );`
`43`		`- const AudioFrame& dst = static_cast<const AudioFrame&>( dstFrame );`
`44`		`-`
`45`	`42`	`_audioConvertContext = AllocResampleContext();`
`46`		`-`
`47`	`43`	`if( !_audioConvertContext )`
`48`	`44`	`{`
`49`	`45`	`throw std::runtime_error( "unable to create audio convert context" );`
`50`	`46`	`}`
`51`		`-`
	`47`	`+`
	`48`	`+ const AudioFrame& src = static_cast<const AudioFrame&>( srcFrame );`
	`49`	`+ const AudioFrame& dst = static_cast<const AudioFrame&>( dstFrame );`
	`50`	`+`
`52`	`51`	`av_opt_set_int( _audioConvertContext, "in_channel_layout", av_get_default_channel_layout( src.desc().getChannels() ), 0 );`
`53`	`52`	`av_opt_set_int( _audioConvertContext, "out_channel_layout", av_get_default_channel_layout( dst.desc().getChannels() ), 0 );`
`54`	`53`	`av_opt_set_int( _audioConvertContext, "in_sample_rate", src.desc().getSampleRate(), 0 );`
`@@ -65,15 +64,19 @@ bool AudioTransform::init( const Frame& srcFrame, const Frame& dstFrame )`
`65`	`64`	`return true;`
`66`	`65`	`}`
`67`	`66`
`68`		`-bool AudioTransform::initFrames( const Frame& srcFrame, Frame& dstFrame )`
	`67`	`+void AudioTransform::initFrames( const Frame& srcFrame, Frame& dstFrame )`
`69`	`68`	`{`
`70`	`69`	`const AudioFrame& src = static_cast<const AudioFrame&>( srcFrame );`
`71`		`- const AudioFrame& dst = static_cast<const AudioFrame&>( dstFrame );`
	`70`	`+ AudioFrame& dst = static_cast<AudioFrame&>( dstFrame );`
`72`	`71`
	`72`	`+ // resize buffer of output frame`
`73`	`73`	`int dstSampleSize = av_get_bytes_per_sample( dst.desc().getSampleFormat() );`
`74`		`- dstFrame.resize( src.getNbSamples() * dstSampleSize );`
	`74`	`+ dstFrame.resize( src.getNbSamples() * src.desc().getChannels() * dstSampleSize );`
	`75`	`+`
	`76`	`+ // set nbSamples of output frame`
	`77`	`+ dst.setNbSamples( src.getNbSamples() );`
	`78`	`+`
`75`	`79`	`_previousProcessedAudioFrameSize = srcFrame.getSize();`
`76`		`- return true;`
`77`	`80`	`}`
`78`	`81`
`79`	`82`	`void AudioTransform::convert( const Frame& srcFrame, Frame& dstFrame )`
`@@ -98,12 +101,6 @@ void AudioTransform::convert( const Frame& srcFrame, Frame& dstFrame )`
`98`	`101`	`{`
`99`	`102`	`throw std::runtime_error( "unable to convert audio samples" );`
`100`	`103`	`}`
`101`		`-`
`102`		`- size_t nbOutputSamples = nbOutputSamplesPerChannel * static_cast<const AudioFrame&>( dstFrame ).desc().getChannels();`
`103`		`-`
`104`		`- if( dstFrame.getSize() != nbOutputSamples )`
`105`		`- dstFrame.resize( nbOutputSamples );`
`106`		`- static_cast<AudioFrame&>( dstFrame ).setNbSamples( static_cast<const AudioFrame&>( srcFrame ).getNbSamples() );`
`107`	`104`	`}`
`108`	`105`
`109`	`106`	`}`