#include <stdlib.h>
#include <string.h>
#include <pthread.h>
+#include <limits.h>
#if LIBAVUTIL_VERSION_INT < (50<<16)
#define PIX_FMT_RGB32 PIX_FMT_RGBA32
#define POSITION_INITIAL (-2)
#define POSITION_INVALID (-1)
+#define MAX_AUDIO_STREAMS (8)
+
void avformat_lock( );
void avformat_unlock( );
{
struct mlt_producer_s parent;
AVFormatContext *dummy_context;
- AVFormatContext *audio_context;
- AVFormatContext *video_context;
- AVCodec *audio_codec;
- AVCodec *video_codec;
+ AVFormatContext *audio_format;
+ AVFormatContext *video_format;
+ AVCodecContext *audio_codec[ MAX_AUDIO_STREAMS ];
+ AVCodecContext *video_codec;
AVFrame *av_frame;
- ReSampleContext *audio_resample;
+ ReSampleContext *audio_resample[ MAX_AUDIO_STREAMS ];
mlt_position audio_expected;
mlt_position video_expected;
int audio_index;
int current_position;
int got_picture;
int top_field_first;
- int16_t *audio_buffer;
- int16_t *decode_buffer;
- int audio_used;
+ int16_t *audio_buffer[ MAX_AUDIO_STREAMS ];
+ size_t audio_buffer_size[ MAX_AUDIO_STREAMS ];
+ int16_t *decode_buffer[ MAX_AUDIO_STREAMS ];
+ int audio_used[ MAX_AUDIO_STREAMS ];
+ int audio_streams;
+ int audio_max_stream;
+ int total_channels;
+ int max_channel;
+ int max_frequency;
+ unsigned int invalid_pts_counter;
+ double resample_factor;
};
typedef struct producer_avformat_s *producer_avformat;
// Forward references.
static int producer_open( producer_avformat this, mlt_profile profile, char *file );
static int producer_get_frame( mlt_producer this, mlt_frame_ptr frame, int index );
-static void producer_file_close( void *context );
+static void producer_format_close( void *context );
static void producer_close( mlt_producer parent );
/** Constructor for libavformat.
// Register our get_frame implementation
producer->get_frame = producer_get_frame;
+
+ this->resample_factor = 1.0;
// Open the file
if ( producer_open( this, profile, file ) != 0 )
{
// Clean up
mlt_producer_close( producer );
- this = NULL;
+ producer = NULL;
}
else
{
// Close the file to release resources for large playlists - reopen later as needed
- producer_file_close( this->dummy_context );
+ producer_format_close( this->dummy_context );
this->dummy_context = NULL;
- producer_file_close( this->audio_context );
- this->audio_context = NULL;
- producer_file_close( this->video_context );
- this->video_context = NULL;
+ producer_format_close( this->audio_format );
+ this->audio_format = NULL;
+ producer_format_close( this->video_format );
+ this->video_format = NULL;
// Default the user-selectable indices from the auto-detected indices
mlt_properties_set_int( properties, "audio_index", this->audio_index );
snprintf( key, sizeof(key), "meta.media.%d.stream.sample_aspect_ratio", i );
mlt_properties_set_double( meta_media, key, av_q2d( context->streams[ i ]->sample_aspect_ratio ) );
#endif
+ snprintf( key, sizeof(key), "meta.media.%d.codec.frame_rate", i );
+ mlt_properties_set_double( meta_media, key, (double) codec_context->time_base.den /
+ ( codec_context->time_base.num == 0 ? 1 : codec_context->time_base.num ) );
snprintf( key, sizeof(key), "meta.media.%d.codec.pix_fmt", i );
mlt_properties_set( meta_media, key, avcodec_get_pix_fmt_name( codec_context->pix_fmt ) );
snprintf( key, sizeof(key), "meta.media.%d.codec.sample_aspect_ratio", i );
mlt_properties_set_int( meta_media, key, codec_context->bit_rate );
// snprintf( key, sizeof(key), "meta.media.%d.codec.time_base", i );
// mlt_properties_set_double( meta_media, key, av_q2d( codec_context->time_base ) );
- snprintf( key, sizeof(key), "meta.media.%d.codec.profile", i );
- mlt_properties_set_int( meta_media, key, codec_context->profile );
- snprintf( key, sizeof(key), "meta.media.%d.codec.level", i );
- mlt_properties_set_int( meta_media, key, codec_context->level );
+// snprintf( key, sizeof(key), "meta.media.%d.codec.profile", i );
+// mlt_properties_set_int( meta_media, key, codec_context->profile );
+// snprintf( key, sizeof(key), "meta.media.%d.codec.level", i );
+// mlt_properties_set_int( meta_media, key, codec_context->level );
}
return meta_media;
/** Producer file destructor.
*/
-static void producer_file_close( void *context )
+static void producer_format_close( void *context )
{
if ( context )
{
{
this->seekable = av_seek_frame( context, -1, this->start_time, AVSEEK_FLAG_BACKWARD ) >= 0;
mlt_properties_set_int( properties, "seekable", this->seekable );
- producer_file_close( this->dummy_context );
+ producer_format_close( this->dummy_context );
this->dummy_context = context;
av_open_input_file( &context, file, NULL, 0, NULL );
av_find_stream_info( context );
// We're going to cheat here - for a/v files, we will have two contexts (reasoning will be clear later)
if ( av == 0 && audio_index != -1 && video_index != -1 )
{
- // We'll use the open one as our video_context
- producer_file_close( this->video_context );
- this->video_context = context;
+ // We'll use the open one as our video_format
+ avformat_unlock();
+ producer_format_close( this->video_format );
+ avformat_lock();
+ this->video_format = context;
// And open again for our audio context
av_open_input_file( &context, file, NULL, 0, NULL );
av_find_stream_info( context );
// Audio context
- producer_file_close( this->audio_context );
- this->audio_context = context;
+ avformat_unlock();
+ producer_format_close( this->audio_format );
+ avformat_lock();
+ this->audio_format = context;
}
else if ( av != 2 && video_index != -1 )
{
// We only have a video context
- producer_file_close( this->video_context );
- this->video_context = context;
+ avformat_unlock();
+ producer_format_close( this->video_format );
+ avformat_lock();
+ this->video_format = context;
}
else if ( audio_index != -1 )
{
// We only have an audio context
- producer_file_close( this->audio_context );
- this->audio_context = context;
+ avformat_unlock();
+ producer_format_close( this->audio_format );
+ avformat_lock();
+ this->audio_format = context;
}
else
{
return ( double )position / mlt_producer_get_fps( this );
}
+ // Collect information about all audio streams
+
+static void get_audio_streams_info( producer_avformat this )
+{
+ // Fetch the audio format context
+ AVFormatContext *context = this->audio_format;
+ int i;
+
+ for ( i = 0;
+ i < context->nb_streams;
+ i++ )
+ {
+ if ( context->streams[i]->codec->codec_type == CODEC_TYPE_AUDIO )
+ {
+ AVCodecContext *codec_context = context->streams[i]->codec;
+ AVCodec *codec = avcodec_find_decoder( codec_context->codec_id );
+
+ // If we don't have a codec and we can't initialise it, we can't do much more...
+ avformat_lock( );
+ if ( codec && avcodec_open( codec_context, codec ) >= 0 )
+ {
+ this->audio_streams++;
+ this->audio_max_stream = i;
+ this->total_channels += codec_context->channels;
+ if ( codec_context->channels > this->max_channel )
+ this->max_channel = codec_context->channels;
+ if ( codec_context->sample_rate > this->max_frequency )
+ this->max_frequency = codec_context->sample_rate;
+ avcodec_close( codec_context );
+ }
+ avformat_unlock( );
+ }
+ }
+ mlt_log_verbose( NULL, "[producer avformat] audio: total_streams %d max_stream %d total_channels %d max_channels %d\n",
+ this->audio_streams, this->audio_max_stream, this->total_channels, this->max_channel );
+}
+
static inline void convert_image( AVFrame *frame, uint8_t *buffer, int pix_fmt, mlt_image_format *format, int width, int height )
{
#ifdef SWSCALE
AVPicture output;
output.data[0] = buffer;
output.data[1] = buffer + width * height;
- output.data[2] = buffer + ( 3 * width * height ) / 2;
+ output.data[2] = buffer + ( 5 * width * height ) / 4;
output.linesize[0] = width;
output.linesize[1] = width >> 1;
output.linesize[2] = width >> 1;
AVPicture pict;
pict.data[0] = buffer;
pict.data[1] = buffer + width * height;
- pict.data[2] = buffer + ( 3 * width * height ) / 2;
+ pict.data[2] = buffer + ( 5 * width * height ) / 4;
pict.linesize[0] = width;
pict.linesize[1] = width >> 1;
pict.linesize[2] = width >> 1;
avformat_lock();
- // Fetch the video_context
- AVFormatContext *context = this->video_context;
+ // Fetch the video format context
+ AVFormatContext *context = this->video_format;
// Get the video stream
AVStream *stream = context->streams[ this->video_index ];
else if ( context->start_time != AV_NOPTS_VALUE )
pts -= context->start_time;
int_position = ( int )( av_q2d( stream->time_base ) * pts * source_fps + 0.1 );
+ if ( pkt.pts == AV_NOPTS_VALUE )
+ {
+ this->invalid_pts_counter++;
+ if ( this->invalid_pts_counter > 20 )
+ {
+ mlt_log_panic( MLT_PRODUCER_SERVICE(producer), "\ainvalid PTS; DISABLING NEW_SEEK!\n" );
+ mlt_properties_set_int( properties, "new_seek", 0 );
+ int_position = req_position;
+ use_new_seek = 0;
+ }
+ }
+ else
+ {
+ this->invalid_pts_counter = 0;
+ }
+ mlt_log_debug( MLT_PRODUCER_SERVICE(producer), "pkt.pts %llu req_pos %d cur_pos %d pkt_pos %d",
+ pkt.pts, req_position, this->current_position, int_position );
}
else
{
codec_context->reordered_opaque = pkt.pts;
if ( int_position >= req_position )
codec_context->skip_loop_filter = AVDISCARD_NONE;
+#if (LIBAVCODEC_VERSION_INT >= ((52<<16)+(26<<8)+0))
+ ret = avcodec_decode_video2( codec_context, this->av_frame, &got_picture, &pkt );
+#else
ret = avcodec_decode_video( codec_context, this->av_frame, &got_picture, pkt.data, pkt.size );
+#endif
// Note: decode may fail at the beginning of MPEGfile (B-frames referencing before first I-frame), so allow a few errors.
if ( ret < 0 )
{
convert_image( this->av_frame, *buffer, codec_context->pix_fmt, format, *width, *height );
if ( !mlt_properties_get( properties, "force_progressive" ) )
mlt_properties_set_int( frame_properties, "progressive", !this->av_frame->interlaced_frame );
- this->top_field_first = this->av_frame->top_field_first;
+ this->top_field_first |= this->av_frame->top_field_first;
this->current_position = int_position;
this->got_picture = 1;
}
}
}
+/** Initialize the video codec context.
+ */
+
+static int video_codec_init( producer_avformat this, int index, mlt_properties properties )
+{
+ // Initialise the codec if necessary
+ if ( !this->video_codec )
+ {
+ // Get the video stream
+ AVStream *stream = this->video_format->streams[ index ];
+
+ // Get codec context
+ AVCodecContext *codec_context = stream->codec;
+
+ // Find the codec
+ AVCodec *codec = avcodec_find_decoder( codec_context->codec_id );
+
+ // Initialise multi-threading
+ int thread_count = mlt_properties_get_int( properties, "threads" );
+ if ( thread_count == 0 && getenv( "MLT_AVFORMAT_THREADS" ) )
+ thread_count = atoi( getenv( "MLT_AVFORMAT_THREADS" ) );
+ if ( thread_count > 1 )
+ {
+ avcodec_thread_init( codec_context, thread_count );
+ codec_context->thread_count = thread_count;
+ }
+
+ // If we don't have a codec and we can't initialise it, we can't do much more...
+ avformat_lock( );
+ if ( codec && avcodec_open( codec_context, codec ) >= 0 )
+ {
+ // Now store the codec with its destructor
+ producer_codec_close( this->video_codec );
+ this->video_codec = codec_context;
+ }
+ else
+ {
+ // Remember that we can't use this later
+ this->video_index = -1;
+ }
+ avformat_unlock( );
+
+ // Process properties as AVOptions
+ apply_properties( codec_context, properties, AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM );
+
+ // Reset some image properties
+ mlt_properties_set_int( properties, "width", this->video_codec->width );
+ mlt_properties_set_int( properties, "height", this->video_codec->height );
+ mlt_properties_set_double( properties, "aspect_ratio", get_aspect_ratio( stream, this->video_codec, NULL ) );
+
+ // Determine the fps first from the codec
+ double source_fps = (double) this->video_codec->time_base.den /
+ ( this->video_codec->time_base.num == 0 ? 1 : this->video_codec->time_base.num );
+
+ if ( mlt_properties_get( properties, "force_fps" ) )
+ {
+ source_fps = mlt_properties_get_double( properties, "force_fps" );
+ stream->time_base = av_d2q( source_fps, 255 );
+ }
+ else
+ {
+ // If the muxer reports a frame rate different than the codec
+ double muxer_fps = av_q2d( stream->r_frame_rate );
+ // Choose the lesser - the wrong tends to be off by some multiple of 10
+ source_fps = FFMIN( source_fps, muxer_fps );
+ }
+
+ // We'll use fps if it's available
+ if ( source_fps > 0 )
+ mlt_properties_set_double( properties, "source_fps", source_fps );
+ else
+ mlt_properties_set_double( properties, "source_fps", mlt_producer_get_fps( &this->parent ) );
+ }
+ return this->video_codec && this->video_index > -1;
+}
+
/** Set up video handling.
*/
// Get the properties
mlt_properties properties = MLT_PRODUCER_PROPERTIES( producer );
- // Fetch the video_context
- AVFormatContext *context = this->video_context;
+ // Fetch the video format context
+ AVFormatContext *context = this->video_format;
// Get the video_index
int index = mlt_properties_get_int( properties, "video_index" );
mlt_events_block( properties, producer );
producer_open( this, mlt_service_profile( MLT_PRODUCER_SERVICE(producer) ),
mlt_properties_get( properties, "resource" ) );
- context = this->video_context;
- producer_file_close( this->dummy_context );
+ context = this->video_format;
+ producer_format_close( this->dummy_context );
this->dummy_context = NULL;
mlt_events_unblock( properties, producer );
+ if ( this->audio_format )
+ get_audio_streams_info( this );
// Process properties as AVOptions
apply_properties( context, properties, AV_OPT_FLAG_DECODING_PARAM );
mlt_properties_set_int( properties, "video_index", index );
}
+ // Update the video properties if the index changed
+ if ( index != this->video_index )
+ {
+ // Reset the video properties if the index changed
+ this->video_index = index;
+ producer_codec_close( this->video_codec );
+ this->video_codec = NULL;
+ }
+
// Get the frame properties
mlt_properties frame_properties = MLT_FRAME_PROPERTIES( frame );
- if ( context && index > -1 )
+ // Get the codec
+ if ( context && index > -1 && video_codec_init( this, index, properties ) )
{
- // Get the video stream
- AVStream *stream = context->streams[ index ];
+ // Set the frame properties
+ double force_aspect_ratio = mlt_properties_get_double( properties, "force_aspect_ratio" );
+ double aspect_ratio = ( force_aspect_ratio > 0.0 ) ?
+ force_aspect_ratio : mlt_properties_get_double( properties, "aspect_ratio" );
+
+ // Set the width and height
+ mlt_properties_set_int( frame_properties, "width", this->video_codec->width );
+ mlt_properties_set_int( frame_properties, "height", this->video_codec->height );
+ mlt_properties_set_int( frame_properties, "real_width", this->video_codec->width );
+ mlt_properties_set_int( frame_properties, "real_height", this->video_codec->height );
+ mlt_properties_set_double( frame_properties, "aspect_ratio", aspect_ratio );
+ if ( mlt_properties_get( properties, "force_progressive" ) )
+ mlt_properties_set_int( frame_properties, "progressive", mlt_properties_get_int( properties, "force_progressive" ) );
+
+ // Add our image operation
+ mlt_frame_push_service( frame, this );
+ mlt_frame_push_get_image( frame, producer_get_image );
+ }
+ else
+ {
+ // If something failed, use test card image
+ mlt_properties_set_int( frame_properties, "test_image", 1 );
+ }
+}
- // Get codec context
- AVCodecContext *codec_context = stream->codec;
+static int seek_audio( producer_avformat this, mlt_position position, double timecode, int *ignore )
+{
+ int paused = 0;
- // Get the codec
- AVCodec *codec = this->video_codec;
+ // Fetch the audio_format
+ AVFormatContext *context = this->audio_format;
- // Update the video properties if the index changed
- if ( index != this->video_index )
+ // Seek if necessary
+ if ( position != this->audio_expected )
+ {
+ if ( position + 1 == this->audio_expected )
{
- // Reset the video properties if the index changed
- this->video_index = index;
- producer_codec_close( this->video_codec );
- this->video_codec = NULL;
- mlt_properties_set_int( properties, "width", codec_context->width );
- mlt_properties_set_int( properties, "height", codec_context->height );
- // TODO: get the first usable AVPacket and reset the stream position
- mlt_properties_set_double( properties, "aspect_ratio",
- get_aspect_ratio( context->streams[ index ], codec_context, NULL ) );
- codec = NULL;
+ // We're paused - silence required
+ paused = 1;
+ }
+ else if ( !this->seekable && position > this->audio_expected && ( position - this->audio_expected ) < 250 )
+ {
+ // Fast forward - seeking is inefficient for small distances - just ignore following frames
+ *ignore = position - this->audio_expected;
+ }
+ else if ( position < this->audio_expected || position - this->audio_expected >= 12 )
+ {
+ int64_t timestamp = ( int64_t )( timecode * AV_TIME_BASE + 0.5 );
+ if ( context->start_time != AV_NOPTS_VALUE )
+ timestamp += context->start_time;
+ if ( timestamp < 0 )
+ timestamp = 0;
+
+ // Set to the real timecode
+ if ( av_seek_frame( context, -1, timestamp, AVSEEK_FLAG_BACKWARD ) != 0 )
+ paused = 1;
+
+ // Clear the usage in the audio buffer
+ int i = MAX_AUDIO_STREAMS + 1;
+ while ( --i )
+ this->audio_used[i - 1] = 0;
}
+ }
+ return paused;
+}
+
+static int decode_audio( producer_avformat this, int *ignore, AVPacket *pkt, int channels, int samples, double timecode, double source_fps )
+{
+ // Fetch the audio_format
+ AVFormatContext *context = this->audio_format;
- // Initialise the codec if necessary
- if ( !codec )
+ // Get the current stream index
+ int index = pkt->stream_index;
+
+ // Get codec context
+ AVCodecContext *codec_context = this->audio_codec[ index ];
+
+ // Obtain the resample context if it exists (not always needed)
+ ReSampleContext *resample = this->audio_resample[ index ];
+
+ // Obtain the audio buffers
+ int16_t *audio_buffer = this->audio_buffer[ index ];
+ int16_t *decode_buffer = this->decode_buffer[ index ];
+
+ int audio_used = this->audio_used[ index ];
+ uint8_t *ptr = pkt->data;
+ int len = pkt->size;
+ int ret = 0;
+
+ while ( ptr && ret >= 0 && len > 0 )
+ {
+ int data_size = sizeof( int16_t ) * AVCODEC_MAX_AUDIO_FRAME_SIZE;
+
+ // Decode the audio
+#if (LIBAVCODEC_VERSION_INT >= ((52<<16)+(26<<8)+0))
+ ret = avcodec_decode_audio3( codec_context, decode_buffer, &data_size, pkt );
+#elif (LIBAVCODEC_VERSION_INT >= ((51<<16)+(29<<8)+0))
+ ret = avcodec_decode_audio2( codec_context, decode_buffer, &data_size, ptr, len );
+#else
+ ret = avcodec_decode_audio( codec_context, decode_buffer, &data_size, ptr, len );
+#endif
+ if ( ret < 0 )
{
- // Initialise multi-threading
- int thread_count = mlt_properties_get_int( properties, "threads" );
- if ( thread_count == 0 && getenv( "MLT_AVFORMAT_THREADS" ) )
- thread_count = atoi( getenv( "MLT_AVFORMAT_THREADS" ) );
- if ( thread_count > 1 )
- {
- avcodec_thread_init( codec_context, thread_count );
- codec_context->thread_count = thread_count;
- }
+ ret = 0;
+ break;
+ }
- // Find the codec
- codec = avcodec_find_decoder( codec_context->codec_id );
+ len -= ret;
+ ptr += ret;
- // If we don't have a codec and we can't initialise it, we can't do much more...
- avformat_lock( );
- if ( codec && avcodec_open( codec_context, codec ) >= 0 )
+ // If decoded successfully
+ if ( data_size > 0 )
+ {
+ // Figure out how many samples will be needed after resampling
+ int convert_samples = data_size / codec_context->channels / ( av_get_bits_per_sample_format( codec_context->sample_fmt ) / 8 );
+ int samples_needed = lrint( this->resample_factor * convert_samples );
+
+ // Resize audio buffer to prevent overflow
+ if ( audio_used * channels + samples_needed > this->audio_buffer_size[ index ] )
+ {
+ this->audio_buffer_size[ index ] *= 2;
+ audio_buffer = this->audio_buffer[ index ] = mlt_pool_realloc( audio_buffer, this->audio_buffer_size[ index ] * sizeof(int16_t) );
+ }
+ if ( resample )
{
- // Now store the codec with its destructor
- producer_codec_close( this->video_codec );
- this->video_codec = codec_context;
+ // Copy to audio buffer while resampling
+ int16_t *source = decode_buffer;
+ int16_t *dest = &audio_buffer[ audio_used * channels ];
+ audio_used += audio_resample( resample, dest, source, convert_samples );
}
else
{
- // Remember that we can't use this later
- this->video_index = index = -1;
+ // Straight copy to audio buffer
+ memcpy( &audio_buffer[ audio_used * codec_context->channels ], decode_buffer, data_size );
+ audio_used += convert_samples;
}
- avformat_unlock( );
-
- // Process properties as AVOptions
- apply_properties( codec_context, properties, AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM );
- }
-
- // No codec, no show...
- if ( codec && index > -1 )
- {
- double source_fps = 0;
- double force_aspect_ratio = mlt_properties_get_double( properties, "force_aspect_ratio" );
- double aspect_ratio = ( force_aspect_ratio > 0.0 ) ?
- force_aspect_ratio : mlt_properties_get_double( properties, "aspect_ratio" );
-
- // Determine the fps
- source_fps = ( double )codec_context->time_base.den /
- ( codec_context->time_base.num == 0 ? 1 : codec_context->time_base.num );
- // If the muxer reports a frame rate different than the codec
- double muxer_fps = av_q2d( context->streams[ index ]->r_frame_rate );
- if ( source_fps != muxer_fps )
- // Choose the lesser - the wrong tends to be off by some multiple of 10
- source_fps = muxer_fps < source_fps ? muxer_fps : source_fps;
-
- // We'll use fps if it's available
- if ( source_fps > 0 )
- mlt_properties_set_double( properties, "source_fps", source_fps );
- else
- mlt_properties_set_double( properties, "source_fps", mlt_producer_get_fps( producer ) );
- mlt_properties_set_double( properties, "aspect_ratio", aspect_ratio );
-
- // Set the width and height
- mlt_properties_set_int( frame_properties, "width", codec_context->width );
- mlt_properties_set_int( frame_properties, "height", codec_context->height );
- mlt_properties_set_int( frame_properties, "real_width", codec_context->width );
- mlt_properties_set_int( frame_properties, "real_height", codec_context->height );
- mlt_properties_set_double( frame_properties, "aspect_ratio", aspect_ratio );
- if ( mlt_properties_get( properties, "force_progressive" ) )
- mlt_properties_set_int( frame_properties, "progressive", mlt_properties_get_int( properties, "force_progressive" ) );
-
- mlt_frame_push_service( frame, this );
- mlt_frame_push_get_image( frame, producer_get_image );
- }
- else
- {
- mlt_properties_set_int( frame_properties, "test_image", 1 );
+ // Handle ignore
+ while ( *ignore && audio_used > samples )
+ {
+ *ignore -= 1;
+ audio_used -= samples;
+ memmove( audio_buffer, &audio_buffer[ samples * (resample? channels : codec_context->channels) ],
+ audio_used * sizeof( int16_t ) );
+ }
}
}
- else
+
+ // If we're behind, ignore this packet
+ if ( pkt->pts >= 0 )
{
- mlt_properties_set_int( frame_properties, "test_image", 1 );
+ double current_pts = av_q2d( context->streams[ index ]->time_base ) * pkt->pts;
+ int req_position = ( int )( timecode * source_fps + 0.5 );
+ int int_position = ( int )( current_pts * source_fps + 0.5 );
+
+ if ( context->start_time != AV_NOPTS_VALUE )
+ int_position -= ( int )( source_fps * context->start_time / AV_TIME_BASE + 0.5 );
+ if ( this->seekable && *ignore == 0 && int_position < req_position )
+ *ignore = 1;
}
+
+ this->audio_used[ index ] = audio_used;
+
+ return ret;
}
/** Get the audio from a frame.
{
// Get the producer
producer_avformat this = mlt_frame_pop_audio( frame );
- mlt_producer producer = &this->parent;
-
- // Get the properties from the frame
- mlt_properties frame_properties = MLT_FRAME_PROPERTIES( frame );
// Obtain the frame number of this frame
- mlt_position position = mlt_properties_get_position( frame_properties, "avformat_position" );
-
- // Get the producer properties
- mlt_properties properties = MLT_PRODUCER_PROPERTIES( producer );
-
- // Fetch the audio_context
- AVFormatContext *context = this->audio_context;
-
- // Get the audio stream
- AVStream *stream = context->streams[ this->audio_index ];
-
- // Get codec context
- AVCodecContext *codec_context = stream->codec;
-
- // Packet
- AVPacket pkt;
-
- // Obtain the resample context if it exists (not always needed)
- ReSampleContext *resample = this->audio_resample;
-
- // Obtain the audio buffers
- int16_t *audio_buffer = this->audio_buffer;
- int16_t *decode_buffer = this->decode_buffer;
-
- // Get amount of audio used
- int audio_used = this->audio_used;
+ mlt_position position = mlt_properties_get_position( MLT_FRAME_PROPERTIES( frame ), "avformat_position" );
// Calculate the real time code
- double real_timecode = producer_time_of_frame( producer, position );
+ double real_timecode = producer_time_of_frame( &this->parent, position );
+
+ // Get the source fps
+ double source_fps = mlt_properties_get_double( MLT_PRODUCER_PROPERTIES( &this->parent ), "source_fps" );
// Number of frames to ignore (for ffwd)
int ignore = 0;
// Flag for paused (silence)
- int paused = 0;
-
- // Check for resample and create if necessary
- if ( !resample && codec_context->channels <= 2 )
- {
- // Create the resampler
-#if (LIBAVCODEC_VERSION_INT >= ((52<<16)+(15<<8)+0))
- resample = av_audio_resample_init( *channels, codec_context->channels, *frequency, codec_context->sample_rate,
- SAMPLE_FMT_S16, codec_context->sample_fmt, 16, 10, 0, 0.8 );
-#else
- resample = audio_resample_init( *channels, codec_context->channels, *frequency, codec_context->sample_rate );
-#endif
-
- // And store it on properties
- if ( this->audio_resample ) audio_resample_close( this->audio_resample );
- this->audio_resample = resample;
- }
- else if ( !resample )
+ int paused = seek_audio( this, position, real_timecode, &ignore );
+
+ // Fetch the audio_format
+ AVFormatContext *context = this->audio_format;
+
+ // Determine the tracks to use
+ int index = this->audio_index;
+ int index_max = this->audio_index + 1;
+ if ( this->audio_index == INT_MAX )
{
- // TODO: uncomment and remove following line when full multi-channel support is ready
- // *channels = codec_context->channels;
- codec_context->request_channels = *channels;
-
- *frequency = codec_context->sample_rate;
+ index = 0;
+ index_max = context->nb_streams;
+ *channels = this->total_channels;
+ *frequency = this->max_frequency;
}
- // Check for audio buffer and create if necessary
- if ( !audio_buffer )
- this->audio_buffer = audio_buffer = mlt_pool_alloc( AVCODEC_MAX_AUDIO_FRAME_SIZE * sizeof( int16_t ) );
-
- // Check for decoder buffer and create if necessary
- if ( !decode_buffer )
- this->decode_buffer = decode_buffer = av_malloc( AVCODEC_MAX_AUDIO_FRAME_SIZE * sizeof( int16_t ) );
-
- // Seek if necessary
- if ( position != this->audio_expected )
+ // Initialize the resamplers and buffers
+ for ( ; index < index_max; index++ )
{
- if ( position + 1 == this->audio_expected )
- {
- // We're paused - silence required
- paused = 1;
- }
- else if ( !this->seekable && position > this->audio_expected && ( position - this->audio_expected ) < 250 )
- {
- // Fast forward - seeking is inefficient for small distances - just ignore following frames
- ignore = position - this->audio_expected;
- }
- else if ( position < this->audio_expected || position - this->audio_expected >= 12 )
+ // Get codec context
+ AVCodecContext *codec_context = this->audio_codec[ index ];
+
+ if ( codec_context && !this->audio_buffer[ index ] )
{
- int64_t timestamp = ( int64_t )( real_timecode * AV_TIME_BASE + 0.5 );
- if ( context->start_time != AV_NOPTS_VALUE )
- timestamp += context->start_time;
- if ( timestamp < 0 )
- timestamp = 0;
+ // Check for resample and create if necessary
+ if ( codec_context->channels <= 2 )
+ {
+ // Determine by how much resampling will increase number of samples
+ double resample_factor = this->audio_index == INT_MAX ? 1 : (double) *channels / codec_context->channels;
+ resample_factor *= (double) *frequency / codec_context->sample_rate;
+ if ( resample_factor > this->resample_factor )
+ this->resample_factor = resample_factor;
+
+ // Create the resampler
+#if (LIBAVCODEC_VERSION_INT >= ((52<<16)+(15<<8)+0))
+ this->audio_resample[ index ] = av_audio_resample_init(
+ this->audio_index == INT_MAX ? codec_context->channels : *channels,
+ codec_context->channels, *frequency, codec_context->sample_rate,
+ SAMPLE_FMT_S16, codec_context->sample_fmt, 16, 10, 0, 0.8 );
+#else
+ this->audio_resample[ index ] = audio_resample_init(
+ this->audio_index == INT_MAX ? codec_context->channels : *channels,
+ codec_context->channels, *frequency, codec_context->sample_rate );
+#endif
+ }
+ else
+ {
+ codec_context->request_channels = this->audio_index == INT_MAX ? codec_context->channels : *channels;
+ }
- // Set to the real timecode
- if ( av_seek_frame( context, -1, timestamp, AVSEEK_FLAG_BACKWARD ) != 0 )
- paused = 1;
+ // Check for audio buffer and create if necessary
+ this->audio_buffer_size[ index ] = AVCODEC_MAX_AUDIO_FRAME_SIZE;
+ this->audio_buffer[ index ] = mlt_pool_alloc( this->audio_buffer_size[ index ] * sizeof( int16_t ) );
- // Clear the usage in the audio buffer
- audio_used = 0;
+ // Check for decoder buffer and create if necessary
+ this->decode_buffer[ index ] = av_malloc( AVCODEC_MAX_AUDIO_FRAME_SIZE * sizeof( int16_t ) );
}
}
{
int ret = 0;
int got_audio = 0;
+ AVPacket pkt;
av_init_packet( &pkt );
- while( ret >= 0 && !got_audio )
+ while ( ret >= 0 && !got_audio )
{
// Check if the buffer already contains the samples required
- if ( audio_used >= *samples && ignore == 0 )
+ if ( this->audio_index != INT_MAX && this->audio_used[ this->audio_index ] >= *samples && ignore == 0 )
{
got_audio = 1;
break;
// Read a packet
ret = av_read_frame( context, &pkt );
- int len = pkt.size;
- uint8_t *ptr = pkt.data;
+ // We only deal with audio from the selected audio index
+ if ( ret >= 0 && pkt.data && pkt.size > 0 && ( pkt.stream_index == this->audio_index ||
+ ( this->audio_index == INT_MAX && context->streams[ pkt.stream_index ]->codec->codec_type == CODEC_TYPE_AUDIO ) ) )
+ ret = decode_audio( this, &ignore, &pkt, *channels, *samples, real_timecode, source_fps );
+ av_free_packet( &pkt );
- // We only deal with audio from the selected audio_index
- while ( ptr && ret >= 0 && pkt.stream_index == this->audio_index && len > 0 )
+ if ( this->audio_index == INT_MAX && ret >= 0 )
{
- int data_size = sizeof( int16_t ) * AVCODEC_MAX_AUDIO_FRAME_SIZE;
-
- // Decode the audio
-#if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(29<<8)+0))
- ret = avcodec_decode_audio2( codec_context, decode_buffer, &data_size, ptr, len );
-#else
- ret = avcodec_decode_audio( codec_context, decode_buffer, &data_size, ptr, len );
-#endif
- if ( ret < 0 )
- {
- ret = 0;
- break;
- }
-
- len -= ret;
- ptr += ret;
-
- if ( data_size > 0 && ( audio_used * *channels + data_size < AVCODEC_MAX_AUDIO_FRAME_SIZE ) )
- {
- if ( resample )
- {
- int16_t *source = decode_buffer;
- int16_t *dest = &audio_buffer[ audio_used * *channels ];
- int convert_samples = data_size / av_get_bits_per_sample_format( codec_context->sample_fmt )
- * 8 / codec_context->channels;
-
- audio_used += audio_resample( resample, dest, source, convert_samples );
- }
- else
- {
- memcpy( &audio_buffer[ audio_used * *channels ], decode_buffer, data_size );
- audio_used += data_size / *channels / av_get_bits_per_sample_format( codec_context->sample_fmt ) * 8;
- }
-
- // Handle ignore
- while ( ignore && audio_used > *samples )
- {
- ignore --;
- audio_used -= *samples;
- memmove( audio_buffer, &audio_buffer[ *samples * *channels ], audio_used * sizeof( int16_t ) );
- }
- }
-
- // If we're behind, ignore this packet
- if ( pkt.pts >= 0 )
+ // Determine if there is enough audio for all streams
+ got_audio = 1;
+ for ( index = 0; index < context->nb_streams; index++ )
{
- double current_pts = av_q2d( stream->time_base ) * pkt.pts;
- double source_fps = mlt_properties_get_double( properties, "source_fps" );
- int req_position = ( int )( real_timecode * source_fps + 0.5 );
- int int_position = ( int )( current_pts * source_fps + 0.5 );
-
- if ( context->start_time != AV_NOPTS_VALUE )
- int_position -= ( int )( context->start_time * source_fps / AV_TIME_BASE + 0.5 );
- if ( this->seekable && !ignore && int_position < req_position )
- ignore = 1;
+ if ( this->audio_codec[ index ] && this->audio_used[ index ] < *samples )
+ got_audio = 0;
}
}
-
- // We're finished with this packet regardless
- av_free_packet( &pkt );
}
- int size = *samples * *channels * sizeof( int16_t );
- *format = mlt_audio_s16;
+ // Allocate and set the frame's audio buffer
+ int size = *samples * *channels * sizeof(int16_t);
*buffer = mlt_pool_alloc( size );
+ *format = mlt_audio_s16;
mlt_frame_set_audio( frame, *buffer, *format, size, mlt_pool_release );
- // Now handle the audio if we have enough
- if ( audio_used >= *samples )
+ // Interleave tracks if audio_index=all
+ if ( this->audio_index == INT_MAX )
{
- memcpy( *buffer, audio_buffer, *samples * *channels * sizeof( int16_t ) );
- audio_used -= *samples;
- memmove( audio_buffer, &audio_buffer[ *samples * *channels ], audio_used * *channels * sizeof( int16_t ) );
+ int16_t *dest = *buffer;
+ int i;
+ for ( i = 0; i < *samples; i++ )
+ {
+ for ( index = 0; index < index_max; index++ )
+ if ( this->audio_codec[ index ] )
+ {
+ int current_channels = this->audio_codec[ index ]->channels;
+ int16_t *src = this->audio_buffer[ index ] + i * current_channels;
+ memcpy( dest, src, current_channels * sizeof(int16_t) );
+ dest += current_channels;
+ }
+ }
+ for ( index = 0; index < index_max; index++ )
+ if ( this->audio_codec[ index ] )
+ {
+ int current_channels = this->audio_codec[ index ]->channels;
+ int16_t *src = this->audio_buffer[ index ] + *samples * current_channels;
+ this->audio_used[index] -= *samples;
+ memmove( this->audio_buffer[ index ], src, this->audio_used[ index ] * current_channels * sizeof(int16_t) );
+ }
}
+ // Copy a single track to the output buffer
else
{
- memset( *buffer, 0, *samples * *channels * sizeof( int16_t ) );
- }
+ index = this->audio_index;
- // Store the number of audio samples still available
- this->audio_used = audio_used;
+ // Now handle the audio if we have enough
+ if ( this->audio_used[ index ] >= *samples )
+ {
+ int16_t *src = this->audio_buffer[ index ];
+ memcpy( *buffer, src, *samples * *channels * sizeof(int16_t) );
+ this->audio_used[ index ] -= *samples;
+ memmove( src, &src[ *samples * *channels ], this->audio_used[ index ] * *channels * sizeof(int16_t) );
+ }
+ else
+ {
+ // Otherwise fill with silence
+ memset( *buffer, 0, *samples * *channels * sizeof(int16_t) );
+ }
+ if ( !this->audio_resample[ index ] )
+ {
+ // TODO: uncomment and remove following line when full multi-channel support is ready
+ // *channels = codec_context->channels;
+ *frequency = this->audio_codec[ index ]->sample_rate;
+ }
+ }
}
else
{
return 0;
}
+/** Initialize the audio codec context.
+*/
+
+static int audio_codec_init( producer_avformat this, int index, mlt_properties properties )
+{
+ // Initialise the codec if necessary
+ if ( !this->audio_codec[ index ] )
+ {
+ // Get codec context
+ AVCodecContext *codec_context = this->audio_format->streams[index]->codec;
+
+ // Find the codec
+ AVCodec *codec = avcodec_find_decoder( codec_context->codec_id );
+
+ // If we don't have a codec and we can't initialise it, we can't do much more...
+ avformat_lock( );
+ if ( codec && avcodec_open( codec_context, codec ) >= 0 )
+ {
+ // Now store the codec with its destructor
+ avformat_unlock();
+ producer_codec_close( this->audio_codec[index] );
+ this->audio_codec[ index ] = codec_context;
+ }
+ else
+ {
+ // Remember that we can't use this later
+ this->audio_index = -1;
+ avformat_unlock( );
+ }
+
+ // Process properties as AVOptions
+ apply_properties( codec_context, properties, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_DECODING_PARAM );
+ }
+ return this->audio_codec[ index ] && this->audio_index > -1;
+}
+
/** Set up audio handling.
*/
// Get the properties
mlt_properties properties = MLT_PRODUCER_PROPERTIES( producer );
- // Fetch the audio_context
- AVFormatContext *context = this->audio_context;
+ // Fetch the audio format context
+ AVFormatContext *context = this->audio_format;
+
+ mlt_properties frame_properties = MLT_FRAME_PROPERTIES( frame );
// Get the audio_index
int index = mlt_properties_get_int( properties, "audio_index" );
+ // Handle all audio tracks
+ if ( mlt_properties_get( properties, "audio_index" ) &&
+ !strcmp( mlt_properties_get( properties, "audio_index" ), "all" ) )
+ index = INT_MAX;
+
// Reopen the file if necessary
if ( !context && index > -1 )
{
mlt_events_block( properties, producer );
producer_open( this, mlt_service_profile( MLT_PRODUCER_SERVICE(producer) ),
mlt_properties_get( properties, "resource" ) );
- context = this->audio_context;
- producer_file_close( this->dummy_context );
+ context = this->audio_format;
+ producer_format_close( this->dummy_context );
this->dummy_context = NULL;
mlt_events_unblock( properties, producer );
+ get_audio_streams_info( this );
}
// Exception handling for audio_index
- if ( context && index >= (int) context->nb_streams )
+ if ( context && index >= (int) context->nb_streams && index < INT_MAX )
{
for ( index = context->nb_streams - 1;
index >= 0 && context->streams[ index ]->codec->codec_type != CODEC_TYPE_AUDIO;
index-- );
mlt_properties_set_int( properties, "audio_index", index );
}
- if ( context && index > -1 && context->streams[ index ]->codec->codec_type != CODEC_TYPE_AUDIO )
+ if ( context && index > -1 && index < INT_MAX &&
+ context->streams[ index ]->codec->codec_type != CODEC_TYPE_AUDIO )
{
index = -1;
mlt_properties_set_int( properties, "audio_index", index );
// Update the audio properties if the index changed
if ( index > -1 && index != this->audio_index )
{
- this->audio_index = index;
- producer_codec_close( this->audio_codec );
- this->audio_codec = NULL;
+ producer_codec_close( this->audio_codec[ this->audio_index ] );
+ this->audio_codec[ this->audio_index ] = NULL;
}
+ this->audio_index = index;
- // Deal with audio context
- if ( context && index > -1 )
+ // Get the codec(s)
+ if ( context && index == INT_MAX )
{
- // Get the frame properties
- mlt_properties frame_properties = MLT_FRAME_PROPERTIES( frame );
-
- // Get the audio stream
- AVStream *stream = context->streams[ index ];
-
- // Get codec context
- AVCodecContext *codec_context = stream->codec;
-
- // Get the codec
- AVCodec *codec = this->audio_codec;
-
- // Initialise the codec if necessary
- if ( !codec )
+ mlt_properties_set_int( frame_properties, "frequency", this->max_frequency );
+ mlt_properties_set_int( frame_properties, "channels", this->total_channels );
+ for ( index = 0; index < context->nb_streams; index++ )
{
- // Find the codec
- codec = avcodec_find_decoder( codec_context->codec_id );
-
- // If we don't have a codec and we can't initialise it, we can't do much more...
- avformat_lock( );
- if ( codec && avcodec_open( codec_context, codec ) >= 0 )
- {
- // Now store the codec with its destructor
- producer_codec_close( this->audio_codec );
- this->audio_codec = codec_context;
-
- }
- else
- {
- // Remember that we can't use this later
- this->audio_index = index = -1;
- }
- avformat_unlock( );
-
- // Process properties as AVOptions
- apply_properties( codec_context, properties, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_DECODING_PARAM );
+ if ( context->streams[ index ]->codec->codec_type == CODEC_TYPE_AUDIO )
+ audio_codec_init( this, index, properties );
}
-
- // No codec, no show...
- if ( codec && index > -1 )
+ }
+ else if ( context && index > -1 && audio_codec_init( this, index, properties ) )
+ {
+ // Set the frame properties
+ if ( index < INT_MAX )
{
- mlt_frame_push_audio( frame, this );
- mlt_frame_push_audio( frame, producer_get_audio );
- mlt_properties_set_int( frame_properties, "frequency", codec_context->sample_rate );
- mlt_properties_set_int( frame_properties, "channels", codec_context->channels );
+ mlt_properties_set_int( frame_properties, "frequency", this->audio_codec[ index ]->sample_rate );
+ mlt_properties_set_int( frame_properties, "channels", this->audio_codec[ index ]->channels );
}
}
+ if ( context && index > -1 )
+ {
+ // Add our audio operation
+ mlt_frame_push_audio( frame, this );
+ mlt_frame_push_audio( frame, producer_get_audio );
+ }
}
/** Our get frame implementation.
// Set up the audio
producer_set_up_audio( this, *frame );
- // Set the aspect_ratio
- mlt_properties_set_double( MLT_FRAME_PROPERTIES( *frame ), "aspect_ratio",
- mlt_properties_get_double( MLT_PRODUCER_PROPERTIES( producer ), "aspect_ratio" ) );
-
// Calculate the next timecode
mlt_producer_prepare_next( producer );
// Close the file
av_free( this->av_frame );
- if ( this->audio_resample )
- audio_resample_close( this->audio_resample );
- mlt_pool_release( this->audio_buffer );
- av_free( this->decode_buffer );
- producer_file_close( this->dummy_context );
- producer_file_close( this->audio_context );
- producer_file_close( this->video_context );
- producer_codec_close( this->audio_codec );
+ int i;
+ for ( i = 0; i < MAX_AUDIO_STREAMS; i++ )
+ {
+ if ( this->audio_resample[i] )
+ audio_resample_close( this->audio_resample[i] );
+ mlt_pool_release( this->audio_buffer[i] );
+ av_free( this->decode_buffer[i] );
+ producer_codec_close( this->audio_codec[i] );
+ }
producer_codec_close( this->video_codec );
+ producer_format_close( this->dummy_context );
+ producer_format_close( this->audio_format );
+ producer_format_close( this->video_format );
// Close the parent
parent->close = NULL;