X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2Fmodules%2Favformat%2Fproducer_avformat.c;h=4ddec9aa47dcf84fb032a2493b636f01d3066c6a;hb=abc7d71a839a7450280f44a278ab90238a7abb36;hp=1349fe09ce25463e7e96557ba356ab6313a7f24f;hpb=d1be09670b5f37b4c5572806d8c13ea180e16e79;p=mlt diff --git a/src/modules/avformat/producer_avformat.c b/src/modules/avformat/producer_avformat.c index 1349fe09..4ddec9aa 100644 --- a/src/modules/avformat/producer_avformat.c +++ b/src/modules/avformat/producer_avformat.c @@ -25,8 +25,12 @@ // ffmpeg Header files #include +#include #ifdef SWSCALE -#include +# include +#endif +#if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(71<<8)+0)) +# include "audioconvert.h" #endif // System header files @@ -94,34 +98,77 @@ mlt_producer producer_avformat_init( mlt_profile profile, char *file ) /** Find the default streams. */ -static void find_default_streams( AVFormatContext *context, int *audio_index, int *video_index ) +static mlt_properties find_default_streams( mlt_properties meta_media, AVFormatContext *context, int *audio_index, int *video_index ) { int i; + char key[200]; + + mlt_properties_set_int( meta_media, "meta.media.nb_streams", context->nb_streams ); // Allow for multiple audio and video streams in the file and select first of each (if available) - for( i = 0; i < context->nb_streams; i++ ) + for( i = 0; i < context->nb_streams; i++ ) { // Get the codec context - AVCodecContext *codec_context = context->streams[ i ]->codec; + AVStream *stream = context->streams[ i ]; + if ( ! stream ) continue; + AVCodecContext *codec_context = stream->codec; + if ( ! codec_context ) continue; + AVCodec *codec = avcodec_find_decoder( codec_context->codec_id ); + if ( ! codec ) continue; - if ( avcodec_find_decoder( codec_context->codec_id ) == NULL ) - continue; + snprintf( key, sizeof(key), "meta.media.%d.stream.type", i ); // Determine the type and obtain the first index of each type - switch( codec_context->codec_type ) + switch( codec_context->codec_type ) { case CODEC_TYPE_VIDEO: if ( *video_index < 0 ) *video_index = i; + mlt_properties_set( meta_media, key, "video" ); + snprintf( key, sizeof(key), "meta.media.%d.stream.frame_rate", i ); + mlt_properties_set_double( meta_media, key, av_q2d( context->streams[ i ]->r_frame_rate ) ); +#if LIBAVFORMAT_VERSION_INT >= ((52<<16)+(21<<8)+0) + 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.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_double( meta_media, key, av_q2d( codec_context->sample_aspect_ratio ) ); break; case CODEC_TYPE_AUDIO: if ( *audio_index < 0 ) *audio_index = i; + mlt_properties_set( meta_media, key, "audio" ); +#if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(71<<8)+0)) + snprintf( key, sizeof(key), "meta.media.%d.codec.sample_fmt", i ); + mlt_properties_set( meta_media, key, avcodec_get_sample_fmt_name( codec_context->sample_fmt ) ); +#endif + snprintf( key, sizeof(key), "meta.media.%d.codec.sample_rate", i ); + mlt_properties_set_int( meta_media, key, codec_context->sample_rate ); + snprintf( key, sizeof(key), "meta.media.%d.codec.channels", i ); + mlt_properties_set_int( meta_media, key, codec_context->channels ); break; default: break; } +// snprintf( key, sizeof(key), "meta.media.%d.stream.time_base", i ); +// mlt_properties_set_double( meta_media, key, av_q2d( context->streams[ i ]->time_base ) ); + snprintf( key, sizeof(key), "meta.media.%d.codec.name", i ); + mlt_properties_set( meta_media, key, codec->name ); + snprintf( key, sizeof(key), "meta.media.%d.codec.long_name", i ); + mlt_properties_set( meta_media, key, codec->long_name ); + snprintf( key, sizeof(key), "meta.media.%d.codec.bit_rate", 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 ); } + + return meta_media; } /** Producer file destructor. @@ -261,7 +308,7 @@ static int producer_open( mlt_producer this, mlt_profile profile, char *file ) // Lock the mutex now avformat_lock( ); - + // If "MRL", then create AVInputFormat AVInputFormat *format = NULL; AVFormatParameters *params = NULL; @@ -270,7 +317,7 @@ static int producer_open( mlt_producer this, mlt_profile profile, char *file ) // AV option (0 = both, 1 = video, 2 = audio) int av = 0; - + // Setting lowest log level av_log_set_level( -1 ); @@ -279,18 +326,18 @@ static int producer_open( mlt_producer this, mlt_profile profile, char *file ) { // 'file' becomes format abbreviation mrl[0] = 0; - + // Lookup the format format = av_find_input_format( file ); - + // Eat the format designator file = ++mrl; - + if ( format ) { // Allocate params params = calloc( sizeof( AVFormatParameters ), 1 ); - + // These are required by video4linux (defaults) params->width = 640; params->height = 480; @@ -299,7 +346,7 @@ static int producer_open( mlt_producer this, mlt_profile profile, char *file ) params->channels = 2; params->sample_rate = 48000; } - + // XXX: this does not work anymore since avdevice // TODO: make producer_avddevice? // Parse out params @@ -343,7 +390,7 @@ static int producer_open( mlt_producer this, mlt_profile profile, char *file ) // Now attempt to open the file error = av_open_input_file( &context, file, format, 0, params ) < 0; - + // Cleanup AVFormatParameters free( standard ); free( params ); @@ -363,7 +410,7 @@ static int producer_open( mlt_producer this, mlt_profile profile, char *file ) int av_bypass = 0; // Now set properties where we can (use default unknowns if required) - if ( context->duration != AV_NOPTS_VALUE ) + if ( context->duration != AV_NOPTS_VALUE ) { // This isn't going to be accurate for all formats mlt_position frames = ( mlt_position )( ( ( double )context->duration / ( double )AV_TIME_BASE ) * fps + 0.5 ); @@ -372,11 +419,11 @@ static int producer_open( mlt_producer this, mlt_profile profile, char *file ) } // Find default audio and video streams - find_default_streams( context, &audio_index, &video_index ); + find_default_streams( properties, context, &audio_index, &video_index ); if ( context->start_time != AV_NOPTS_VALUE ) mlt_properties_set_double( properties, "_start_time", context->start_time ); - + // Check if we're seekable (something funny about mpeg here :-/) if ( strcmp( file, "pipe:" ) && strncmp( file, "http://", 6 ) && strncmp( file, "udp:", 4 ) && strncmp( file, "tcp:", 4 ) && strncmp( file, "rtsp:", 5 ) && strncmp( file, "rtp:", 4 ) ) { @@ -424,21 +471,21 @@ static int producer_open( mlt_producer this, mlt_profile profile, char *file ) } // Read Metadata - if (context->title != NULL) + if (context->title != NULL) mlt_properties_set(properties, "meta.attr.title.markup", context->title ); - if (context->author != NULL) + if (context->author != NULL) mlt_properties_set(properties, "meta.attr.author.markup", context->author ); - if (context->copyright != NULL) + if (context->copyright != NULL) mlt_properties_set(properties, "meta.attr.copyright.markup", context->copyright ); - if (context->comment != NULL) + if (context->comment != NULL) mlt_properties_set(properties, "meta.attr.comment.markup", context->comment ); - if (context->album != NULL) + if (context->album != NULL) mlt_properties_set(properties, "meta.attr.album.markup", context->album ); - if (context->year != 0) + if (context->year != 0) mlt_properties_set_int(properties, "meta.attr.year.markup", context->year ); - if (context->track != 0) + if (context->track != 0) mlt_properties_set_int(properties, "meta.attr.track.markup", context->track ); - + // 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 ) { @@ -601,7 +648,7 @@ static int producer_get_image( mlt_frame frame, uint8_t **buffer, mlt_image_form // Obtain the frame number of this frame mlt_position position = mlt_properties_get_position( frame_properties, "avformat_position" ); - // Get the producer + // Get the producer mlt_producer this = mlt_properties_get_data( frame_properties, "avformat_producer", NULL ); // Get the producer properties @@ -681,7 +728,7 @@ static int producer_get_image( mlt_frame frame, uint8_t **buffer, mlt_image_form // Set to the timestamp av_seek_frame( context, -1, timestamp, AVSEEK_FLAG_BACKWARD ); - + // Remove the cached info relating to the previous position mlt_properties_set_int( properties, "_current_position", -1 ); mlt_properties_set_int( properties, "_last_position", -1 ); @@ -711,10 +758,7 @@ static int producer_get_image( mlt_frame frame, uint8_t **buffer, mlt_image_form // Construct an AVFrame for YUV422 conversion if ( av_frame == NULL ) - { av_frame = avcodec_alloc_frame( ); - mlt_properties_set_data( properties, "av_frame", av_frame, 0, av_free, NULL ); - } while( ret >= 0 && !got_picture ) { @@ -771,6 +815,7 @@ static int producer_get_image( mlt_frame frame, uint8_t **buffer, mlt_image_form mlt_properties_set_int( properties, "top_field_first", av_frame->top_field_first ); mlt_properties_set_int( properties, "_current_position", int_position ); mlt_properties_set_int( properties, "_got_picture", 1 ); + mlt_properties_set_data( properties, "av_frame", av_frame, 0, av_free, NULL ); } else { @@ -796,6 +841,28 @@ static int producer_get_image( mlt_frame frame, uint8_t **buffer, mlt_image_form return 0; } +/** Process properties as AVOptions and apply to AV context obj +*/ + +static void apply_properties( void *obj, mlt_properties properties, int flags ) +{ + int i; + int count = mlt_properties_count( properties ); + for ( i = 0; i < count; i++ ) + { + const char *opt_name = mlt_properties_get_name( properties, i ); + const AVOption *opt = av_find_opt( obj, opt_name, NULL, flags, flags ); + if ( opt != NULL ) +#if LIBAVCODEC_VERSION_INT >= ((52<<16)+(7<<8)+0) + av_set_string3( obj, opt_name, mlt_properties_get( properties, opt_name), 0, NULL ); +#elif LIBAVCODEC_VERSION_INT >= ((51<<16)+(59<<8)+0) + av_set_string2( obj, opt_name, mlt_properties_get( properties, opt_name), 0 ); +#else + av_set_string( obj, opt_name, mlt_properties_get( properties, opt_name) ); +#endif + } +} + /** Set up video handling. */ @@ -819,6 +886,9 @@ static void producer_set_up_video( mlt_producer this, mlt_frame frame ) context = mlt_properties_get_data( properties, "video_context", NULL ); mlt_properties_set_data( properties, "dummy_context", NULL, 0, NULL, NULL ); mlt_events_unblock( properties, this ); + + // Process properties as AVOptions + apply_properties( context, properties, AV_OPT_FLAG_DECODING_PARAM ); } // Exception handling for video_index @@ -862,11 +932,11 @@ static void producer_set_up_video( mlt_producer this, mlt_frame frame ) get_aspect_ratio( context->streams[ index ], codec_context, NULL ) ); codec = NULL; } - + // Initialise the codec if necessary if ( codec == NULL ) { - // Initialise multi-threading + // 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" ) ); @@ -875,7 +945,7 @@ static void producer_set_up_video( mlt_producer this, mlt_frame frame ) avcodec_thread_init( codec_context, thread_count ); codec_context->thread_count = thread_count; } - + // Find the codec codec = avcodec_find_decoder( codec_context->codec_id ); @@ -893,6 +963,9 @@ static void producer_set_up_video( mlt_producer this, mlt_frame frame ) index = -1; } 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... @@ -912,10 +985,12 @@ static void producer_set_up_video( mlt_producer this, mlt_frame frame ) else mlt_properties_set_double( properties, "source_fps", mlt_producer_get_fps( this ) ); 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 ); mlt_frame_push_get_image( frame, producer_get_image ); @@ -943,7 +1018,7 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form // Obtain the frame number of this frame mlt_position position = mlt_properties_get_position( frame_properties, "avformat_position" ); - // Get the producer + // Get the producer mlt_producer this = mlt_properties_get_data( frame_properties, "avformat_producer", NULL ); // Get the producer properties @@ -964,8 +1039,15 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form // Obtain the resample context if it exists (not always needed) ReSampleContext *resample = mlt_properties_get_data( properties, "audio_resample", NULL ); - // Obtain the audio buffer +#if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(71<<8)+0)) + // Get the format converter context if it exists + AVAudioConvert *convert = mlt_properties_get_data( properties, "audio_convert", NULL ); +#endif + + // Obtain the audio buffers int16_t *audio_buffer = mlt_properties_get_data( properties, "audio_buffer", NULL ); + int16_t *decode_buffer = mlt_properties_get_data( properties, "decode_buffer", NULL ); + int16_t *convert_buffer = mlt_properties_get_data( properties, "convert_buffer", NULL ); // Get amount of audio used int audio_used = mlt_properties_get_int( properties, "_audio_used" ); @@ -985,7 +1067,7 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form // Number of frames to ignore (for ffwd) int ignore = 0; - // Flag for paused (silence) + // Flag for paused (silence) int paused = 0; // Check for resample and create if necessary @@ -1003,6 +1085,17 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form *frequency = codec_context->sample_rate; } +#if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(71<<8)+0)) + // Check for audio format converter and create if necessary + // TODO: support higher resolutions than 16-bit. + if ( convert == NULL && codec_context->sample_fmt != SAMPLE_FMT_S16 ) + { + // Create single channel converter for interleaved with no mixing matrix + convert = av_audio_convert_alloc( SAMPLE_FMT_S16, 1, codec_context->sample_fmt, 1, NULL, 0 ); + mlt_properties_set_data( properties, "audio_convert", convert, 0, ( mlt_destructor )av_audio_convert_free, NULL ); + } +#endif + // Check for audio buffer and create if necessary if ( audio_buffer == NULL ) { @@ -1013,6 +1106,28 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form mlt_properties_set_data( properties, "audio_buffer", audio_buffer, 0, ( mlt_destructor )mlt_pool_release, NULL ); } + // Check for decoder buffer and create if necessary + if ( decode_buffer == NULL ) + { + // Allocate the audio buffer + decode_buffer = av_malloc( AVCODEC_MAX_AUDIO_FRAME_SIZE * sizeof( int16_t ) ); + + // And store it on properties for reuse + mlt_properties_set_data( properties, "decode_buffer", decode_buffer, 0, ( mlt_destructor )av_free, NULL ); + } + +#if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(71<<8)+0)) + // Check for format converter buffer and create if necessary + if ( resample && convert && convert_buffer == NULL ) + { + // Allocate the audio buffer + convert_buffer = mlt_pool_alloc( AVCODEC_MAX_AUDIO_FRAME_SIZE * sizeof( int16_t ) ); + + // And store it on properties for reuse + mlt_properties_set_data( properties, "convert_buffer", convert_buffer, 0, ( mlt_destructor )mlt_pool_release, NULL ); + } +#endif + // Seek if necessary if ( position != expected ) { @@ -1042,7 +1157,6 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form { int ret = 0; int got_audio = 0; - int16_t *temp = av_malloc( sizeof( int16_t ) * AVCODEC_MAX_AUDIO_FRAME_SIZE ); av_init_packet( &pkt ); @@ -1068,9 +1182,9 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form // Decode the audio #if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(29<<8)+0)) - ret = avcodec_decode_audio2( codec_context, temp, &data_size, ptr, len ); + ret = avcodec_decode_audio2( codec_context, decode_buffer, &data_size, ptr, len ); #else - ret = avcodec_decode_audio( codec_context, temp, &data_size, ptr, len ); + ret = avcodec_decode_audio( codec_context, decode_buffer, &data_size, ptr, len ); #endif if ( ret < 0 ) { @@ -1083,14 +1197,41 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form if ( data_size > 0 ) { - if ( resample != NULL ) + int src_stride[6]= { av_get_bits_per_sample_format( codec_context->sample_fmt ) / 8 }; + int dst_stride[6]= { av_get_bits_per_sample_format( SAMPLE_FMT_S16 ) / 8 }; + + if ( resample ) { - audio_used += audio_resample( resample, &audio_buffer[ audio_used * *channels ], temp, data_size / ( codec_context->channels * sizeof( int16_t ) ) ); + int16_t *source = decode_buffer; + int16_t *dest = &audio_buffer[ audio_used * *channels ]; + int convert_samples = data_size / src_stride[0]; + +#if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(71<<8)+0)) + if ( convert ) + { + const void *src_buf[6] = { decode_buffer }; + void *dst_buf[6] = { convert_buffer }; + av_audio_convert( convert, dst_buf, dst_stride, src_buf, src_stride, convert_samples ); + source = convert_buffer; + } +#endif + audio_used += audio_resample( resample, dest, source, convert_samples / codec_context->channels ); } else { - memcpy( &audio_buffer[ audio_used * *channels ], temp, data_size ); - audio_used += data_size / ( codec_context->channels * sizeof( int16_t ) ); +#if (LIBAVCODEC_VERSION_INT >= ((51<<16)+(71<<8)+0)) + if ( convert ) + { + const void *src_buf[6] = { decode_buffer }; + void *dst_buf[6] = { &audio_buffer[ audio_used * *channels ] }; + av_audio_convert( convert, dst_buf, dst_stride, src_buf, src_stride, data_size / src_stride[0] ); + } + else +#endif + { + memcpy( &audio_buffer[ audio_used * *channels ], decode_buffer, data_size ); + } + audio_used += data_size / *channels / src_stride[0]; } // Handle ignore @@ -1129,12 +1270,9 @@ static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_form { memset( *buffer, 0, *samples * *channels * sizeof( int16_t ) ); } - + // Store the number of audio samples still available mlt_properties_set_int( properties, "_audio_used", audio_used ); - - // Release the temporary audio - av_free( temp ); } else { @@ -1229,6 +1367,9 @@ static void producer_set_up_audio( mlt_producer this, mlt_frame frame ) index = -1; } avformat_unlock( ); + + // Process properties as AVOptions + apply_properties( codec_context, properties, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_DECODING_PARAM ); } // No codec, no show...