add support for ffmpeg libswscale
[melted] / src / modules / avformat / producer_avformat.c
1 /*
2 * producer_avformat.c -- avformat producer
3 * Copyright (C) 2003-2004 Ushodaya Enterprises Limited
4 * Author: Charles Yates <charles.yates@pandora.be>
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software Foundation,
18 * Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
19 */
20
21 // Local header files
22 #include "producer_avformat.h"
23
24 // MLT Header files
25 #include <framework/mlt_frame.h>
26
27 // ffmpeg Header files
28 #include <avformat.h>
29 #ifdef SWSCALE
30 #include <swscale.h>
31 #endif
32
33 // System header files
34 #include <stdlib.h>
35 #include <string.h>
36 #include <pthread.h>
37 #include <math.h>
38
39 void avformat_lock( );
40 void avformat_unlock( );
41
42 // Forward references.
43 static int producer_open( mlt_producer this, char *file );
44 static int producer_get_frame( mlt_producer this, mlt_frame_ptr frame, int index );
45
46 /** Constructor for libavformat.
47 */
48
49 mlt_producer producer_avformat_init( char *file )
50 {
51 mlt_producer this = NULL;
52
53 // Check that we have a non-NULL argument
54 if ( file != NULL )
55 {
56 // Construct the producer
57 this = calloc( 1, sizeof( struct mlt_producer_s ) );
58
59 // Initialise it
60 if ( mlt_producer_init( this, NULL ) == 0 )
61 {
62 // Get the properties
63 mlt_properties properties = MLT_PRODUCER_PROPERTIES( this );
64
65 // Set the resource property (required for all producers)
66 mlt_properties_set( properties, "resource", file );
67
68 // Register our get_frame implementation
69 this->get_frame = producer_get_frame;
70
71 // Open the file
72 if ( producer_open( this, file ) != 0 )
73 {
74 // Clean up
75 mlt_producer_close( this );
76 this = NULL;
77 }
78 }
79 }
80
81 return this;
82 }
83
84 /** Find the default streams.
85 */
86
87 static void find_default_streams( AVFormatContext *context, int *audio_index, int *video_index )
88 {
89 int i;
90
91 // Allow for multiple audio and video streams in the file and select first of each (if available)
92 for( i = 0; i < context->nb_streams; i++ )
93 {
94 // Get the codec context
95 AVCodecContext *codec_context = context->streams[ i ]->codec;
96
97 if ( avcodec_find_decoder( codec_context->codec_id ) == NULL )
98 continue;
99
100 // Determine the type and obtain the first index of each type
101 switch( codec_context->codec_type )
102 {
103 case CODEC_TYPE_VIDEO:
104 if ( *video_index < 0 )
105 *video_index = i;
106 break;
107 case CODEC_TYPE_AUDIO:
108 if ( *audio_index < 0 )
109 *audio_index = i;
110 break;
111 default:
112 break;
113 }
114 }
115 }
116
117 /** Producer file destructor.
118 */
119
120 static void producer_file_close( void *context )
121 {
122 if ( context != NULL )
123 {
124 // Lock the mutex now
125 avformat_lock( );
126
127 // Close the file
128 av_close_input_file( context );
129
130 // Unlock the mutex now
131 avformat_unlock( );
132 }
133 }
134
135 /** Producer file destructor.
136 */
137
138 static void producer_codec_close( void *codec )
139 {
140 if ( codec != NULL )
141 {
142 // Lock the mutex now
143 avformat_lock( );
144
145 // Close the file
146 avcodec_close( codec );
147
148 // Unlock the mutex now
149 avformat_unlock( );
150 }
151 }
152
153 /** Open the file.
154 */
155
156 static int producer_open( mlt_producer this, char *file )
157 {
158 // Return an error code (0 == no error)
159 int error = 0;
160
161 // Context for avformat
162 AVFormatContext *context = NULL;
163
164 // Get the properties
165 mlt_properties properties = MLT_PRODUCER_PROPERTIES( this );
166
167 // We will treat everything with the producer fps
168 double fps = mlt_properties_get_double( properties, "fps" );
169
170 // Lock the mutex now
171 avformat_lock( );
172
173 // If "MRL", then create AVInputFormat
174 AVInputFormat *format = NULL;
175 AVFormatParameters *params = NULL;
176 char *standard = NULL;
177 char *mrl = strchr( file, ':' );
178
179 // AV option (0 = both, 1 = video, 2 = audio)
180 int av = 0;
181
182 // Setting lowest log level
183 av_log_set_level( -1 );
184
185 // Only if there is not a protocol specification that avformat can handle
186 if ( mrl && !url_exist( file ) )
187 {
188 // 'file' becomes format abbreviation
189 mrl[0] = 0;
190
191 // Lookup the format
192 format = av_find_input_format( file );
193
194 // Eat the format designator
195 file = ++mrl;
196
197 if ( format )
198 {
199 // Allocate params
200 params = calloc( sizeof( AVFormatParameters ), 1 );
201
202 // These are required by video4linux (defaults)
203 params->width = 640;
204 params->height = 480;
205 params->time_base= (AVRational){1,25};
206 params->device = file;
207 params->channels = 2;
208 params->sample_rate = 48000;
209 }
210
211 // Parse out params
212 mrl = strchr( file, '?' );
213 while ( mrl )
214 {
215 mrl[0] = 0;
216 char *name = strdup( ++mrl );
217 char *value = strchr( name, ':' );
218 if ( value )
219 {
220 value[0] = 0;
221 value++;
222 char *t = strchr( value, '&' );
223 if ( t )
224 t[0] = 0;
225 if ( !strcmp( name, "frame_rate" ) )
226 params->time_base.den = atoi( value );
227 else if ( !strcmp( name, "frame_rate_base" ) )
228 params->time_base.num = atoi( value );
229 else if ( !strcmp( name, "sample_rate" ) )
230 params->sample_rate = atoi( value );
231 else if ( !strcmp( name, "channels" ) )
232 params->channels = atoi( value );
233 else if ( !strcmp( name, "width" ) )
234 params->width = atoi( value );
235 else if ( !strcmp( name, "height" ) )
236 params->height = atoi( value );
237 else if ( !strcmp( name, "standard" ) )
238 {
239 standard = strdup( value );
240 params->standard = standard;
241 }
242 else if ( !strcmp( name, "av" ) )
243 av = atoi( value );
244 }
245 free( name );
246 mrl = strchr( mrl, '&' );
247 }
248 }
249
250 // Now attempt to open the file
251 error = av_open_input_file( &context, file, format, 0, params ) < 0;
252
253 // Cleanup AVFormatParameters
254 free( standard );
255 free( params );
256
257 // If successful, then try to get additional info
258 if ( error == 0 )
259 {
260 // Get the stream info
261 error = av_find_stream_info( context ) < 0;
262
263 // Continue if no error
264 if ( error == 0 )
265 {
266 // We will default to the first audio and video streams found
267 int audio_index = -1;
268 int video_index = -1;
269 int av_bypass = 0;
270
271 // Now set properties where we can (use default unknowns if required)
272 if ( context->duration != AV_NOPTS_VALUE )
273 {
274 // This isn't going to be accurate for all formats
275 mlt_position frames = ( mlt_position )( ( ( double )context->duration / ( double )AV_TIME_BASE ) * fps );
276 mlt_properties_set_position( properties, "out", frames - 2 );
277 mlt_properties_set_position( properties, "length", frames - 1 );
278 }
279
280 // Find default audio and video streams
281 find_default_streams( context, &audio_index, &video_index );
282
283 if ( context->start_time != AV_NOPTS_VALUE )
284 mlt_properties_set_double( properties, "_start_time", context->start_time );
285
286 // Check if we're seekable (something funny about mpeg here :-/)
287 if ( strcmp( file, "pipe:" ) && strncmp( file, "http://", 6 ) )
288 mlt_properties_set_int( properties, "seekable", av_seek_frame( context, -1, mlt_properties_get_double( properties, "_start_time" ), AVSEEK_FLAG_BACKWARD ) >= 0 );
289 else
290 av_bypass = 1;
291
292 // Store selected audio and video indexes on properties
293 mlt_properties_set_int( properties, "audio_index", audio_index );
294 mlt_properties_set_int( properties, "video_index", video_index );
295 mlt_properties_set_int( properties, "_last_position", -1 );
296
297 // Fetch the width, height and aspect ratio
298 if ( video_index != -1 )
299 {
300 AVCodecContext *codec_context = context->streams[ video_index ]->codec;
301 mlt_properties_set_int( properties, "width", codec_context->width );
302 mlt_properties_set_int( properties, "height", codec_context->height );
303 mlt_properties_set_double( properties, "aspect_ratio", av_q2d( codec_context->sample_aspect_ratio ) );
304 }
305
306 // Read Metadata
307 if (context->title != NULL)
308 mlt_properties_set(properties, "meta.attr.title.markup", context->title );
309 if (context->author != NULL)
310 mlt_properties_set(properties, "meta.attr.author.markup", context->author );
311 if (context->copyright != NULL)
312 mlt_properties_set(properties, "meta.attr.copyright.markup", context->copyright );
313 if (context->comment != NULL)
314 mlt_properties_set(properties, "meta.attr.comment.markup", context->comment );
315 if (context->album != NULL)
316 mlt_properties_set(properties, "meta.attr.album.markup", context->album );
317 if (context->year != 0)
318 mlt_properties_set_int(properties, "meta.attr.year.markup", context->year );
319 if (context->track != 0)
320 mlt_properties_set_int(properties, "meta.attr.track.markup", context->track );
321
322 // We're going to cheat here - for a/v files, we will have two contexts (reasoning will be clear later)
323 if ( av == 0 && !av_bypass && audio_index != -1 && video_index != -1 )
324 {
325 // We'll use the open one as our video_context
326 mlt_properties_set_data( properties, "video_context", context, 0, producer_file_close, NULL );
327 av_seek_frame( context, -1, 0, AVSEEK_FLAG_BACKWARD );
328
329 // And open again for our audio context
330 av_open_input_file( &context, file, NULL, 0, NULL );
331 av_find_stream_info( context );
332
333 // Audio context
334 mlt_properties_set_data( properties, "audio_context", context, 0, producer_file_close, NULL );
335 }
336 else if ( av != 2 && video_index != -1 )
337 {
338 // We only have a video context
339 mlt_properties_set_data( properties, "video_context", context, 0, producer_file_close, NULL );
340 av_seek_frame( context, -1, 0, AVSEEK_FLAG_BACKWARD );
341 }
342 else if ( audio_index != -1 )
343 {
344 // We only have an audio context
345 mlt_properties_set_data( properties, "audio_context", context, 0, producer_file_close, NULL );
346 }
347 else
348 {
349 // Something has gone wrong
350 error = -1;
351 }
352
353 mlt_properties_set_int( properties, "av_bypass", av_bypass );
354 }
355 }
356
357 // Unlock the mutex now
358 avformat_unlock( );
359
360 return error;
361 }
362
363 /** Convert a frame position to a time code.
364 */
365
366 static double producer_time_of_frame( mlt_producer this, mlt_position position )
367 {
368 // Get the properties
369 mlt_properties properties = MLT_PRODUCER_PROPERTIES( this );
370
371 // Obtain the fps
372 double fps = mlt_properties_get_double( properties, "fps" );
373
374 // Do the calc
375 return ( double )position / fps;
376 }
377
378 static inline void convert_image( AVFrame *frame, uint8_t *buffer, int pix_fmt, mlt_image_format format, int width, int height )
379 {
380 #ifdef SWSCALE
381 if ( format == mlt_image_yuv420p )
382 {
383 struct SwsContext *context = sws_getContext( width, height, pix_fmt,
384 width, height, PIX_FMT_YUV420P, SWS_FAST_BILINEAR, NULL, NULL, NULL);
385 AVPicture output;
386 output.data[0] = buffer;
387 output.data[1] = buffer + width * height;
388 output.data[2] = buffer + ( 3 * width * height ) / 2;
389 output.linesize[0] = width;
390 output.linesize[1] = width >> 1;
391 output.linesize[2] = width >> 1;
392 sws_scale( context, frame->data, frame->linesize, 0, height,
393 output.data, output.linesize);
394 sws_freeContext( context );
395 }
396 else if ( format == mlt_image_rgb24 )
397 {
398 struct SwsContext *context = sws_getContext( width, height, pix_fmt,
399 width, height, PIX_FMT_RGB24, SWS_FAST_BILINEAR, NULL, NULL, NULL);
400 AVPicture output;
401 avpicture_fill( &output, buffer, PIX_FMT_RGB24, width, height );
402 sws_scale( context, frame->data, frame->linesize, 0, height,
403 output.data, output.linesize);
404 sws_freeContext( context );
405 }
406 else
407 {
408 struct SwsContext *context = sws_getContext( width, height, pix_fmt,
409 width, height, PIX_FMT_YUYV422, SWS_FAST_BILINEAR, NULL, NULL, NULL);
410 AVPicture output;
411 avpicture_fill( &output, buffer, PIX_FMT_YUYV422, width, height );
412 sws_scale( context, frame->data, frame->linesize, 0, height,
413 output.data, output.linesize);
414 sws_freeContext( context );
415 }
416 #else
417 if ( format == mlt_image_yuv420p )
418 {
419 AVPicture pict;
420 pict.data[0] = buffer;
421 pict.data[1] = buffer + width * height;
422 pict.data[2] = buffer + ( 3 * width * height ) / 2;
423 pict.linesize[0] = width;
424 pict.linesize[1] = width >> 1;
425 pict.linesize[2] = width >> 1;
426 img_convert( &pict, PIX_FMT_YUV420P, (AVPicture *)frame, pix_fmt, width, height );
427 }
428 else if ( format == mlt_image_rgb24 )
429 {
430 AVPicture output;
431 avpicture_fill( &output, buffer, PIX_FMT_RGB24, width, height );
432 img_convert( &output, PIX_FMT_RGB24, (AVPicture *)frame, pix_fmt, width, height );
433 }
434 else
435 {
436 AVPicture output;
437 avpicture_fill( &output, buffer, PIX_FMT_YUV422, width, height );
438 img_convert( &output, PIX_FMT_YUV422, (AVPicture *)frame, pix_fmt, width, height );
439 }
440 #endif
441 }
442
443 /** Get an image from a frame.
444 */
445
446 static int producer_get_image( mlt_frame frame, uint8_t **buffer, mlt_image_format *format, int *width, int *height, int writable )
447 {
448 // Get the properties from the frame
449 mlt_properties frame_properties = MLT_FRAME_PROPERTIES( frame );
450
451 // Obtain the frame number of this frame
452 mlt_position position = mlt_properties_get_position( frame_properties, "avformat_position" );
453
454 // Get the producer
455 mlt_producer this = mlt_properties_get_data( frame_properties, "avformat_producer", NULL );
456
457 // Get the producer properties
458 mlt_properties properties = MLT_PRODUCER_PROPERTIES( this );
459
460 // Fetch the video_context
461 AVFormatContext *context = mlt_properties_get_data( properties, "video_context", NULL );
462
463 // Get the video_index
464 int index = mlt_properties_get_int( properties, "video_index" );
465
466 // Obtain the expected frame numer
467 mlt_position expected = mlt_properties_get_position( properties, "_video_expected" );
468
469 // Calculate the real time code
470 double real_timecode = producer_time_of_frame( this, position );
471
472 // Get the video stream
473 AVStream *stream = context->streams[ index ];
474
475 // Get codec context
476 AVCodecContext *codec_context = stream->codec;
477
478 // Packet
479 AVPacket pkt;
480
481 // Get the conversion frame
482 AVFrame *av_frame = mlt_properties_get_data( properties, "av_frame", NULL );
483
484 // Special case pause handling flag
485 int paused = 0;
486
487 // Special case ffwd handling
488 int ignore = 0;
489
490 // Current time calcs
491 int current_position = mlt_properties_get_double( properties, "_current_position" );
492
493 // We may want to use the source fps if available
494 double source_fps = mlt_properties_get_double( properties, "source_fps" );
495 double fps = mlt_properties_get_double( properties, "fps" );
496
497 // This is the physical frame position in the source
498 int req_position = ( int )( position / fps * source_fps );
499
500 // Get the seekable status
501 int seekable = mlt_properties_get_int( properties, "seekable" );
502
503 // Generate the size in bytes
504 int size = 0;
505
506 // Hopefully provide better support for streams...
507 int av_bypass = mlt_properties_get_int( properties, "av_bypass" );
508
509 // Determines if we have to decode all frames in a sequence
510 int must_decode = 1;
511
512 // Set the result arguments that we know here (only *buffer is now required)
513 *width = codec_context->width;
514 *height = codec_context->height;
515
516 switch ( *format )
517 {
518 case mlt_image_yuv420p:
519 size = *width * 3 * ( *height + 1 ) / 2;
520 break;
521 case mlt_image_rgb24:
522 size = *width * ( *height + 1 ) * 3;
523 break;
524 default:
525 *format = mlt_image_yuv422;
526 size = *width * ( *height + 1 ) * 2;
527 break;
528 }
529
530 // Set this on the frame properties
531 mlt_properties_set_int( frame_properties, "width", *width );
532 mlt_properties_set_int( frame_properties, "height", *height );
533
534 // Construct the output image
535 *buffer = mlt_pool_alloc( size );
536
537 // Temporary hack to improve intra frame only
538 must_decode = strcmp( codec_context->codec->name, "mjpeg" ) &&
539 strcmp( codec_context->codec->name, "rawvideo" ) &&
540 strcmp( codec_context->codec->name, "dvvideo" );
541
542 // Seek if necessary
543 if ( position != expected )
544 {
545 if ( av_frame != NULL && position + 1 == expected )
546 {
547 // We're paused - use last image
548 paused = 1;
549 }
550 else if ( !seekable && position > expected && ( position - expected ) < 250 )
551 {
552 // Fast forward - seeking is inefficient for small distances - just ignore following frames
553 ignore = ( int )( ( position - expected ) / fps * source_fps );
554 }
555 else if ( seekable && ( position < expected || position - expected >= 12 ) )
556 {
557 // Calculate the timestamp for the requested frame
558 int64_t timestamp = ( int64_t )( ( double )req_position / source_fps * AV_TIME_BASE );
559 if ( ( uint64_t )context->start_time != AV_NOPTS_VALUE )
560 timestamp += context->start_time;
561 if ( must_decode )
562 timestamp -= AV_TIME_BASE;
563 if ( timestamp < 0 )
564 timestamp = 0;
565
566 // Set to the timestamp
567 av_seek_frame( context, -1, timestamp, AVSEEK_FLAG_BACKWARD );
568
569 // Remove the cached info relating to the previous position
570 mlt_properties_set_int( properties, "_current_position", -1 );
571 mlt_properties_set_int( properties, "_last_position", -1 );
572 mlt_properties_set_data( properties, "av_frame", NULL, 0, NULL, NULL );
573 av_frame = NULL;
574 }
575 }
576
577 // Duplicate the last image if necessary (see comment on rawvideo below)
578 if ( av_frame != NULL && ( paused || mlt_properties_get_int( properties, "_current_position" ) >= req_position ) && av_bypass == 0 )
579 {
580 // Duplicate it
581 convert_image( av_frame, *buffer, codec_context->pix_fmt, *format, *width, *height );
582
583 // Set this on the frame properties
584 mlt_properties_set_data( frame_properties, "image", *buffer, size, ( mlt_destructor )mlt_pool_release, NULL );
585 }
586 else
587 {
588 int ret = 0;
589 int got_picture = 0;
590 int int_position = 0;
591
592 av_init_packet( &pkt );
593
594 // Construct an AVFrame for YUV422 conversion
595 if ( av_frame == NULL )
596 {
597 av_frame = avcodec_alloc_frame( );
598 mlt_properties_set_data( properties, "av_frame", av_frame, 0, av_free, NULL );
599 }
600
601 while( ret >= 0 && !got_picture )
602 {
603 // Read a packet
604 ret = av_read_frame( context, &pkt );
605
606 // We only deal with video from the selected video_index
607 if ( ret >= 0 && pkt.stream_index == index && pkt.size > 0 )
608 {
609 // Determine time code of the packet
610 int_position = ( int )( av_q2d( stream->time_base ) * pkt.dts * source_fps );
611 if ( context->start_time != AV_NOPTS_VALUE )
612 int_position -= ( int )( context->start_time * source_fps / AV_TIME_BASE );
613
614 int last_position = mlt_properties_get_int( properties, "_last_position" );
615 if ( int_position == last_position )
616 int_position = last_position + 1;
617 mlt_properties_set_int( properties, "_last_position", int_position );
618
619 // Decode the image
620 if ( must_decode || int_position >= req_position )
621 ret = avcodec_decode_video( codec_context, av_frame, &got_picture, pkt.data, pkt.size );
622
623 if ( got_picture )
624 {
625 // Handle ignore
626 if ( int_position < req_position )
627 {
628 ignore = 0;
629 got_picture = 0;
630 }
631 else if ( int_position >= req_position )
632 {
633 ignore = 0;
634 }
635 else if ( ignore -- )
636 {
637 got_picture = 0;
638 }
639 }
640 }
641
642 // Now handle the picture if we have one
643 if ( got_picture )
644 {
645 mlt_properties_set_int( frame_properties, "progressive", !av_frame->interlaced_frame );
646 mlt_properties_set_int( frame_properties, "top_field_first", av_frame->top_field_first );
647 convert_image( av_frame, *buffer, codec_context->pix_fmt, *format, *width, *height );
648 mlt_properties_set_data( frame_properties, "image", *buffer, size, (mlt_destructor)mlt_pool_release, NULL );
649 mlt_properties_set_double( properties, "_current_position", int_position );
650 }
651
652 // We're finished with this packet regardless
653 av_free_packet( &pkt );
654 }
655 }
656
657 // Very untidy - for rawvideo, the packet contains the frame, hence the free packet
658 // above will break the pause behaviour - so we wipe the frame now
659 if ( !strcmp( codec_context->codec->name, "rawvideo" ) )
660 mlt_properties_set_data( properties, "av_frame", NULL, 0, NULL, NULL );
661
662 // Set the field order property for this frame
663 mlt_properties_set_int( frame_properties, "top_field_first", mlt_properties_get_int( properties, "top_field_first" ) );
664
665 // Regardless of speed, we expect to get the next frame (cos we ain't too bright)
666 mlt_properties_set_position( properties, "_video_expected", position + 1 );
667
668 return 0;
669 }
670
671 /** Set up video handling.
672 */
673
674 static void producer_set_up_video( mlt_producer this, mlt_frame frame )
675 {
676 // Get the properties
677 mlt_properties properties = MLT_PRODUCER_PROPERTIES( this );
678
679 // Fetch the video_context
680 AVFormatContext *context = mlt_properties_get_data( properties, "video_context", NULL );
681
682 // Get the video_index
683 int index = mlt_properties_get_int( properties, "video_index" );
684
685 // Get the frame properties
686 mlt_properties frame_properties = MLT_FRAME_PROPERTIES( frame );
687
688 if ( context != NULL && index != -1 )
689 {
690 // Get the video stream
691 AVStream *stream = context->streams[ index ];
692
693 // Get codec context
694 AVCodecContext *codec_context = stream->codec;
695
696 // Get the codec
697 AVCodec *codec = mlt_properties_get_data( properties, "video_codec", NULL );
698
699 // Initialise the codec if necessary
700 if ( codec == NULL )
701 {
702 // Find the codec
703 codec = avcodec_find_decoder( codec_context->codec_id );
704
705 // If we don't have a codec and we can't initialise it, we can't do much more...
706 if ( codec != NULL && avcodec_open( codec_context, codec ) >= 0 )
707 {
708 // Now store the codec with its destructor
709 mlt_properties_set_data( properties, "video_codec", codec_context, 0, producer_codec_close, NULL );
710 }
711 else
712 {
713 // Remember that we can't use this later
714 mlt_properties_set_int( properties, "video_index", -1 );
715 }
716 }
717
718 // No codec, no show...
719 if ( codec != NULL )
720 {
721 double source_fps = 0;
722 int norm_aspect_ratio = mlt_properties_get_int( properties, "norm_aspect_ratio" );
723 double force_aspect_ratio = mlt_properties_get_double( properties, "force_aspect_ratio" );
724 double aspect_ratio;
725
726 // XXX: We won't know the real aspect ratio until an image is decoded
727 // but we do need it now (to satisfy filter_resize) - take a guess based
728 // on pal/ntsc
729 if ( force_aspect_ratio > 0.0 )
730 {
731 aspect_ratio = force_aspect_ratio;
732 }
733 else if ( !norm_aspect_ratio && codec_context->sample_aspect_ratio.num > 0 )
734 {
735 aspect_ratio = av_q2d( codec_context->sample_aspect_ratio );
736 }
737 else
738 {
739 int is_pal = mlt_properties_get_double( properties, "fps" ) == 25.0;
740 aspect_ratio = is_pal ? 59.0/54.0 : 10.0/11.0;
741 }
742
743 // Determine the fps
744 source_fps = ( double )codec_context->time_base.den / ( codec_context->time_base.num == 0 ? 1 : codec_context->time_base.num );
745
746 // We'll use fps if it's available
747 if ( source_fps > 0 && source_fps < 30 )
748 mlt_properties_set_double( properties, "source_fps", source_fps );
749 else
750 mlt_properties_set_double( properties, "source_fps", mlt_properties_get_double( properties, "fps" ) );
751 mlt_properties_set_double( properties, "aspect_ratio", aspect_ratio );
752
753 // Set the width and height
754 mlt_properties_set_int( frame_properties, "width", codec_context->width );
755 mlt_properties_set_int( frame_properties, "height", codec_context->height );
756 mlt_properties_set_double( frame_properties, "aspect_ratio", aspect_ratio );
757
758 mlt_frame_push_get_image( frame, producer_get_image );
759 mlt_properties_set_data( frame_properties, "avformat_producer", this, 0, NULL, NULL );
760 }
761 else
762 {
763 mlt_properties_set_int( frame_properties, "test_image", 1 );
764 }
765 }
766 else
767 {
768 mlt_properties_set_int( frame_properties, "test_image", 1 );
769 }
770 }
771
772 /** Get the audio from a frame.
773 */
774
775 static int producer_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_format *format, int *frequency, int *channels, int *samples )
776 {
777 // Get the properties from the frame
778 mlt_properties frame_properties = MLT_FRAME_PROPERTIES( frame );
779
780 // Obtain the frame number of this frame
781 mlt_position position = mlt_properties_get_position( frame_properties, "avformat_position" );
782
783 // Get the producer
784 mlt_producer this = mlt_properties_get_data( frame_properties, "avformat_producer", NULL );
785
786 // Get the producer properties
787 mlt_properties properties = MLT_PRODUCER_PROPERTIES( this );
788
789 // Fetch the audio_context
790 AVFormatContext *context = mlt_properties_get_data( properties, "audio_context", NULL );
791
792 // Get the audio_index
793 int index = mlt_properties_get_int( properties, "audio_index" );
794
795 // Get the seekable status
796 int seekable = mlt_properties_get_int( properties, "seekable" );
797
798 // Obtain the expected frame numer
799 mlt_position expected = mlt_properties_get_position( properties, "_audio_expected" );
800
801 // Obtain the resample context if it exists (not always needed)
802 ReSampleContext *resample = mlt_properties_get_data( properties, "audio_resample", NULL );
803
804 // Obtain the audio buffer
805 int16_t *audio_buffer = mlt_properties_get_data( properties, "audio_buffer", NULL );
806
807 // Get amount of audio used
808 int audio_used = mlt_properties_get_int( properties, "_audio_used" );
809
810 // Calculate the real time code
811 double real_timecode = producer_time_of_frame( this, position );
812
813 // Get the audio stream
814 AVStream *stream = context->streams[ index ];
815
816 // Get codec context
817 AVCodecContext *codec_context = stream->codec;
818
819 // Packet
820 AVPacket pkt;
821
822 // Number of frames to ignore (for ffwd)
823 int ignore = 0;
824
825 // Flag for paused (silence)
826 int paused = 0;
827
828 // Check for resample and create if necessary
829 if ( resample == NULL && codec_context->channels <= 2 )
830 {
831 // Create the resampler
832 resample = audio_resample_init( *channels, codec_context->channels, *frequency, codec_context->sample_rate );
833
834 // And store it on properties
835 mlt_properties_set_data( properties, "audio_resample", resample, 0, ( mlt_destructor )audio_resample_close, NULL );
836 }
837 else if ( resample == NULL )
838 {
839 *channels = codec_context->channels;
840 *frequency = codec_context->sample_rate;
841 }
842
843 // Check for audio buffer and create if necessary
844 if ( audio_buffer == NULL )
845 {
846 // Allocate the audio buffer
847 audio_buffer = mlt_pool_alloc( AVCODEC_MAX_AUDIO_FRAME_SIZE * sizeof( int16_t ) );
848
849 // And store it on properties for reuse
850 mlt_properties_set_data( properties, "audio_buffer", audio_buffer, 0, ( mlt_destructor )mlt_pool_release, NULL );
851 }
852
853 // Seek if necessary
854 if ( position != expected )
855 {
856 if ( position + 1 == expected )
857 {
858 // We're paused - silence required
859 paused = 1;
860 }
861 else if ( !seekable && position > expected && ( position - expected ) < 250 )
862 {
863 // Fast forward - seeking is inefficient for small distances - just ignore following frames
864 ignore = position - expected;
865 }
866 else if ( position < expected || position - expected >= 12 )
867 {
868 // Set to the real timecode
869 if ( av_seek_frame( context, -1, mlt_properties_get_double( properties, "_start_time" ) + real_timecode * 1000000.0, AVSEEK_FLAG_BACKWARD ) != 0 )
870 paused = 1;
871
872 // Clear the usage in the audio buffer
873 audio_used = 0;
874 }
875 }
876
877 // Get the audio if required
878 if ( !paused )
879 {
880 int ret = 0;
881 int got_audio = 0;
882 int16_t *temp = mlt_pool_alloc( sizeof( int16_t ) * AVCODEC_MAX_AUDIO_FRAME_SIZE );
883
884 av_init_packet( &pkt );
885
886 while( ret >= 0 && !got_audio )
887 {
888 // Check if the buffer already contains the samples required
889 if ( audio_used >= *samples && ignore == 0 )
890 {
891 got_audio = 1;
892 break;
893 }
894
895 // Read a packet
896 ret = av_read_frame( context, &pkt );
897
898 int len = pkt.size;
899 uint8_t *ptr = pkt.data;
900 int data_size;
901
902 // We only deal with audio from the selected audio_index
903 while ( ptr != NULL && ret >= 0 && pkt.stream_index == index && len > 0 )
904 {
905 // Decode the audio
906 ret = avcodec_decode_audio( codec_context, temp, &data_size, ptr, len );
907
908 if ( ret < 0 )
909 {
910 ret = 0;
911 break;
912 }
913
914 len -= ret;
915 ptr += ret;
916
917 if ( data_size > 0 )
918 {
919 if ( resample != NULL )
920 {
921 audio_used += audio_resample( resample, &audio_buffer[ audio_used * *channels ], temp, data_size / ( codec_context->channels * sizeof( int16_t ) ) );
922 }
923 else
924 {
925 memcpy( &audio_buffer[ audio_used * *channels ], temp, data_size );
926 audio_used += data_size / ( codec_context->channels * sizeof( int16_t ) );
927 }
928
929 // Handle ignore
930 while ( ignore && audio_used > *samples )
931 {
932 ignore --;
933 audio_used -= *samples;
934 memmove( audio_buffer, &audio_buffer[ *samples * *channels ], audio_used * sizeof( int16_t ) );
935 }
936 }
937
938 // If we're behind, ignore this packet
939 float current_pts = av_q2d( stream->time_base ) * pkt.pts;
940 if ( seekable && ( !ignore && current_pts <= ( real_timecode - 0.02 ) ) )
941 ignore = 1;
942 }
943
944 // We're finished with this packet regardless
945 av_free_packet( &pkt );
946 }
947
948 *buffer = mlt_pool_alloc( *samples * *channels * sizeof( int16_t ) );
949 mlt_properties_set_data( frame_properties, "audio", *buffer, 0, ( mlt_destructor )mlt_pool_release, NULL );
950
951 // Now handle the audio if we have enough
952 if ( audio_used >= *samples )
953 {
954 memcpy( *buffer, audio_buffer, *samples * *channels * sizeof( int16_t ) );
955 audio_used -= *samples;
956 memmove( audio_buffer, &audio_buffer[ *samples * *channels ], audio_used * *channels * sizeof( int16_t ) );
957 }
958 else
959 {
960 memset( *buffer, 0, *samples * *channels * sizeof( int16_t ) );
961 }
962
963 // Store the number of audio samples still available
964 mlt_properties_set_int( properties, "_audio_used", audio_used );
965
966 // Release the temporary audio
967 mlt_pool_release( temp );
968 }
969 else
970 {
971 // Get silence and don't touch the context
972 mlt_frame_get_audio( frame, buffer, format, frequency, channels, samples );
973 }
974
975 // Regardless of speed (other than paused), we expect to get the next frame
976 if ( !paused )
977 mlt_properties_set_position( properties, "_audio_expected", position + 1 );
978
979 return 0;
980 }
981
982 /** Set up audio handling.
983 */
984
985 static void producer_set_up_audio( mlt_producer this, mlt_frame frame )
986 {
987 // Get the properties
988 mlt_properties properties = MLT_PRODUCER_PROPERTIES( this );
989
990 // Fetch the audio_context
991 AVFormatContext *context = mlt_properties_get_data( properties, "audio_context", NULL );
992
993 // Get the audio_index
994 int index = mlt_properties_get_int( properties, "audio_index" );
995
996 // Deal with audio context
997 if ( context != NULL && index != -1 )
998 {
999 // Get the frame properties
1000 mlt_properties frame_properties = MLT_FRAME_PROPERTIES( frame );
1001
1002 // Get the audio stream
1003 AVStream *stream = context->streams[ index ];
1004
1005 // Get codec context
1006 AVCodecContext *codec_context = stream->codec;
1007
1008 // Get the codec
1009 AVCodec *codec = mlt_properties_get_data( properties, "audio_codec", NULL );
1010
1011 // Initialise the codec if necessary
1012 if ( codec == NULL )
1013 {
1014 // Find the codec
1015 codec = avcodec_find_decoder( codec_context->codec_id );
1016
1017 // If we don't have a codec and we can't initialise it, we can't do much more...
1018 if ( codec != NULL && avcodec_open( codec_context, codec ) >= 0 )
1019 {
1020 // Now store the codec with its destructor
1021 mlt_properties_set_data( properties, "audio_codec", codec_context, 0, producer_codec_close, NULL );
1022
1023 }
1024 else
1025 {
1026 // Remember that we can't use this later
1027 mlt_properties_set_int( properties, "audio_index", -1 );
1028 }
1029 }
1030
1031 // No codec, no show...
1032 if ( codec != NULL )
1033 {
1034 mlt_frame_push_audio( frame, producer_get_audio );
1035 mlt_properties_set_data( frame_properties, "avformat_producer", this, 0, NULL, NULL );
1036 mlt_properties_set_int( frame_properties, "frequency", codec_context->sample_rate );
1037 mlt_properties_set_int( frame_properties, "channels", codec_context->channels );
1038 }
1039 }
1040 }
1041
1042 /** Our get frame implementation.
1043 */
1044
1045 static int producer_get_frame( mlt_producer this, mlt_frame_ptr frame, int index )
1046 {
1047 // Create an empty frame
1048 *frame = mlt_frame_init( );
1049
1050 // Update timecode on the frame we're creating
1051 mlt_frame_set_position( *frame, mlt_producer_position( this ) );
1052
1053 // Set the position of this producer
1054 mlt_properties_set_position( MLT_FRAME_PROPERTIES( *frame ), "avformat_position", mlt_producer_frame( this ) );
1055
1056 // Set up the video
1057 producer_set_up_video( this, *frame );
1058
1059 // Set up the audio
1060 producer_set_up_audio( this, *frame );
1061
1062 // Set the aspect_ratio
1063 mlt_properties_set_double( MLT_FRAME_PROPERTIES( *frame ), "aspect_ratio", mlt_properties_get_double( MLT_PRODUCER_PROPERTIES( this ), "aspect_ratio" ) );
1064
1065 // Calculate the next timecode
1066 mlt_producer_prepare_next( this );
1067
1068 return 0;
1069 }