2 Copyright (C) 2012 Carl Hetherington <cth@carlh.net>
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 /** @file src/ffmpeg_decoder.cc
21 * @brief A decoder using FFmpeg to decode content.
30 #include <boost/lexical_cast.hpp>
33 #include <libavcodec/avcodec.h>
34 #include <libavformat/avformat.h>
35 #include <libswscale/swscale.h>
36 #include <libpostproc/postprocess.h>
41 #include "transcoder.h"
45 #include "exceptions.h"
49 #include "ffmpeg_decoder.h"
55 using std::stringstream;
56 using boost::shared_ptr;
58 FFmpegDecoder::FFmpegDecoder (shared_ptr<Film> f, shared_ptr<const Options> o, Job* j, bool minimal, bool ignore_length)
59 : Decoder (f, o, j, minimal, ignore_length)
63 , _subtitle_stream (-1)
65 , _video_codec_context (0)
67 , _audio_codec_context (0)
69 , _subtitle_codec_context (0)
78 FFmpegDecoder::~FFmpegDecoder ()
80 if (_audio_codec_context) {
81 avcodec_close (_audio_codec_context);
84 if (_video_codec_context) {
85 avcodec_close (_video_codec_context);
88 if (_subtitle_codec_context) {
89 avcodec_close (_subtitle_codec_context);
93 avformat_close_input (&_format_context);
97 FFmpegDecoder::setup_general ()
103 if ((r = avformat_open_input (&_format_context, _film->content_path().c_str(), 0, 0)) != 0) {
104 throw OpenFileError (_film->content_path ());
107 if (avformat_find_stream_info (_format_context, 0) < 0) {
108 throw DecodeError ("could not find stream information");
111 /* Find video, audio and subtitle streams and choose the first of each */
113 for (uint32_t i = 0; i < _format_context->nb_streams; ++i) {
114 AVStream* s = _format_context->streams[i];
115 if (s->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
117 } else if (s->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
118 if (_audio_stream == -1) {
121 _audio_streams.push_back (AudioStream (stream_name (s), i, s->codec->channels));
122 } else if (s->codec->codec_type == AVMEDIA_TYPE_SUBTITLE) {
123 if (_subtitle_stream == -1) {
124 _subtitle_stream = i;
126 _subtitle_streams.push_back (SubtitleStream (stream_name (s), i));
130 /* Now override audio and subtitle streams with those from the Film, if it has any */
132 if (_film->audio_stream_index() != -1) {
133 _audio_stream = _film->audio_stream().id();
136 if (_film->subtitle_stream_index() != -1) {
137 _subtitle_stream = _film->subtitle_stream().id ();
140 if (_video_stream < 0) {
141 throw DecodeError ("could not find video stream");
144 _frame = avcodec_alloc_frame ();
146 throw DecodeError ("could not allocate frame");
151 FFmpegDecoder::setup_video ()
153 _video_codec_context = _format_context->streams[_video_stream]->codec;
154 _video_codec = avcodec_find_decoder (_video_codec_context->codec_id);
156 if (_video_codec == 0) {
157 throw DecodeError ("could not find video decoder");
160 /* I think this prevents problems with green hash on decodes and
161 "changing frame properties on the fly is not supported by all filters"
162 messages with some content. Although I'm not sure; needs checking.
164 AVDictionary* opts = 0;
165 av_dict_set (&opts, "threads", "1", 0);
167 if (avcodec_open2 (_video_codec_context, _video_codec, &opts) < 0) {
168 throw DecodeError ("could not open video decoder");
173 FFmpegDecoder::setup_audio ()
175 if (_audio_stream < 0) {
179 _audio_codec_context = _format_context->streams[_audio_stream]->codec;
180 _audio_codec = avcodec_find_decoder (_audio_codec_context->codec_id);
182 if (_audio_codec == 0) {
183 throw DecodeError ("could not find audio decoder");
186 if (avcodec_open2 (_audio_codec_context, _audio_codec, 0) < 0) {
187 throw DecodeError ("could not open audio decoder");
190 /* This is a hack; sometimes it seems that _audio_codec_context->channel_layout isn't set up,
191 so bodge it here. No idea why we should have to do this.
194 if (_audio_codec_context->channel_layout == 0) {
195 _audio_codec_context->channel_layout = av_get_default_channel_layout (audio_channels ());
200 FFmpegDecoder::setup_subtitle ()
202 if (_subtitle_stream < 0) {
206 _subtitle_codec_context = _format_context->streams[_subtitle_stream]->codec;
207 _subtitle_codec = avcodec_find_decoder (_subtitle_codec_context->codec_id);
209 if (_subtitle_codec == 0) {
210 throw DecodeError ("could not find subtitle decoder");
213 if (avcodec_open2 (_subtitle_codec_context, _subtitle_codec, 0) < 0) {
214 throw DecodeError ("could not open subtitle decoder");
220 FFmpegDecoder::do_pass ()
222 int r = av_read_frame (_format_context, &_packet);
225 if (r != AVERROR_EOF) {
226 throw DecodeError ("error on av_read_frame");
229 /* Get any remaining frames */
234 /* XXX: should we reset _packet.data and size after each *_decode_* call? */
238 while (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
239 process_video (_frame);
242 if (_audio_stream >= 0 && _opt->decode_audio) {
243 while (avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
244 int const data_size = av_samples_get_buffer_size (
245 0, _audio_codec_context->channels, _frame->nb_samples, audio_sample_format (), 1
248 assert (_audio_codec_context->channels == _film->audio_channels());
249 process_audio (_frame->data[0], data_size);
256 double const pts_seconds = av_q2d (_format_context->streams[_packet.stream_index]->time_base) * _packet.pts;
258 avcodec_get_frame_defaults (_frame);
260 if (_packet.stream_index == _video_stream) {
263 _first_video = pts_seconds;
267 if (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
269 /* Where we are in the output, in seconds */
270 double const out_pts_seconds = last_video_frame() / frames_per_second();
272 /* Difference between where we are and where we should be */
273 double const delta = pts_seconds - out_pts_seconds;
274 double const one_frame = 1 / frames_per_second();
276 /* Insert frames if required to get out_pts_seconds up to pts_seconds */
277 if (delta > one_frame) {
278 int const extra = rint (delta / one_frame);
279 for (int i = 0; i < extra; ++i) {
280 _film->log()->log (String::compose ("Extra frame inserted at %1s", out_pts_seconds));
281 process_video (_frame);
285 if (delta > -one_frame) {
286 /* Process this frame */
287 process_video (_frame);
289 /* Otherwise we are omitting a frame to keep things right */
290 _film->log()->log (String::compose ("Frame removed at %1s", out_pts_seconds));
294 } else if (_audio_stream >= 0 && _packet.stream_index == _audio_stream && _opt->decode_audio && _first_video && _first_video.get() <= pts_seconds) {
296 /* Note: We only decode audio if we've had our first video packet through, and if it
297 was before this packet. Until then audio is thrown away.
301 _first_audio = pts_seconds;
303 /* This is our first audio packet, and if we've arrived here we must have had our
304 first video packet. Push some silence to make up the gap between our first
305 video packet and our first audio.
308 /* frames of silence that we must push */
309 int const s = rint ((_first_audio.get() - _first_video.get()) * audio_sample_rate ());
313 "First video at %1, first audio at %2, pushing %3 frames of silence for %4 channels (%5 bytes per sample)",
314 _first_video.get(), _first_audio.get(), s, audio_channels(), bytes_per_audio_sample()
319 int const b = s * audio_channels() * bytes_per_audio_sample();
321 /* XXX: this assumes that it won't be too much, and there are shaky assumptions
322 that all sound representations are silent with memset()ed zero data.
325 memset (silence, 0, b);
326 process_audio (silence, b);
330 if (avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
331 int const data_size = av_samples_get_buffer_size (
332 0, _audio_codec_context->channels, _frame->nb_samples, audio_sample_format (), 1
335 assert (_audio_codec_context->channels == _film->audio_channels());
336 process_audio (_frame->data[0], data_size);
339 } else if (_subtitle_stream >= 0 && _packet.stream_index == _subtitle_stream && _opt->decode_subtitles && _first_video) {
343 if (avcodec_decode_subtitle2 (_subtitle_codec_context, &sub, &got_subtitle, &_packet) && got_subtitle) {
344 /* Sometimes we get an empty AVSubtitle, which is used by some codecs to
345 indicate that the previous subtitle should stop.
347 if (sub.num_rects > 0) {
348 process_subtitle (shared_ptr<TimedSubtitle> (new TimedSubtitle (sub, _first_video.get())));
350 process_subtitle (shared_ptr<TimedSubtitle> ());
352 avsubtitle_free (&sub);
356 av_free_packet (&_packet);
361 FFmpegDecoder::frames_per_second () const
363 AVStream* s = _format_context->streams[_video_stream];
365 if (s->avg_frame_rate.num && s->avg_frame_rate.den) {
366 return av_q2d (s->avg_frame_rate);
369 return av_q2d (s->r_frame_rate);
373 FFmpegDecoder::audio_channels () const
375 if (_audio_codec_context == 0) {
379 return _audio_codec_context->channels;
383 FFmpegDecoder::audio_sample_rate () const
385 if (_audio_codec_context == 0) {
389 return _audio_codec_context->sample_rate;
393 FFmpegDecoder::audio_sample_format () const
395 if (_audio_codec_context == 0) {
396 return (AVSampleFormat) 0;
399 return _audio_codec_context->sample_fmt;
403 FFmpegDecoder::audio_channel_layout () const
405 if (_audio_codec_context == 0) {
409 return _audio_codec_context->channel_layout;
413 FFmpegDecoder::native_size () const
415 return Size (_video_codec_context->width, _video_codec_context->height);
419 FFmpegDecoder::pixel_format () const
421 return _video_codec_context->pix_fmt;
425 FFmpegDecoder::time_base_numerator () const
427 return _video_codec_context->time_base.num;
431 FFmpegDecoder::time_base_denominator () const
433 return _video_codec_context->time_base.den;
437 FFmpegDecoder::sample_aspect_ratio_numerator () const
439 return _video_codec_context->sample_aspect_ratio.num;
443 FFmpegDecoder::sample_aspect_ratio_denominator () const
445 return _video_codec_context->sample_aspect_ratio.den;
449 FFmpegDecoder::has_subtitles () const
451 return (_subtitle_stream != -1);
455 FFmpegDecoder::audio_streams () const
457 return _audio_streams;
460 vector<SubtitleStream>
461 FFmpegDecoder::subtitle_streams () const
463 return _subtitle_streams;
467 FFmpegDecoder::stream_name (AVStream* s) const
471 AVDictionaryEntry const * lang = av_dict_get (s->metadata, "language", 0, 0);
476 AVDictionaryEntry const * title = av_dict_get (s->metadata, "title", 0, 0);
478 if (!n.str().empty()) {
484 if (n.str().empty()) {