2 Copyright (C) 2013-2015 Carl Hetherington <cth@carlh.net>
4 This file is part of DCP-o-matic.
6 DCP-o-matic 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.
11 DCP-o-matic 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.
16 You should have received a copy of the GNU General Public License
17 along with DCP-o-matic. If not, see <http://www.gnu.org/licenses/>.
21 #include "dcpomatic_log.h"
22 #include "ffmpeg_examiner.h"
23 #include "ffmpeg_content.h"
25 #include "ffmpeg_audio_stream.h"
26 #include "ffmpeg_subtitle_stream.h"
29 DCPOMATIC_DISABLE_WARNINGS
31 #include <libavcodec/avcodec.h>
32 #include <libavformat/avformat.h>
33 #include <libavutil/pixfmt.h>
34 #include <libavutil/pixdesc.h>
35 #include <libavutil/eval.h>
36 #include <libavutil/display.h>
38 DCPOMATIC_ENABLE_WARNINGS
47 using std::shared_ptr;
48 using boost::optional;
49 using namespace dcpomatic;
52 /* This is how many frames from the start of any video that we will examine to see if we
53 * can spot soft 2:3 pull-down ("telecine").
55 static const int PULLDOWN_CHECK_FRAMES = 16;
58 /** @param job job that the examiner is operating in, or 0 */
59 FFmpegExaminer::FFmpegExaminer (shared_ptr<const FFmpegContent> c, shared_ptr<Job> job)
62 , _need_video_length (false)
65 /* Find audio and subtitle streams */
67 for (uint32_t i = 0; i < _format_context->nb_streams; ++i) {
68 AVStream* s = _format_context->streams[i];
69 DCPOMATIC_DISABLE_WARNINGS
70 if (s->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
72 /* This is a hack; sometimes it seems that _audio_codec_context->channel_layout isn't set up,
73 so bodge it here. No idea why we should have to do this.
76 if (s->codec->channel_layout == 0) {
77 s->codec->channel_layout = av_get_default_channel_layout (s->codec->channels);
80 DCPOMATIC_ASSERT (_format_context->duration != AV_NOPTS_VALUE);
81 DCPOMATIC_ASSERT (s->codec->codec);
82 DCPOMATIC_ASSERT (s->codec->codec->name);
84 _audio_streams.push_back (
85 shared_ptr<FFmpegAudioStream> (
86 new FFmpegAudioStream (
88 s->codec->codec->name,
90 s->codec->sample_rate,
91 llrint ((double (_format_context->duration) / AV_TIME_BASE) * s->codec->sample_rate),
97 } else if (s->codec->codec_type == AVMEDIA_TYPE_SUBTITLE) {
98 _subtitle_streams.push_back (shared_ptr<FFmpegSubtitleStream> (new FFmpegSubtitleStream (subtitle_stream_name (s), s->id)));
103 /* See if the header has duration information in it */
104 _need_video_length = _format_context->duration == AV_NOPTS_VALUE;
105 if (!_need_video_length) {
106 _video_length = llrint ((double (_format_context->duration) / AV_TIME_BASE) * video_frame_rate().get());
110 if (job && _need_video_length) {
111 job->sub (_("Finding length"));
114 /* Run through until we find:
116 * - the first audio for each stream.
117 * - the top-field-first and repeat-first-frame values ("temporal_reference") for the first PULLDOWN_CHECK_FRAMES video frames.
120 int64_t const len = _file_group.length ();
121 /* A string which we build up to describe the top-field-first and repeat-first-frame values for the first few frames.
122 * It would be nicer to use something like vector<bool> here but we want to search the array for a pattern later,
123 * and a string seems a reasonably neat way to do that.
125 string temporal_reference;
127 auto packet = av_packet_alloc ();
128 DCPOMATIC_ASSERT (packet);
129 int r = av_read_frame (_format_context, packet);
131 av_packet_free (&packet);
137 job->set_progress (float (_format_context->pb->pos) / len);
139 job->set_progress_unknown ();
143 auto context = _format_context->streams[packet->stream_index]->codec;
144 DCPOMATIC_ENABLE_WARNINGS
146 if (_video_stream && packet->stream_index == _video_stream.get()) {
147 video_packet (context, temporal_reference, packet);
150 bool got_all_audio = true;
152 for (size_t i = 0; i < _audio_streams.size(); ++i) {
153 if (_audio_streams[i]->uses_index(_format_context, packet->stream_index)) {
154 audio_packet (context, _audio_streams[i], packet);
156 if (!_audio_streams[i]->first_audio) {
157 got_all_audio = false;
161 av_packet_free (&packet);
163 if (_first_video && got_all_audio && temporal_reference.size() >= (PULLDOWN_CHECK_FRAMES * 2)) {
171 av_init_packet (&packet);
172 packet.data = nullptr;
174 DCPOMATIC_DISABLE_WARNINGS
175 auto context = _format_context->streams[*_video_stream]->codec;
176 DCPOMATIC_ENABLE_WARNINGS
177 while (video_packet(context, temporal_reference, &packet)) {}
180 for (auto i: _audio_streams) {
182 av_init_packet (&packet);
183 packet.data = nullptr;
185 DCPOMATIC_DISABLE_WARNINGS
186 audio_packet (i->stream(_format_context)->codec, i, &packet);
187 DCPOMATIC_ENABLE_WARNINGS
191 /* This code taken from get_rotation() in ffmpeg:cmdutils.c */
192 auto stream = _format_context->streams[*_video_stream];
193 auto rotate_tag = av_dict_get (stream->metadata, "rotate", 0, 0);
194 uint8_t* displaymatrix = av_stream_get_side_data (stream, AV_PKT_DATA_DISPLAYMATRIX, 0);
197 if (rotate_tag && *rotate_tag->value && strcmp(rotate_tag->value, "0")) {
199 _rotation = av_strtod (rotate_tag->value, &tail);
205 if (displaymatrix && !_rotation) {
206 _rotation = - av_display_rotation_get ((int32_t*) displaymatrix);
209 _rotation = *_rotation - 360 * floor (*_rotation / 360 + 0.9 / 360);
212 LOG_GENERAL("Temporal reference was %1", temporal_reference);
213 if (temporal_reference.find("T2T3B2B3T2T3B2B3") != string::npos || temporal_reference.find("B2B3T2T3B2B3T2T3") != string::npos) {
214 /* The magical sequence (taken from mediainfo) suggests that 2:3 pull-down is in use */
216 LOG_GENERAL_NC("Suggest that this may be 2:3 pull-down (soft telecine)");
221 /** @param temporal_reference A string to which we should add two characters per frame;
222 * the first is T or B depending on whether it's top- or bottom-field first,
223 * the second is 3 or 2 depending on whether "repeat_pict" is true or not.
224 * @return true if some video was decoded, otherwise false.
227 FFmpegExaminer::video_packet (AVCodecContext* context, string& temporal_reference, AVPacket* packet)
229 DCPOMATIC_ASSERT (_video_stream);
231 if (_first_video && !_need_video_length && temporal_reference.size() >= (PULLDOWN_CHECK_FRAMES * 2)) {
236 DCPOMATIC_DISABLE_WARNINGS
237 if (avcodec_decode_video2 (context, _frame, &frame_finished, packet) < 0 || !frame_finished) {
240 DCPOMATIC_ENABLE_WARNINGS
243 _first_video = frame_time (_format_context->streams[_video_stream.get()]);
245 if (_need_video_length) {
246 _video_length = frame_time (
247 _format_context->streams[_video_stream.get()]
248 ).get_value_or (ContentTime ()).frames_round (video_frame_rate().get ());
250 if (temporal_reference.size() < (PULLDOWN_CHECK_FRAMES * 2)) {
251 temporal_reference += (_frame->top_field_first ? "T" : "B");
252 temporal_reference += (_frame->repeat_pict ? "3" : "2");
260 FFmpegExaminer::audio_packet (AVCodecContext* context, shared_ptr<FFmpegAudioStream> stream, AVPacket* packet)
262 if (stream->first_audio) {
267 DCPOMATIC_DISABLE_WARNINGS
268 if (avcodec_decode_audio4 (context, _frame, &frame_finished, packet) >= 0 && frame_finished) {
269 DCPOMATIC_ENABLE_WARNINGS
270 stream->first_audio = frame_time (stream->stream (_format_context));
275 optional<ContentTime>
276 FFmpegExaminer::frame_time (AVStream* s) const
278 optional<ContentTime> t;
280 DCPOMATIC_DISABLE_WARNINGS
281 int64_t const bet = av_frame_get_best_effort_timestamp (_frame);
282 DCPOMATIC_ENABLE_WARNINGS
283 if (bet != AV_NOPTS_VALUE) {
284 t = ContentTime::from_seconds (bet * av_q2d (s->time_base));
292 FFmpegExaminer::video_frame_rate () const
294 DCPOMATIC_ASSERT (_video_stream);
295 return av_q2d(av_guess_frame_rate(_format_context, _format_context->streams[_video_stream.get()], 0));
299 FFmpegExaminer::video_size () const
301 return dcp::Size (video_codec_context()->width, video_codec_context()->height);
304 /** @return Length according to our content's header */
306 FFmpegExaminer::video_length () const
308 return max (Frame (1), _video_length);
312 FFmpegExaminer::sample_aspect_ratio () const
314 DCPOMATIC_ASSERT (_video_stream);
315 AVRational sar = av_guess_sample_aspect_ratio (_format_context, _format_context->streams[_video_stream.get()], 0);
317 /* I assume this means that we don't know */
320 return double (sar.num) / sar.den;
324 FFmpegExaminer::subtitle_stream_name (AVStream* s) const
326 auto n = stream_name (s);
336 FFmpegExaminer::stream_name (AVStream* s) const
341 AVDictionaryEntry const * lang = av_dict_get (s->metadata, "language", 0, 0);
346 AVDictionaryEntry const * title = av_dict_get (s->metadata, "title", 0, 0);
359 FFmpegExaminer::bits_per_pixel () const
361 if (video_codec_context()->pix_fmt == -1) {
362 return optional<int>();
365 AVPixFmtDescriptor const * d = av_pix_fmt_desc_get (video_codec_context()->pix_fmt);
366 DCPOMATIC_ASSERT (d);
367 return av_get_bits_per_pixel (d);
371 FFmpegExaminer::yuv () const
373 switch (video_codec_context()->pix_fmt) {
374 case AV_PIX_FMT_YUV420P:
375 case AV_PIX_FMT_YUYV422:
376 case AV_PIX_FMT_YUV422P:
377 case AV_PIX_FMT_YUV444P:
378 case AV_PIX_FMT_YUV410P:
379 case AV_PIX_FMT_YUV411P:
380 case AV_PIX_FMT_YUVJ420P:
381 case AV_PIX_FMT_YUVJ422P:
382 case AV_PIX_FMT_YUVJ444P:
383 case AV_PIX_FMT_UYVY422:
384 case AV_PIX_FMT_UYYVYY411:
385 case AV_PIX_FMT_NV12:
386 case AV_PIX_FMT_NV21:
387 case AV_PIX_FMT_YUV440P:
388 case AV_PIX_FMT_YUVJ440P:
389 case AV_PIX_FMT_YUVA420P:
390 case AV_PIX_FMT_YUV420P16LE:
391 case AV_PIX_FMT_YUV420P16BE:
392 case AV_PIX_FMT_YUV422P16LE:
393 case AV_PIX_FMT_YUV422P16BE:
394 case AV_PIX_FMT_YUV444P16LE:
395 case AV_PIX_FMT_YUV444P16BE:
396 case AV_PIX_FMT_YUV420P9BE:
397 case AV_PIX_FMT_YUV420P9LE:
398 case AV_PIX_FMT_YUV420P10BE:
399 case AV_PIX_FMT_YUV420P10LE:
400 case AV_PIX_FMT_YUV422P10BE:
401 case AV_PIX_FMT_YUV422P10LE:
402 case AV_PIX_FMT_YUV444P9BE:
403 case AV_PIX_FMT_YUV444P9LE:
404 case AV_PIX_FMT_YUV444P10BE:
405 case AV_PIX_FMT_YUV444P10LE:
406 case AV_PIX_FMT_YUV422P9BE:
407 case AV_PIX_FMT_YUV422P9LE:
408 case AV_PIX_FMT_YUVA420P9BE:
409 case AV_PIX_FMT_YUVA420P9LE:
410 case AV_PIX_FMT_YUVA422P9BE:
411 case AV_PIX_FMT_YUVA422P9LE:
412 case AV_PIX_FMT_YUVA444P9BE:
413 case AV_PIX_FMT_YUVA444P9LE:
414 case AV_PIX_FMT_YUVA420P10BE:
415 case AV_PIX_FMT_YUVA420P10LE:
416 case AV_PIX_FMT_YUVA422P10BE:
417 case AV_PIX_FMT_YUVA422P10LE:
418 case AV_PIX_FMT_YUVA444P10BE:
419 case AV_PIX_FMT_YUVA444P10LE:
420 case AV_PIX_FMT_YUVA420P16BE:
421 case AV_PIX_FMT_YUVA420P16LE:
422 case AV_PIX_FMT_YUVA422P16BE:
423 case AV_PIX_FMT_YUVA422P16LE:
424 case AV_PIX_FMT_YUVA444P16BE:
425 case AV_PIX_FMT_YUVA444P16LE:
426 case AV_PIX_FMT_NV16:
427 case AV_PIX_FMT_NV20LE:
428 case AV_PIX_FMT_NV20BE:
429 case AV_PIX_FMT_YVYU422:
430 case AV_PIX_FMT_YUVA444P:
431 case AV_PIX_FMT_YUVA422P:
432 case AV_PIX_FMT_YUV420P12BE:
433 case AV_PIX_FMT_YUV420P12LE:
434 case AV_PIX_FMT_YUV420P14BE:
435 case AV_PIX_FMT_YUV420P14LE:
436 case AV_PIX_FMT_YUV422P12BE:
437 case AV_PIX_FMT_YUV422P12LE:
438 case AV_PIX_FMT_YUV422P14BE:
439 case AV_PIX_FMT_YUV422P14LE:
440 case AV_PIX_FMT_YUV444P12BE:
441 case AV_PIX_FMT_YUV444P12LE:
442 case AV_PIX_FMT_YUV444P14BE:
443 case AV_PIX_FMT_YUV444P14LE:
444 case AV_PIX_FMT_YUVJ411P:
452 FFmpegExaminer::has_video () const
454 return static_cast<bool> (_video_stream);
458 FFmpegExaminer::range () const
460 switch (color_range()) {
461 case AVCOL_RANGE_MPEG:
462 case AVCOL_RANGE_UNSPECIFIED:
463 return VideoRange::VIDEO;
464 case AVCOL_RANGE_JPEG:
466 return VideoRange::FULL;