X-Git-Url: https://main.carlh.net/gitweb/?p=dcpomatic.git;a=blobdiff_plain;f=src%2Flib%2Fffmpeg_examiner.cc;h=581fead97cb05510bd22abfd15a8df5ebe08691e;hp=4fd1b6fe8a8495ed9d326036ef07d712fb8efb8a;hb=81646d1dd5b28fa05d8f134142dcbaed6314ebab;hpb=27bd2c44278ccf181b11ba961649a3e9b3ae03af diff --git a/src/lib/ffmpeg_examiner.cc b/src/lib/ffmpeg_examiner.cc index 4fd1b6fe8..581fead97 100644 --- a/src/lib/ffmpeg_examiner.cc +++ b/src/lib/ffmpeg_examiner.cc @@ -1,56 +1,75 @@ /* - Copyright (C) 2013-2015 Carl Hetherington + Copyright (C) 2013-2021 Carl Hetherington - This program is free software; you can redistribute it and/or modify + This file is part of DCP-o-matic. + + DCP-o-matic is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. - This program is distributed in the hope that it will be useful, + DCP-o-matic is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License - along with this program; if not, write to the Free Software - Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + along with DCP-o-matic. If not, see . */ -extern "C" { -#include -#include -#include -#include -} + +#include "dcpomatic_log.h" #include "ffmpeg_examiner.h" #include "ffmpeg_content.h" #include "job.h" #include "ffmpeg_audio_stream.h" #include "ffmpeg_subtitle_stream.h" #include "util.h" -#include "safe_stringstream.h" -#include +#include "warnings.h" +DCPOMATIC_DISABLE_WARNINGS +extern "C" { +#include +#include +#include +#include +#include +#include +} +DCPOMATIC_ENABLE_WARNINGS #include #include "i18n.h" -using std::string; + using std::cout; +using std::make_shared; using std::max; -using boost::shared_ptr; +using std::shared_ptr; +using std::string; +using std::vector; using boost::optional; +using namespace dcpomatic; + + +/* This is how many frames from the start of any video that we will examine to see if we + * can spot soft 2:3 pull-down ("telecine"). + */ +static const int PULLDOWN_CHECK_FRAMES = 16; + /** @param job job that the examiner is operating in, or 0 */ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptr job) : FFmpeg (c) , _video_length (0) , _need_video_length (false) + , _pulldown (false) { /* Find audio and subtitle streams */ for (uint32_t i = 0; i < _format_context->nb_streams; ++i) { - AVStream* s = _format_context->streams[i]; + auto s = _format_context->streams[i]; +DCPOMATIC_DISABLE_WARNINGS if (s->codec->codec_type == AVMEDIA_TYPE_AUDIO) { /* This is a hack; sometimes it seems that _audio_codec_context->channel_layout isn't set up, @@ -61,9 +80,18 @@ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptrcodec->channel_layout = av_get_default_channel_layout (s->codec->channels); } + DCPOMATIC_ASSERT (_format_context->duration != AV_NOPTS_VALUE); + DCPOMATIC_ASSERT (s->codec->codec); + DCPOMATIC_ASSERT (s->codec->codec->name); + _audio_streams.push_back ( - shared_ptr ( - new FFmpegAudioStream (audio_stream_name (s), s->id, s->codec->sample_rate, s->codec->channels) + make_shared( + stream_name (s), + s->codec->codec->name, + s->id, + s->codec->sample_rate, + llrint ((double(_format_context->duration) / AV_TIME_BASE) * s->codec->sample_rate), + s->codec->channels ) ); @@ -72,34 +100,36 @@ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptrduration == AV_NOPTS_VALUE; - if (!_need_video_length) { - _video_length = (double (_format_context->duration) / AV_TIME_BASE) * video_frame_rate().get (); + if (has_video ()) { + /* See if the header has duration information in it */ + _need_video_length = _format_context->duration == AV_NOPTS_VALUE; + if (!_need_video_length) { + _video_length = llrint ((double (_format_context->duration) / AV_TIME_BASE) * video_frame_rate().get()); + } } - if (job) { - if (_need_video_length) { - job->sub (_("Finding length and subtitles")); - } else { - job->sub (_("Finding subtitles")); - } + if (job && _need_video_length) { + job->sub (_("Finding length")); } /* Run through until we find: * - the first video. * - the first audio for each stream. - * - the subtitle periods for each stream. - * - * We have to note subtitle periods as otherwise we have no way of knowing - * where we should look for subtitles (video and audio are always present, - * so they are ok). + * - the top-field-first and repeat-first-frame values ("temporal_reference") for the first PULLDOWN_CHECK_FRAMES video frames. */ int64_t const len = _file_group.length (); + /* A string which we build up to describe the top-field-first and repeat-first-frame values for the first few frames. + * It would be nicer to use something like vector here but we want to search the array for a pattern later, + * and a string seems a reasonably neat way to do that. + */ + string temporal_reference; while (true) { - int r = av_read_frame (_format_context, &_packet); + auto packet = av_packet_alloc (); + DCPOMATIC_ASSERT (packet); + int r = av_read_frame (_format_context, packet); if (r < 0) { + av_packet_free (&packet); break; } @@ -111,128 +141,144 @@ FFmpegExaminer::FFmpegExaminer (shared_ptr c, shared_ptrstreams[_packet.stream_index]->codec; + auto context = _format_context->streams[packet->stream_index]->codec; +DCPOMATIC_ENABLE_WARNINGS - if (_packet.stream_index == _video_stream) { - video_packet (context); + if (_video_stream && packet->stream_index == _video_stream.get()) { + video_packet (context, temporal_reference, packet); } bool got_all_audio = true; for (size_t i = 0; i < _audio_streams.size(); ++i) { - if (_audio_streams[i]->uses_index (_format_context, _packet.stream_index)) { - audio_packet (context, _audio_streams[i]); + if (_audio_streams[i]->uses_index(_format_context, packet->stream_index)) { + audio_packet (context, _audio_streams[i], packet); } if (!_audio_streams[i]->first_audio) { got_all_audio = false; } } - for (size_t i = 0; i < _subtitle_streams.size(); ++i) { - if (_subtitle_streams[i]->uses_index (_format_context, _packet.stream_index)) { - subtitle_packet (context, _subtitle_streams[i]); - } - } - - av_free_packet (&_packet); + av_packet_free (&packet); - if (_first_video && got_all_audio && _subtitle_streams.empty ()) { + if (_first_video && got_all_audio && temporal_reference.size() >= (PULLDOWN_CHECK_FRAMES * 2)) { /* All done */ break; } } - for (LastSubtitleMap::const_iterator i = _last_subtitle_start.begin(); i != _last_subtitle_start.end(); ++i) { - if (i->second) { - i->first->add_subtitle ( - ContentTimePeriod ( - i->second.get (), - ContentTime::from_frames (video_length(), video_frame_rate().get_value_or (24)) - ) - ); - } + if (_video_stream) { + AVPacket packet; + av_init_packet (&packet); + packet.data = nullptr; + packet.size = 0; +DCPOMATIC_DISABLE_WARNINGS + auto context = _format_context->streams[*_video_stream]->codec; +DCPOMATIC_ENABLE_WARNINGS + while (video_packet(context, temporal_reference, &packet)) {} + } + + for (auto i: _audio_streams) { + AVPacket packet; + av_init_packet (&packet); + packet.data = nullptr; + packet.size = 0; +DCPOMATIC_DISABLE_WARNINGS + audio_packet (i->stream(_format_context)->codec, i, &packet); +DCPOMATIC_ENABLE_WARNINGS } - /* We just added subtitles to our streams without taking the PTS offset into account; - this is because we might not know the PTS offset when the first subtitle is seen. - Now we know the PTS offset so we can apply it to those subtitles. - */ - if (video_frame_rate()) { - BOOST_FOREACH (shared_ptr i, _subtitle_streams) { - i->add_offset (pts_offset (_audio_streams, _first_video, video_frame_rate().get())); + if (_video_stream) { + /* This code taken from get_rotation() in ffmpeg:cmdutils.c */ + auto stream = _format_context->streams[*_video_stream]; + auto rotate_tag = av_dict_get (stream->metadata, "rotate", 0, 0); + uint8_t* displaymatrix = av_stream_get_side_data (stream, AV_PKT_DATA_DISPLAYMATRIX, 0); + _rotation = 0; + + if (rotate_tag && *rotate_tag->value && strcmp(rotate_tag->value, "0")) { + char *tail; + _rotation = av_strtod (rotate_tag->value, &tail); + if (*tail) { + _rotation = 0; + } + } + + if (displaymatrix && !_rotation) { + _rotation = - av_display_rotation_get ((int32_t*) displaymatrix); } + + _rotation = *_rotation - 360 * floor (*_rotation / 360 + 0.9 / 360); + } + + LOG_GENERAL("Temporal reference was %1", temporal_reference); + if (temporal_reference.find("T2T3B2B3T2T3B2B3") != string::npos || temporal_reference.find("B2B3T2T3B2B3T2T3") != string::npos) { + /* The magical sequence (taken from mediainfo) suggests that 2:3 pull-down is in use */ + _pulldown = true; + LOG_GENERAL_NC("Suggest that this may be 2:3 pull-down (soft telecine)"); } } -void -FFmpegExaminer::video_packet (AVCodecContext* context) + +/** @param temporal_reference A string to which we should add two characters per frame; + * the first is T or B depending on whether it's top- or bottom-field first, + * the second is 3 or 2 depending on whether "repeat_pict" is true or not. + * @return true if some video was decoded, otherwise false. + */ +bool +FFmpegExaminer::video_packet (AVCodecContext* context, string& temporal_reference, AVPacket* packet) { - if (_first_video && !_need_video_length) { - return; + DCPOMATIC_ASSERT (_video_stream); + + if (_first_video && !_need_video_length && temporal_reference.size() >= (PULLDOWN_CHECK_FRAMES * 2)) { + return false; } int frame_finished; - if (avcodec_decode_video2 (context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) { - if (!_first_video) { - _first_video = frame_time (_format_context->streams[_video_stream]); - } - if (_need_video_length) { - _video_length = frame_time ( - _format_context->streams[_video_stream] - ).get_value_or (ContentTime ()).frames_round (video_frame_rate().get ()); - } +DCPOMATIC_DISABLE_WARNINGS + if (avcodec_decode_video2 (context, _frame, &frame_finished, packet) < 0 || !frame_finished) { + return false; + } +DCPOMATIC_ENABLE_WARNINGS + + if (!_first_video) { + _first_video = frame_time (_format_context->streams[_video_stream.get()]); } + if (_need_video_length) { + _video_length = frame_time ( + _format_context->streams[_video_stream.get()] + ).get_value_or (ContentTime ()).frames_round (video_frame_rate().get ()); + } + if (temporal_reference.size() < (PULLDOWN_CHECK_FRAMES * 2)) { + temporal_reference += (_frame->top_field_first ? "T" : "B"); + temporal_reference += (_frame->repeat_pict ? "3" : "2"); + } + + return true; } + void -FFmpegExaminer::audio_packet (AVCodecContext* context, shared_ptr stream) +FFmpegExaminer::audio_packet (AVCodecContext* context, shared_ptr stream, AVPacket* packet) { if (stream->first_audio) { return; } int frame_finished; - if (avcodec_decode_audio4 (context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) { +DCPOMATIC_DISABLE_WARNINGS + if (avcodec_decode_audio4 (context, _frame, &frame_finished, packet) >= 0 && frame_finished) { +DCPOMATIC_ENABLE_WARNINGS stream->first_audio = frame_time (stream->stream (_format_context)); } } -void -FFmpegExaminer::subtitle_packet (AVCodecContext* context, shared_ptr stream) -{ - int frame_finished; - AVSubtitle sub; - if (avcodec_decode_subtitle2 (context, &sub, &frame_finished, &_packet) >= 0 && frame_finished) { - FFmpegSubtitlePeriod const period = subtitle_period (sub); - LastSubtitleMap::iterator last = _last_subtitle_start.find (stream); - if (last != _last_subtitle_start.end() && last->second) { - /* We have seen the start of a subtitle but not yet the end. Whatever this is - finishes the previous subtitle, so add it */ - stream->add_subtitle (ContentTimePeriod (last->second.get (), period.from)); - if (sub.num_rects == 0) { - /* This is a `proper' end-of-subtitle */ - _last_subtitle_start[stream] = optional (); - } else { - /* This is just another subtitle, so we start again */ - _last_subtitle_start[stream] = period.from; - } - } else if (sub.num_rects == 1) { - if (period.to) { - stream->add_subtitle (ContentTimePeriod (period.from, period.to.get ())); - } else { - _last_subtitle_start[stream] = period.from; - } - } - avsubtitle_free (&sub); - } -} optional FFmpegExaminer::frame_time (AVStream* s) const { optional t; - int64_t const bet = av_frame_get_best_effort_timestamp (_frame); + int64_t const bet = _frame->best_effort_timestamp; if (bet != AV_NOPTS_VALUE) { t = ContentTime::from_seconds (bet * av_q2d (s->time_base)); } @@ -240,22 +286,22 @@ FFmpegExaminer::frame_time (AVStream* s) const return t; } + optional FFmpegExaminer::video_frame_rate () const { - /* This use of r_frame_rate is debateable; there's a few different - * frame rates in the format context, but this one seems to be the most - * reliable. - */ - return av_q2d (av_stream_get_r_frame_rate (_format_context->streams[_video_stream])); + DCPOMATIC_ASSERT (_video_stream); + return av_q2d(av_guess_frame_rate(_format_context, _format_context->streams[_video_stream.get()], 0)); } + dcp::Size FFmpegExaminer::video_size () const { return dcp::Size (video_codec_context()->width, video_codec_context()->height); } + /** @return Length according to our content's header */ Frame FFmpegExaminer::video_length () const @@ -263,72 +309,168 @@ FFmpegExaminer::video_length () const return max (Frame (1), _video_length); } + optional FFmpegExaminer::sample_aspect_ratio () const { - AVRational sar = av_guess_sample_aspect_ratio (_format_context, _format_context->streams[_video_stream], 0); + DCPOMATIC_ASSERT (_video_stream); + auto sar = av_guess_sample_aspect_ratio (_format_context, _format_context->streams[_video_stream.get()], 0); if (sar.num == 0) { /* I assume this means that we don't know */ - return optional (); + return {}; } return double (sar.num) / sar.den; } -string -FFmpegExaminer::audio_stream_name (AVStream* s) const -{ - SafeStringStream n; - - n << stream_name (s); - - if (!n.str().empty()) { - n << "; "; - } - - n << s->codec->channels << " channels"; - - return n.str (); -} string FFmpegExaminer::subtitle_stream_name (AVStream* s) const { - SafeStringStream n; - - n << stream_name (s); + auto n = stream_name (s); - if (n.str().empty()) { - n << _("unknown"); + if (n.empty()) { + n = _("unknown"); } - return n.str (); + return n; } + string FFmpegExaminer::stream_name (AVStream* s) const { - SafeStringStream n; + string n; if (s->metadata) { - AVDictionaryEntry const * lang = av_dict_get (s->metadata, "language", 0, 0); + auto const lang = av_dict_get (s->metadata, "language", 0, 0); if (lang) { - n << lang->value; + n = lang->value; } - AVDictionaryEntry const * title = av_dict_get (s->metadata, "title", 0, 0); + auto const title = av_dict_get (s->metadata, "title", 0, 0); if (title) { - if (!n.str().empty()) { - n << " "; + if (!n.empty()) { + n += " "; } - n << title->value; + n += title->value; } } - return n.str (); + return n; } -int + +optional FFmpegExaminer::bits_per_pixel () const { - return av_get_bits_per_pixel (av_pix_fmt_desc_get (video_codec_context()->pix_fmt)); + if (video_codec_context()->pix_fmt == -1) { + return {}; + } + + auto const d = av_pix_fmt_desc_get (video_codec_context()->pix_fmt); + DCPOMATIC_ASSERT (d); + return av_get_bits_per_pixel (d); +} + + +bool +FFmpegExaminer::yuv () const +{ + switch (video_codec_context()->pix_fmt) { + case AV_PIX_FMT_YUV420P: + case AV_PIX_FMT_YUYV422: + case AV_PIX_FMT_YUV422P: + case AV_PIX_FMT_YUV444P: + case AV_PIX_FMT_YUV410P: + case AV_PIX_FMT_YUV411P: + case AV_PIX_FMT_YUVJ420P: + case AV_PIX_FMT_YUVJ422P: + case AV_PIX_FMT_YUVJ444P: + case AV_PIX_FMT_UYVY422: + case AV_PIX_FMT_UYYVYY411: + case AV_PIX_FMT_NV12: + case AV_PIX_FMT_NV21: + case AV_PIX_FMT_YUV440P: + case AV_PIX_FMT_YUVJ440P: + case AV_PIX_FMT_YUVA420P: + case AV_PIX_FMT_YUV420P16LE: + case AV_PIX_FMT_YUV420P16BE: + case AV_PIX_FMT_YUV422P16LE: + case AV_PIX_FMT_YUV422P16BE: + case AV_PIX_FMT_YUV444P16LE: + case AV_PIX_FMT_YUV444P16BE: + case AV_PIX_FMT_YUV420P9BE: + case AV_PIX_FMT_YUV420P9LE: + case AV_PIX_FMT_YUV420P10BE: + case AV_PIX_FMT_YUV420P10LE: + case AV_PIX_FMT_YUV422P10BE: + case AV_PIX_FMT_YUV422P10LE: + case AV_PIX_FMT_YUV444P9BE: + case AV_PIX_FMT_YUV444P9LE: + case AV_PIX_FMT_YUV444P10BE: + case AV_PIX_FMT_YUV444P10LE: + case AV_PIX_FMT_YUV422P9BE: + case AV_PIX_FMT_YUV422P9LE: + case AV_PIX_FMT_YUVA420P9BE: + case AV_PIX_FMT_YUVA420P9LE: + case AV_PIX_FMT_YUVA422P9BE: + case AV_PIX_FMT_YUVA422P9LE: + case AV_PIX_FMT_YUVA444P9BE: + case AV_PIX_FMT_YUVA444P9LE: + case AV_PIX_FMT_YUVA420P10BE: + case AV_PIX_FMT_YUVA420P10LE: + case AV_PIX_FMT_YUVA422P10BE: + case AV_PIX_FMT_YUVA422P10LE: + case AV_PIX_FMT_YUVA444P10BE: + case AV_PIX_FMT_YUVA444P10LE: + case AV_PIX_FMT_YUVA420P16BE: + case AV_PIX_FMT_YUVA420P16LE: + case AV_PIX_FMT_YUVA422P16BE: + case AV_PIX_FMT_YUVA422P16LE: + case AV_PIX_FMT_YUVA444P16BE: + case AV_PIX_FMT_YUVA444P16LE: + case AV_PIX_FMT_NV16: + case AV_PIX_FMT_NV20LE: + case AV_PIX_FMT_NV20BE: + case AV_PIX_FMT_YVYU422: + case AV_PIX_FMT_YUVA444P: + case AV_PIX_FMT_YUVA422P: + case AV_PIX_FMT_YUV420P12BE: + case AV_PIX_FMT_YUV420P12LE: + case AV_PIX_FMT_YUV420P14BE: + case AV_PIX_FMT_YUV420P14LE: + case AV_PIX_FMT_YUV422P12BE: + case AV_PIX_FMT_YUV422P12LE: + case AV_PIX_FMT_YUV422P14BE: + case AV_PIX_FMT_YUV422P14LE: + case AV_PIX_FMT_YUV444P12BE: + case AV_PIX_FMT_YUV444P12LE: + case AV_PIX_FMT_YUV444P14BE: + case AV_PIX_FMT_YUV444P14LE: + case AV_PIX_FMT_YUVJ411P: + return true; + default: + return false; + } +} + + +bool +FFmpegExaminer::has_video () const +{ + return static_cast(_video_stream); +} + + +VideoRange +FFmpegExaminer::range () const +{ + switch (color_range()) { + case AVCOL_RANGE_MPEG: + case AVCOL_RANGE_UNSPECIFIED: + return VideoRange::VIDEO; + case AVCOL_RANGE_JPEG: + default: + return VideoRange::FULL; + } }