/*
- Copyright (C) 2012-2015 Carl Hetherington <cth@carlh.net>
+ Copyright (C) 2012-2016 Carl Hetherington <cth@carlh.net>
- 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 <http://www.gnu.org/licenses/>.
*/
#include "util.h"
#include "log.h"
#include "ffmpeg_decoder.h"
+#include "subtitle_decoder.h"
#include "ffmpeg_audio_stream.h"
#include "ffmpeg_subtitle_stream.h"
#include "video_filter_graph.h"
#include "audio_buffers.h"
#include "ffmpeg_content.h"
#include "raw_image_proxy.h"
+#include "video_decoder.h"
#include "film.h"
+#include "md5_digester.h"
+#include "audio_decoder.h"
#include "compose.hpp"
+#include <dcp/subtitle_string.h>
+#include <sub/ssa_reader.h>
+#include <sub/subtitle.h>
+#include <sub/collect.h>
extern "C" {
#include <libavcodec/avcodec.h>
#include <libavformat/avformat.h>
}
#include <boost/foreach.hpp>
+#include <boost/algorithm/string.hpp>
#include <vector>
#include <iomanip>
#include <iostream>
#define LOG_WARNING(...) _log->log (String::compose (__VA_ARGS__), LogEntry::TYPE_WARNING);
using std::cout;
+using std::string;
using std::vector;
using std::list;
using std::min;
using std::pair;
using std::max;
+using std::map;
using boost::shared_ptr;
+using boost::is_any_of;
+using boost::split;
using dcp::Size;
FFmpegDecoder::FFmpegDecoder (shared_ptr<const FFmpegContent> c, shared_ptr<Log> log, bool fast)
- : VideoDecoder (c)
- , AudioDecoder (c, fast)
- , SubtitleDecoder (c)
- , FFmpeg (c)
+ : FFmpeg (c)
, _log (log)
- , _pts_offset (pts_offset (c->ffmpeg_audio_streams(), c->first_video(), c->video_frame_rate()))
{
+ if (c->video) {
+ video.reset (new VideoDecoder (this, c, log));
+ _pts_offset = pts_offset (c->ffmpeg_audio_streams(), c->first_video(), c->active_video_frame_rate());
+ } else {
+ _pts_offset = ContentTime ();
+ }
+
+ if (c->audio) {
+ audio.reset (new AudioDecoder (this, c->audio, fast, log));
+ }
+ if (c->subtitle) {
+ subtitle.reset (
+ new SubtitleDecoder (
+ this,
+ c->subtitle,
+ bind (&FFmpegDecoder::image_subtitles_during, this, _1, _2),
+ bind (&FFmpegDecoder::text_subtitles_during, this, _1, _2)
+ )
+ );
+ }
}
void
/* XXX: should we reset _packet.data and size after each *_decode_* call? */
- while (decode_video_packet ()) {}
+ while (video && decode_video_packet ()) {}
- decode_audio_packet ();
- AudioDecoder::flush ();
+ if (audio) {
+ decode_audio_packet ();
+ audio->flush ();
+ }
}
bool
int const si = _packet.stream_index;
shared_ptr<const FFmpegContent> fc = _ffmpeg_content;
- if (si == _video_stream && !_ignore_video && (accurate || reason != PASS_REASON_SUBTITLE)) {
+ if (_video_stream && si == _video_stream.get() && !video->ignore() && (accurate || reason != PASS_REASON_SUBTITLE)) {
decode_video_packet ();
} else if (fc->subtitle_stream() && fc->subtitle_stream()->uses_index (_format_context, si)) {
decode_subtitle_packet ();
decode_audio_packet ();
}
- av_free_packet (&_packet);
+ av_packet_unref (&_packet);
return false;
}
* Only the first buffer will be used for non-planar data, otherwise there will be one per channel.
*/
shared_ptr<AudioBuffers>
-FFmpegDecoder::deinterleave_audio (shared_ptr<FFmpegAudioStream> stream, uint8_t** data, int size)
+FFmpegDecoder::deinterleave_audio (shared_ptr<FFmpegAudioStream> stream) const
{
DCPOMATIC_ASSERT (bytes_per_audio_sample (stream));
+ int const size = av_samples_get_buffer_size (
+ 0, stream->stream(_format_context)->codec->channels, _frame->nb_samples, audio_sample_format (stream), 1
+ );
+
/* Deinterleave and convert to float */
/* total_samples and frames will be rounded down here, so if there are stray samples at the end
switch (audio_sample_format (stream)) {
case AV_SAMPLE_FMT_U8:
{
- uint8_t* p = reinterpret_cast<uint8_t *> (data[0]);
+ uint8_t* p = reinterpret_cast<uint8_t *> (_frame->data[0]);
int sample = 0;
int channel = 0;
for (int i = 0; i < total_samples; ++i) {
case AV_SAMPLE_FMT_S16:
{
- int16_t* p = reinterpret_cast<int16_t *> (data[0]);
+ int16_t* p = reinterpret_cast<int16_t *> (_frame->data[0]);
int sample = 0;
int channel = 0;
for (int i = 0; i < total_samples; ++i) {
case AV_SAMPLE_FMT_S16P:
{
- int16_t** p = reinterpret_cast<int16_t **> (data);
+ int16_t** p = reinterpret_cast<int16_t **> (_frame->data);
for (int i = 0; i < stream->channels(); ++i) {
for (int j = 0; j < frames; ++j) {
audio->data(i)[j] = static_cast<float>(p[i][j]) / (1 << 15);
case AV_SAMPLE_FMT_S32:
{
- int32_t* p = reinterpret_cast<int32_t *> (data[0]);
+ int32_t* p = reinterpret_cast<int32_t *> (_frame->data[0]);
int sample = 0;
int channel = 0;
for (int i = 0; i < total_samples; ++i) {
- audio->data(channel)[sample] = static_cast<float>(*p++) / (1 << 31);
+ audio->data(channel)[sample] = static_cast<float>(*p++) / 2147483648;
++channel;
if (channel == stream->channels()) {
}
break;
+ case AV_SAMPLE_FMT_S32P:
+ {
+ int32_t** p = reinterpret_cast<int32_t **> (_frame->data);
+ for (int i = 0; i < stream->channels(); ++i) {
+ for (int j = 0; j < frames; ++j) {
+ audio->data(i)[j] = static_cast<float>(p[i][j]) / 2147483648;
+ }
+ }
+ }
+ break;
+
case AV_SAMPLE_FMT_FLT:
{
- float* p = reinterpret_cast<float*> (data[0]);
+ float* p = reinterpret_cast<float*> (_frame->data[0]);
int sample = 0;
int channel = 0;
for (int i = 0; i < total_samples; ++i) {
case AV_SAMPLE_FMT_FLTP:
{
- float** p = reinterpret_cast<float**> (data);
- for (int i = 0; i < stream->channels(); ++i) {
+ float** p = reinterpret_cast<float**> (_frame->data);
+ /* Sometimes there aren't as many channels in the _frame as in the stream */
+ for (int i = 0; i < _frame->channels; ++i) {
memcpy (audio->data(i), p[i], frames * sizeof(float));
}
+ for (int i = _frame->channels; i < stream->channels(); ++i) {
+ audio->make_silent (i);
+ }
}
break;
void
FFmpegDecoder::seek (ContentTime time, bool accurate)
{
- VideoDecoder::seek (time, accurate);
- AudioDecoder::seek (time, accurate);
- SubtitleDecoder::seek (time, accurate);
+ if (video) {
+ video->seek (time, accurate);
+ }
+
+ if (audio) {
+ audio->seek (time, accurate);
+ }
+
+ if (subtitle) {
+ subtitle->seek (time, accurate);
+ }
/* If we are doing an `accurate' seek, we need to use pre-roll, as
we don't really know what the seek will give us.
http://www.mjbshaw.com/2012/04/seeking-in-ffmpeg-know-your-timestamp.html
*/
+ DCPOMATIC_ASSERT (_video_stream);
+
ContentTime u = time - _pts_offset;
if (u < ContentTime ()) {
u = ContentTime ();
}
- av_seek_frame (_format_context, _video_stream, u.seconds() / av_q2d (_format_context->streams[_video_stream]->time_base), AVSEEK_FLAG_BACKWARD);
+ av_seek_frame (
+ _format_context,
+ _video_stream.get(),
+ u.seconds() / av_q2d (_format_context->streams[_video_stream.get()]->time_base),
+ AVSEEK_FLAG_BACKWARD
+ );
avcodec_flush_buffers (video_codec_context());
av_q2d ((*stream)->stream (_format_context)->time_base))
+ _pts_offset;
- int const data_size = av_samples_get_buffer_size (
- 0, (*stream)->stream(_format_context)->codec->channels, _frame->nb_samples, audio_sample_format (*stream), 1
- );
-
- shared_ptr<AudioBuffers> data = deinterleave_audio (*stream, _frame->data, data_size);
+ shared_ptr<AudioBuffers> data = deinterleave_audio (*stream);
if (ct < ContentTime ()) {
/* Discard audio data that comes before time 0 */
}
if (data->frames() > 0) {
- audio (*stream, data, ct);
+ audio->give (*stream, data, ct);
}
}
bool
FFmpegDecoder::decode_video_packet ()
{
+ DCPOMATIC_ASSERT (_video_stream);
+
int frame_finished;
if (avcodec_decode_video2 (video_codec_context(), _frame, &frame_finished, &_packet) < 0 || !frame_finished) {
return false;
shared_ptr<Image> image = i->first;
if (i->second != AV_NOPTS_VALUE) {
- double const pts = i->second * av_q2d (_format_context->streams[_video_stream]->time_base) + _pts_offset.seconds ();
- video (
+ double const pts = i->second * av_q2d (_format_context->streams[_video_stream.get()]->time_base) + _pts_offset.seconds ();
+ video->give (
shared_ptr<ImageProxy> (new RawImageProxy (image)),
- llrint (pts * _ffmpeg_content->video_frame_rate ())
+ llrint (pts * _ffmpeg_content->active_video_frame_rate ())
);
} else {
LOG_WARNING_NC ("Dropping frame without PTS");
period.to = sub_period.to.get() + _pts_offset;
} else {
/* We have to look up the `to' time in the stream's records */
- period.to = ffmpeg_content()->subtitle_stream()->find_subtitle_to (period.from);
+ period.to = ffmpeg_content()->subtitle_stream()->find_subtitle_to (subtitle_id (sub));
}
for (unsigned int i = 0; i < sub.num_rects; ++i) {
cout << "XXX: SUBTITLE_TEXT " << rect->text << "\n";
break;
case SUBTITLE_ASS:
- cout << "XXX: SUBTITLE_ASS " << rect->ass << "\n";
+ decode_ass_subtitle (rect->ass, period);
break;
}
}
list<ContentTimePeriod>
FFmpegDecoder::image_subtitles_during (ContentTimePeriod p, bool starting) const
{
- return _ffmpeg_content->subtitles_during (p, starting);
+ return _ffmpeg_content->image_subtitles_during (p, starting);
}
list<ContentTimePeriod>
-FFmpegDecoder::text_subtitles_during (ContentTimePeriod, bool) const
+FFmpegDecoder::text_subtitles_during (ContentTimePeriod p, bool starting) const
{
- return list<ContentTimePeriod> ();
+ return _ffmpeg_content->text_subtitles_during (p, starting);
}
void
*/
shared_ptr<Image> image (new Image (AV_PIX_FMT_RGBA, dcp::Size (rect->w, rect->h), true));
+#ifdef DCPOMATIC_HAVE_AVSUBTITLERECT_PICT
/* Start of the first line in the subtitle */
uint8_t* sub_p = rect->pict.data[0];
/* sub_p looks up into a BGRA palette which is here
(i.e. first byte B, second G, third R, fourth A)
*/
uint32_t const * palette = (uint32_t *) rect->pict.data[1];
+#else
+ /* Start of the first line in the subtitle */
+ uint8_t* sub_p = rect->data[0];
+ /* sub_p looks up into a BGRA palette which is here
+ (i.e. first byte B, second G, third R, fourth A)
+ */
+ uint32_t const * palette = (uint32_t *) rect->data[1];
+#endif
+ /* And the stream has a map of those palette colours to colours
+ chosen by the user; created a `mapped' palette from those settings.
+ */
+ map<RGBA, RGBA> colour_map = ffmpeg_content()->subtitle_stream()->colours ();
+ vector<RGBA> mapped_palette (rect->nb_colors);
+ for (int i = 0; i < rect->nb_colors; ++i) {
+ RGBA c ((palette[i] & 0xff0000) >> 16, (palette[i] & 0xff00) >> 8, palette[i] & 0xff, (palette[i] & 0xff000000) >> 24);
+ map<RGBA, RGBA>::const_iterator j = colour_map.find (c);
+ if (j != colour_map.end ()) {
+ mapped_palette[i] = j->second;
+ } else {
+ /* This colour was not found in the FFmpegSubtitleStream's colour map; probably because
+ it is from a project that was created before this stuff was added. Just use the
+ colour straight from the original palette.
+ */
+ mapped_palette[i] = c;
+ }
+ }
+
/* Start of the output data */
uint32_t* out_p = (uint32_t *) image->data()[0];
uint8_t* sub_line_p = sub_p;
uint32_t* out_line_p = out_p;
for (int x = 0; x < rect->w; ++x) {
- uint32_t const p = palette[*sub_line_p++];
- *out_line_p++ = ((p & 0xff) << 16) | (p & 0xff00) | ((p & 0xff0000) >> 16) | (p & 0xff000000);
+ RGBA const p = mapped_palette[*sub_line_p++];
+ /* XXX: this seems to be wrong to me (isn't the output image RGBA?) but it looks right on screen */
+ *out_line_p++ = (p.a << 24) | (p.r << 16) | (p.g << 8) | p.b;
}
+#ifdef DCPOMATIC_HAVE_AVSUBTITLERECT_PICT
sub_p += rect->pict.linesize[0];
+#else
+ sub_p += rect->linesize[0];
+#endif
out_p += image->stride()[0] / sizeof (uint32_t);
}
- dcp::Size const vs = _ffmpeg_content->video_size ();
+ dcp::Size const vs = _ffmpeg_content->video->size ();
dcpomatic::Rect<double> const scaled_rect (
static_cast<double> (rect->x) / vs.width,
static_cast<double> (rect->y) / vs.height,
static_cast<double> (rect->h) / vs.height
);
- image_subtitle (period, image, scaled_rect);
+ subtitle->give_image (period, image, scaled_rect);
+}
+
+void
+FFmpegDecoder::decode_ass_subtitle (string ass, ContentTimePeriod period)
+{
+ /* We have no styles and no Format: line, so I'm assuming that FFmpeg
+ produces a single format of Dialogue: lines...
+ */
+
+ vector<string> bits;
+ split (bits, ass, is_any_of (","));
+ if (bits.size() < 10) {
+ return;
+ }
+
+ sub::RawSubtitle base;
+ list<sub::RawSubtitle> raw = sub::SSAReader::parse_line (base, bits[9]);
+ list<sub::Subtitle> subs = sub::collect<list<sub::Subtitle> > (raw);
+
+ /* XXX: lots of this is copied from TextSubtitle; there should probably be some sharing */
+
+ /* Highest line index in this subtitle */
+ int highest = 0;
+ BOOST_FOREACH (sub::Subtitle i, subs) {
+ BOOST_FOREACH (sub::Line j, i.lines) {
+ DCPOMATIC_ASSERT (j.vertical_position.reference && j.vertical_position.reference.get() == sub::TOP_OF_SUBTITLE);
+ DCPOMATIC_ASSERT (j.vertical_position.line);
+ highest = max (highest, j.vertical_position.line.get());
+ }
+ }
+
+ list<dcp::SubtitleString> ss;
+
+ BOOST_FOREACH (sub::Subtitle i, sub::collect<list<sub::Subtitle> > (sub::SSAReader::parse_line (base, bits[9]))) {
+ BOOST_FOREACH (sub::Line j, i.lines) {
+ BOOST_FOREACH (sub::Block k, j.blocks) {
+ ss.push_back (
+ dcp::SubtitleString (
+ boost::optional<string> (),
+ k.italic,
+ k.bold,
+ dcp::Colour (255, 255, 255),
+ /* 48pt is 1/22nd of the screen height */
+ 48,
+ 1,
+ dcp::Time (i.from.seconds(), 1000),
+ dcp::Time (i.to.seconds(), 1000),
+ 0,
+ dcp::HALIGN_CENTER,
+ /* This 1.015 is an arbitrary value to lift the bottom sub off the bottom
+ of the screen a bit to a pleasing degree.
+ */
+ 1.015 - ((1 + highest - j.vertical_position.line.get()) * 1.5 / 22),
+ dcp::VALIGN_TOP,
+ dcp::DIRECTION_LTR,
+ k.text,
+ static_cast<dcp::Effect> (0),
+ dcp::Colour (255, 255, 255),
+ dcp::Time (),
+ dcp::Time ()
+ )
+ );
+ }
+ }
+ }
+
+ subtitle->give_text (period, ss);
}
projects / dcpomatic.git / blobdiff