#include "transcoder.h"
#include "job.h"
#include "filter.h"
-#include "film_state.h"
#include "options.h"
#include "exceptions.h"
#include "image.h"
using namespace std;
using namespace boost;
-FFmpegDecoder::FFmpegDecoder (boost::shared_ptr<const FilmState> s, boost::shared_ptr<const Options> o, Job* j, Log* l, bool minimal, bool ignore_length)
- : Decoder (s, o, j, l, minimal, ignore_length)
+FFmpegDecoder::FFmpegDecoder (boost::shared_ptr<Film> f, boost::shared_ptr<const Options> o, Job* j, bool minimal, bool ignore_length)
+ : Decoder (f, o, j, minimal, ignore_length)
, _format_context (0)
, _video_stream (-1)
, _audio_stream (-1)
av_register_all ();
- if ((r = avformat_open_input (&_format_context, _fs->content_path().c_str(), 0, 0)) != 0) {
- throw OpenFileError (_fs->content_path ());
+ if ((r = avformat_open_input (&_format_context, _film->content_path().c_str(), 0, 0)) != 0) {
+ throw OpenFileError (_film->content_path ());
}
if (avformat_find_stream_info (_format_context, 0) < 0) {
/* Now override audio and subtitle streams with those from the Film, if it has any */
- if (_fs->audio_stream_index() != -1) {
- _audio_stream = _fs->audio_stream_decoder_id ();
+ if (_film->audio_stream_index() != -1) {
+ _audio_stream = _film->audio_stream().id();
}
- if (_fs->subtitle_stream_index() != -1) {
- _subtitle_stream = _fs->subtitle_stream_decoder_id ();
+ if (_film->subtitle_stream_index() != -1) {
+ _subtitle_stream = _film->subtitle_stream().id ();
}
if (_video_stream < 0) {
FFmpegDecoder::do_pass ()
{
int r = av_read_frame (_format_context, &_packet);
+
if (r < 0) {
if (r != AVERROR_EOF) {
throw DecodeError ("error on av_read_frame");
_packet.data = 0;
_packet.size = 0;
+ /* XXX: should we reset _packet.data and size after each *_decode_* call? */
+
int frame_finished;
while (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
int const data_size = av_samples_get_buffer_size (
0, _audio_codec_context->channels, _frame->nb_samples, audio_sample_format (), 1
);
-
- assert (_audio_codec_context->channels == _fs->audio_channels());
+
+ assert (_audio_codec_context->channels == _film->audio_channels());
process_audio (_frame->data[0], data_size);
}
}
return true;
}
+ double const pts_seconds = av_q2d (_format_context->streams[_packet.stream_index]->time_base) * _packet.pts;
+
if (_packet.stream_index == _video_stream) {
+ if (!_first_video) {
+ _first_video = pts_seconds;
+ }
+
int frame_finished;
if (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
process_video (_frame);
}
- } else if (_audio_stream >= 0 && _packet.stream_index == _audio_stream && _opt->decode_audio) {
+ } else if (_audio_stream >= 0 && _packet.stream_index == _audio_stream && _opt->decode_audio && _first_video && _first_video.get() <= pts_seconds) {
+
+ /* Note: We only decode audio if we've had our first video packet through, and if it
+ was before this packet. Until then audio is thrown away.
+ */
+
+ if (!_first_audio) {
+ _first_audio = pts_seconds;
+
+ /* This is our first audio packet, and if we've arrived here we must have had our
+ first video packet. Push some silence to make up the gap between our first
+ video packet and our first audio.
+ */
+
+ /* frames of silence that we must push */
+ int const s = rint ((_first_audio.get() - _first_video.get()) * audio_sample_rate ());
+
+ _log->log (
+ String::compose (
+ "First video at %1, first audio at %2, pushing %3 frames of silence for %4 channels (%5 bytes per sample)",
+ _first_video.get(), _first_audio.get(), s, audio_channels(), bytes_per_audio_sample()
+ )
+ );
+
+ /* hence bytes */
+ int const b = s * audio_channels() * bytes_per_audio_sample();
+
+ /* XXX: this assumes that it won't be too much, and there are shaky assumptions
+ that all sound representations are silent with memset()ed zero data.
+ */
+ uint8_t silence[b];
+ memset (silence, 0, b);
+ process_audio (silence, b);
+ }
avcodec_get_frame_defaults (_frame);
int const data_size = av_samples_get_buffer_size (
0, _audio_codec_context->channels, _frame->nb_samples, audio_sample_format (), 1
);
-
- assert (_audio_codec_context->channels == _fs->audio_channels());
+
+ assert (_audio_codec_context->channels == _film->audio_channels());
process_audio (_frame->data[0], data_size);
}
-
- } else if (_subtitle_stream >= 0 && _packet.stream_index == _subtitle_stream && _opt->decode_subtitles) {
+
+ } else if (_subtitle_stream >= 0 && _packet.stream_index == _subtitle_stream && _opt->decode_subtitles && _first_video) {
int got_subtitle;
AVSubtitle sub;
if (avcodec_decode_subtitle2 (_subtitle_codec_context, &sub, &got_subtitle, &_packet) && got_subtitle) {
- /* I'm not entirely sure why, but sometimes we get an AVSubtitle with
- no AVSubtitleRects.
+ /* Sometimes we get an empty AVSubtitle, which is used by some codecs to
+ indicate that the previous subtitle should stop.
*/
if (sub.num_rects > 0) {
- process_subtitle (shared_ptr<TimedSubtitle> (new TimedSubtitle (sub)));
+ process_subtitle (shared_ptr<TimedSubtitle> (new TimedSubtitle (sub, _first_video.get())));
+ } else {
+ process_subtitle (shared_ptr<TimedSubtitle> ());
}
avsubtitle_free (&sub);
}
return false;
}
-int
-FFmpegDecoder::length_in_frames () const
-{
- return (_format_context->duration / AV_TIME_BASE) * frames_per_second ();
-}
-
float
FFmpegDecoder::frames_per_second () const
{
return n.str ();
}
+