, _audio_codec (0)
, _subtitle_codec_context (0)
, _subtitle_codec (0)
+ , _first_video_pts (-1)
+ , _first_audio_pts (-1)
{
setup_general ();
setup_video ();
throw DecodeError ("could not find stream information");
}
+ /* Find video, audio and subtitle streams and choose the first of each */
+
for (uint32_t i = 0; i < _format_context->nb_streams; ++i) {
- if (_format_context->streams[i]->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
+ AVStream* s = _format_context->streams[i];
+ if (s->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
_video_stream = i;
- } else if (_format_context->streams[i]->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
- _audio_stream = i;
- } else if (_format_context->streams[i]->codec->codec_type == AVMEDIA_TYPE_SUBTITLE) {
- _subtitle_stream = i;
+ } else if (s->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
+ if (_audio_stream == -1) {
+ _audio_stream = i;
+ }
+ _audio_streams.push_back (AudioStream (stream_name (s), i, s->codec->channels));
+ } else if (s->codec->codec_type == AVMEDIA_TYPE_SUBTITLE) {
+ if (_subtitle_stream == -1) {
+ _subtitle_stream = i;
+ }
+ _subtitle_streams.push_back (SubtitleStream (stream_name (s), i));
}
}
+ /* 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().id();
+ }
+
+ if (_fs->subtitle_stream_index() != -1) {
+ _subtitle_stream = _fs->subtitle_stream().id ();
+ }
+
if (_video_stream < 0) {
throw DecodeError ("could not find video stream");
}
_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 == _fs->audio_channels());
process_audio (_frame->data[0], data_size);
}
}
if (_packet.stream_index == _video_stream) {
+ if (_first_video_pts == -1) {
+ _first_video_pts = _packet.pts;
+ }
+
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_pts != -1 && _packet.pts > _first_video_pts)) {
+
+ /* Note: We only decode audio if we've had our first video packet through, and if this
+ packet comes after it. Until then it is thrown away.
+ */
+
+ if (_first_audio_pts == -1) {
+ _first_audio_pts = _packet.pts;
+
+ /* 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.
+ */
+
+ AVStream* v = _format_context->streams[_video_stream];
+ AVStream* a = _format_context->streams[_audio_stream];
+
+ assert (v->time_base.num == a->time_base.num);
+ assert (v->time_base.den == a->time_base.den);
+
+ /* samples of silence that we must push */
+ int const s = rint (av_q2d (v->time_base) * (_first_audio_pts - _first_video_pts) * audio_sample_rate ());
+
+ _log->log (
+ String::compose (
+ "First video at %1, first audio at %2, pushing %3 samples of silence",
+ _first_video_pts, _first_audio_pts, s
+ )
+ );
+
+ /* 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 frame_finished;
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 == _fs->audio_channels());
process_audio (_frame->data[0], data_size);
}
int got_subtitle;
AVSubtitle sub;
if (avcodec_decode_subtitle2 (_subtitle_codec_context, &sub, &got_subtitle, &_packet) && got_subtitle) {
- process_subtitle (shared_ptr<Subtitle> (new Subtitle (sub)));
+ /* I'm not entirely sure why, but sometimes we get an AVSubtitle with
+ no AVSubtitleRects.
+ */
+ if (sub.num_rects > 0) {
+ process_subtitle (shared_ptr<TimedSubtitle> (new TimedSubtitle (sub)));
+ }
avsubtitle_free (&sub);
}
}
{
return (_subtitle_stream != -1);
}
+
+vector<AudioStream>
+FFmpegDecoder::audio_streams () const
+{
+ return _audio_streams;
+}
+
+vector<SubtitleStream>
+FFmpegDecoder::subtitle_streams () const
+{
+ return _subtitle_streams;
+}
+
+string
+FFmpegDecoder::stream_name (AVStream* s) const
+{
+ stringstream n;
+
+ AVDictionaryEntry const * lang = av_dict_get (s->metadata, "language", 0, 0);
+ if (lang) {
+ n << lang->value;
+ }
+
+ AVDictionaryEntry const * title = av_dict_get (s->metadata, "title", 0, 0);
+ if (title) {
+ if (!n.str().empty()) {
+ n << " ";
+ }
+ n << title->value;
+ }
+
+ if (n.str().empty()) {
+ n << "unknown";
+ }
+
+ return n.str ();
+}
+