Try again to sort out the audio padding / alignment.

[dcpomatic.git] / src / lib / j2k_wav_encoder.cc

diff --git a/src/lib/j2k_wav_encoder.cc b/src/lib/j2k_wav_encoder.cc
index 58f8a101f7440c0be6312846114c9ebeed474011..c0c785d118d8254be4eb3489bca2806ca8f9a230 100644 (file)
--- a/src/lib/j2k_wav_encoder.cc
+++ b/src/lib/j2k_wav_encoder.cc
@@ -49,6 +49,7 @@ J2KWAVEncoder::J2KWAVEncoder (shared_ptr<const FilmState> s, shared_ptr<const Op
 #ifdef HAVE_SWRESAMPLE   
        , _swr_context (0)
 #endif   
+       , _audio_frames_written (0)
        , _process_end (false)
 {
        /* Create sound output files with .tmp suffixes; we will rename
@@ -216,13 +217,13 @@ J2KWAVEncoder::encoder_thread (ServerDescription* server)
 }
 
 void
-J2KWAVEncoder::process_begin (int64_t audio_channel_layout, AVSampleFormat audio_sample_format)
+J2KWAVEncoder::process_begin (int64_t audio_channel_layout)
 {
-       if (_fs->audio_sample_rate() != _fs->target_sample_rate()) {
+       if (_fs->audio_sample_rate() != _fs->target_audio_sample_rate()) {
 #ifdef HAVE_SWRESAMPLE
 
                stringstream s;
-               s << "Will resample audio from " << _fs->audio_sample_rate() << " to " << _fs->target_sample_rate();
+               s << "Will resample audio from " << _fs->audio_sample_rate() << " to " << _fs->target_audio_sample_rate();
                _log->log (s.str ());
 
                /* We will be using planar float data when we call the resampler */
@@ -230,7 +231,7 @@ J2KWAVEncoder::process_begin (int64_t audio_channel_layout, AVSampleFormat audio
                        0,
                        audio_channel_layout,
                        AV_SAMPLE_FMT_FLTP,
-                       _fs->target_sample_rate(),
+                       _fs->target_audio_sample_rate(),
                        audio_channel_layout,
                        AV_SAMPLE_FMT_FLTP,
                        _fs->audio_sample_rate(),
@@ -303,13 +304,10 @@ J2KWAVEncoder::process_end ()
 #if HAVE_SWRESAMPLE    
        if (_swr_context) {
 
-               float* out[_fs->audio_channels()];
-               for (int i = 0; i < _fs->audio_channels(); ++i) {
-                       out[i] = new float[256];
-               }
+               shared_ptr<AudioBuffers> out (new AudioBuffers (_fs->audio_channels(), 256));
                        
                while (1) {
-                       int const frames = swr_convert (_swr_context, (uint8_t **) out, 256, 0, 0);
+                       int const frames = swr_convert (_swr_context, (uint8_t **) out->data(), 256, 0, 0);
 
                        if (frames < 0) {
                                throw EncodeError ("could not run sample-rate converter");
@@ -319,16 +317,18 @@ J2KWAVEncoder::process_end ()
                                break;
                        }
 
-                       write_audio (out, frames);
-               }
-
-               for (int i = 0; i < _fs->audio_channels(); ++i) {
-                       delete[] out[i];
+                       write_audio (out);
                }
 
                swr_free (&_swr_context);
        }
-#endif 
+#endif
+
+       int const dcp_sr = dcp_audio_sample_rate (_fs->audio_sample_rate ());
+       int64_t const extra_audio_frames = dcp_sr - (_audio_frames_written % dcp_sr);
+       shared_ptr<AudioBuffers> silence (new AudioBuffers (_fs->audio_channels(), extra_audio_frames));
+       silence->make_silent ();
+       write_audio (silence);
        
        close_sound_files ();
 
@@ -342,50 +342,45 @@ J2KWAVEncoder::process_end ()
 }
 
 void
-J2KWAVEncoder::process_audio (float** data, int frames)
+J2KWAVEncoder::process_audio (shared_ptr<const AudioBuffers> audio)
 {
-       float* resampled[_fs->audio_channels()];
+       shared_ptr<AudioBuffers> resampled;
        
-#if HAVE_SWRESAMPLE    
+#if HAVE_SWRESAMPLE
        /* Maybe sample-rate convert */
        if (_swr_context) {
 
                /* Compute the resampled frames count and add 32 for luck */
-               int const resampled_frames = ceil (frames * _fs->target_sample_rate() / _fs->audio_sample_rate()) + 32;
+               int const max_resampled_frames = ceil (audio->frames() * _fs->target_audio_sample_rate() / _fs->audio_sample_rate()) + 32;
 
-               /* Make a buffer to put the result in */
-               for (int i = 0; i < _fs->audio_channels(); ++i) {
-                       resampled[i] = new float[resampled_frames];
-               }
+               resampled.reset (new AudioBuffers (_fs->audio_channels(), max_resampled_frames));
 
                /* Resample audio */
-               int out_frames = swr_convert (_swr_context, (uint8_t **) resampled, resampled_frames, (uint8_t const **) data, frames);
-               if (out_frames < 0) {
+               int const resampled_frames = swr_convert (
+                       _swr_context, (uint8_t **) resampled->data(), max_resampled_frames, (uint8_t const **) audio->data(), audio->frames()
+                       );
+               
+               if (resampled_frames < 0) {
                        throw EncodeError ("could not run sample-rate converter");
                }
 
+               resampled->set_frames (resampled_frames);
+               
                /* And point our variables at the resampled audio */
-               data = resampled;
-               frames = resampled_frames;
+               audio = resampled;
        }
 #endif
 
-       write_audio (data, frames);
-
-#if HAVE_SWRESAMPLE
-       if (_swr_context) {
-               for (int i = 0; i < _fs->audio_channels(); ++i) {
-                       delete[] resampled[i];
-               }
-       }
-#endif 
+       write_audio (audio);
 }
 
 void
-J2KWAVEncoder::write_audio (float** data, int frames)
+J2KWAVEncoder::write_audio (shared_ptr<const AudioBuffers> audio)
 {
        for (int i = 0; i < _fs->audio_channels(); ++i) {
-               sf_write_float (_sound_files[i], data[i], frames);
+               sf_write_float (_sound_files[i], audio->data(i), audio->frames());
        }
+
+       _audio_frames_written += audio->frames ();
 }