// TODO maybe send num-1 frames from here already
}
+typedef int channelmapping_t[8];
+channelmapping_t mapping[8] =
+{
+ { 0, -1, -1, -1, -1, -1, -1, -1 }, // mono
+ { 0, 1, -1, -1, -1, -1, -1, -1 }, // stereo
+ { 0, 1, 2, -1, -1, -1, -1, -1 }, // L C R
+ { 0, 1, 2, 3, -1, -1, -1, -1 }, // surround40
+ { 0, 2, 3, 4, 1, -1, -1, -1 }, // FL FC FR RL RR
+ { 0, 2, 3, 4, 1, 5, -1, -1 }, // surround51
+ { 0, 2, 3, 4, 1, 5, 6, -1 }, // (not defined by vorbis spec)
+ { 0, 2, 3, 4, 1, 5, 6, 7 } // surround71 (not defined by vorbis spec)
+};
+
static void SCR_CaptureVideo_Ogg_SoundFrame(const portable_sampleframe_t *paintbuffer, size_t length)
{
LOAD_FORMATSPECIFIC_OGG();
size_t i;
int j;
ogg_packet pt;
+ int *map = mapping[bound(1, cls.capturevideo.soundchannels, 8) - 1];
vorbis_buffer = qvorbis_analysis_buffer(&format->vd, length);
- for(i = 0; i < length; ++i)
+ for(j = 0; j < cls.capturevideo.soundchannels; ++j)
{
- for(j = 0; j < cls.capturevideo.soundchannels; ++j)
- vorbis_buffer[j][i] = paintbuffer[i].sample[j] / 32768.0f;
+ float *b = vorbis_buffer[map[j]];
+ for(i = 0; i < length; ++i)
+ b[i] = paintbuffer[i].sample[j] / 32768.0f;
}
qvorbis_analysis_wrote(&format->vd, length);
ti.keyframe_auto_threshold = bound(0, cl_capturevideo_ogg_theora_keyframe_auto_threshold.integer, 100);
ti.keyframe_frequency_force = ti.keyframe_frequency;
- ti.keyframe_auto_p = (ti.keyframe_frequency != ti.keyframe_mindistance);
+ ti.keyframe_auto_p = (ti.keyframe_frequency != ti.keyframe_mindistance + 1);
qtheora_encode_init(&format->ts, &ti);
qtheora_info_clear(&ti);