bool obs_output_initialize_encoders(obs_output_t *output, uint32_t flags)
{
bool encoded, has_video, has_audio, has_service;
size_t num_mixes = num_audio_mixes(output);
if (!output) return false;
if (output->active) return false;
convert_flags(output, flags, &encoded, &has_video, &has_audio,
&has_service);
if (!encoded)
return false;
if (has_service && !obs_service_initialize(output->service, output))
return false;
if (has_video && !obs_encoder_initialize(output->video_encoder))
return false;
if (has_audio && !initialize_audio_encoders(output, num_mixes))
return false;
if (has_video && has_audio) {
if (!pair_encoders(output, num_mixes)) {
return false;
}
}
return true;
}
static bool initialize_interleaved_packets(struct obs_output *output)
{
struct encoder_packet *video;
struct encoder_packet *audio[MAX_AUDIO_MIXES];
struct encoder_packet *last_audio[MAX_AUDIO_MIXES];
size_t audio_mixes = num_audio_mixes(output);
size_t start_idx;
if (!get_audio_and_video_packets(output, &video, audio, audio_mixes))
return false;
for (size_t i = 0; i < audio_mixes; i++)
last_audio[i] = find_last_packet_type(output, OBS_ENCODER_AUDIO,
i);
/* ensure that there is audio past the first video packet */
for (size_t i = 0; i < audio_mixes; i++) {
if (last_audio[i]->dts_usec < video->dts_usec) {
output->received_audio = false;
return false;
}
}
/* clear out excess starting audio if it hasn't been already */
start_idx = get_interleaved_start_idx(output);
if (start_idx) {
discard_to_idx(output, start_idx);
if (!get_audio_and_video_packets(output, &video, audio,
audio_mixes))
return false;
}
/* get new offsets */
output->video_offset = video->dts;
for (size_t i = 0; i < audio_mixes; i++)
output->audio_offsets[i] = audio[i]->dts;
#if DEBUG_STARTING_PACKETS == 1
int64_t v = video->dts_usec;
int64_t a = audio[0]->dts_usec;
int64_t diff = v - a;
blog(LOG_DEBUG, "output '%s' offset for video: %lld, audio: %lld, "
"diff: %lldms", output->context.name, v, a,
diff / 1000LL);
#endif
/* subtract offsets from highest TS offset variables */
output->highest_audio_ts -= audio[0]->dts_usec;
output->highest_video_ts -= video->dts_usec;
/* apply new offsets to all existing packet DTS/PTS values */
for (size_t i = 0; i < output->interleaved_packets.num; i++) {
struct encoder_packet *packet =
&output->interleaved_packets.array[i];
apply_interleaved_packet_offset(output, packet);
}
return true;
}
static inline void start_audio_encoders(struct obs_output *output,
encoded_callback_t encoded_callback)
{
size_t num_mixes = num_audio_mixes(output);
for (size_t i = 0; i < num_mixes; i++) {
obs_encoder_start(output->audio_encoders[i],
encoded_callback, output);
}
}
static inline bool audio_valid(const struct obs_output *output, bool encoded)
{
if (encoded) {
size_t mix_count = num_audio_mixes(output);
if (!mix_count)
return false;
for (size_t i = 0; i < mix_count; i++) {
if (!output->audio_encoders[i]) {
return false;
}
}
} else {
if (!output->audio)
return false;
}
return true;
}
static int prune_premature_packets(struct obs_output *output)
{
size_t audio_mixes = num_audio_mixes(output);
struct encoder_packet *video;
int video_idx;
int max_idx;
int64_t duration_usec;
int64_t max_diff = 0;
int64_t diff = 0;
video_idx = find_first_packet_type_idx(output, OBS_ENCODER_VIDEO, 0);
if (video_idx == -1) {
output->received_video = false;
return -1;
}
max_idx = video_idx;
video = &output->interleaved_packets.array[video_idx];
duration_usec = video->timebase_num * 1000000LL / video->timebase_den;
for (size_t i = 0; i < audio_mixes; i++) {
struct encoder_packet *audio;
int audio_idx;
audio_idx = find_first_packet_type_idx(output,
OBS_ENCODER_AUDIO, i);
if (audio_idx == -1) {
output->received_audio = false;
return -1;
}
audio = &output->interleaved_packets.array[audio_idx];
if (audio_idx > max_idx)
max_idx = audio_idx;
diff = audio->dts_usec - video->dts_usec;
if (diff > max_diff)
max_diff = diff;
}
return diff > duration_usec ? max_idx + 1 : 0;
}
static bool initialize_interleaved_packets(struct obs_output *output)
{
struct encoder_packet *video;
struct encoder_packet *audio[MAX_AUDIO_MIXES];
size_t audio_mixes = num_audio_mixes(output);
video = find_first_packet_type(output, OBS_ENCODER_VIDEO, 0);
if (!video)
output->received_video = false;
for (size_t i = 0; i < audio_mixes; i++) {
audio[i] = find_first_packet_type(output, OBS_ENCODER_AUDIO, i);
if (!audio[i]) {
output->received_audio = false;
return false;
}
}
if (!video) {
return false;
}
/* get new offsets */
output->video_offset = video->dts;
for (size_t i = 0; i < audio_mixes; i++)
output->audio_offsets[i] = audio[i]->dts;
/* subtract offsets from highest TS offset variables */
output->highest_audio_ts -= audio[0]->dts_usec;
output->highest_video_ts -= video->dts_usec;
/* apply new offsets to all existing packet DTS/PTS values */
for (size_t i = 0; i < output->interleaved_packets.num; i++) {
struct encoder_packet *packet =
&output->interleaved_packets.array[i];
apply_interleaved_packet_offset(output, packet);
}
return true;
}
