libobs: Fix starting video packet offset in outputs

(This commit also modifies the obs-outputs module)

The first video packet video offset (the value used to set the starting
point of video data) would be set to the DTS value of the first video
packet.  However, when b-frames are used, the first DTS value will be
negative.  This was originally done because FLV muxing requires that the
first packet's DTS start from 0.  Unfortunately, this would also
effectively cause the first packet's PTS/DTS value to be shifted forward
by the negative amount, which would cause video sync to be off by a
video frame or two.

This fixes it to start at the PTS value instead and preserve any
negative offsets.  Additionally, the FLV muxing code has been fixed to
ensure that it adjusts the starting video DTS to 0, and now correctly
adjusts the first audio packet's timestamp according to that DTS as well
(which it didn't do before).

libobs/obs-output.c

@@ -1301,7 +1301,7 @@ static bool initialize_interleaved_packets(struct obs_output *output)
 	}
 
 	/* get new offsets */
-	output->video_offset = video->dts;
+	output->video_offset = video->pts;
 	for (size_t i = 0; i < audio_mixes; i++)
 		output->audio_offsets[i] = audio[i]->dts;
 

plugins/obs-outputs/flv-mux.c

@@ -177,11 +177,11 @@ bool flv_meta_data(obs_output_t *context, uint8_t **output, size_t *size,
 static int32_t last_time = 0;
 #endif
 
-static void flv_video(struct serializer *s, struct encoder_packet *packet,
-		bool is_header)
+static void flv_video(struct serializer *s, int32_t dts_offset,
+		struct encoder_packet *packet, bool is_header)
 {
 	int64_t offset  = packet->pts - packet->dts;
-	int32_t time_ms = get_ms_time(packet, packet->dts);
+	int32_t time_ms = get_ms_time(packet, packet->dts) - dts_offset;
 
 	if (!packet->data || !packet->size)
 		return;
@@ -212,10 +212,10 @@ static void flv_video(struct serializer *s, struct encoder_packet *packet,
 	s_wb32(s, (uint32_t)serializer_get_pos(s) + 4 - 1);
 }
 
-static void flv_audio(struct serializer *s, struct encoder_packet *packet,
-		bool is_header)
+static void flv_audio(struct serializer *s, int32_t dts_offset,
+		struct encoder_packet *packet, bool is_header)
 {
-	int32_t time_ms = get_ms_time(packet, packet->dts);
+	int32_t time_ms = get_ms_time(packet, packet->dts) - dts_offset;
 
 	if (!packet->data || !packet->size)
 		return;
@@ -245,7 +245,7 @@ static void flv_audio(struct serializer *s, struct encoder_packet *packet,
 	s_wb32(s, (uint32_t)serializer_get_pos(s) + 4 - 1);
 }
 
-void flv_packet_mux(struct encoder_packet *packet,
+void flv_packet_mux(struct encoder_packet *packet, int32_t dts_offset,
 		uint8_t **output, size_t *size, bool is_header)
 {
 	struct array_output_data data;
@@ -254,9 +254,9 @@ void flv_packet_mux(struct encoder_packet *packet,
 	array_output_serializer_init(&s, &data);
 
 	if (packet->type == OBS_ENCODER_VIDEO)
-		flv_video(&s, packet, is_header);
+		flv_video(&s, dts_offset, packet, is_header);
 	else
-		flv_audio(&s, packet, is_header);
+		flv_audio(&s, dts_offset, packet, is_header);
 
 	*output = data.bytes.array;
 	*size   = data.bytes.num;

plugins/obs-outputs/flv-mux.h

@@ -21,14 +21,14 @@
 
 #define MILLISECOND_DEN   1000
 
-static uint32_t get_ms_time(struct encoder_packet *packet, int64_t val)
+static int32_t get_ms_time(struct encoder_packet *packet, int64_t val)
 {
-	return (uint32_t)(val * MILLISECOND_DEN / packet->timebase_den);
+	return (int32_t)(val * MILLISECOND_DEN / packet->timebase_den);
 }
 
 extern void write_file_info(FILE *file, int64_t duration_ms, int64_t size);
 
 extern bool flv_meta_data(obs_output_t *context, uint8_t **output, size_t *size,
 		bool write_header, size_t audio_idx);
-extern void flv_packet_mux(struct encoder_packet *packet,
+extern void flv_packet_mux(struct encoder_packet *packet, int32_t dts_offset,
 		uint8_t **output, size_t *size, bool is_header);

plugins/obs-outputs/flv-output.c

@@ -32,12 +32,15 @@
 #define info(format, ...)  do_log(LOG_INFO,    format, ##__VA_ARGS__)
 
 struct flv_output {
-	obs_output_t *output;
-	struct dstr  path;
-	FILE         *file;
-	bool         active;
-	bool         sent_headers;
-	int64_t      last_packet_ts;
+	obs_output_t    *output;
+	struct dstr     path;
+	FILE            *file;
+	volatile bool   active;
+	bool            sent_headers;
+	int64_t         last_packet_ts;
+
+	bool            got_first_video;
+	int32_t         start_dts_offset;
 };
 
 static const char *flv_output_getname(void *unused)
@@ -98,7 +101,8 @@ static int write_packet(struct flv_output *stream,
 
 	stream->last_packet_ts = get_ms_time(packet, packet->dts);
 
-	flv_packet_mux(packet, &data, &size, is_header);
+	flv_packet_mux(packet, is_header ? 0 : stream->start_dts_offset,
+			&data, &size, is_header);
 	fwrite(data, 1, size, stream->file);
 	bfree(data);
 	obs_encoder_packet_release(packet);
@@ -168,6 +172,8 @@ static bool flv_output_start(void *data)
 	if (!obs_output_initialize_encoders(stream->output, 0))
 		return false;
 
+	stream->got_first_video = false;
+
 	/* get path */
 	settings = obs_output_get_settings(stream->output);
 	path = obs_data_get_string(settings, "path");
@@ -199,6 +205,12 @@ static void flv_output_data(void *data, struct encoder_packet *packet)
 	}
 
 	if (packet->type == OBS_ENCODER_VIDEO) {
+		if (!stream->got_first_video) {
+			stream->start_dts_offset =
+				get_ms_time(packet, packet->dts);
+			stream->got_first_video = true;
+		}
+
 		obs_parse_avc_packet(&parsed_packet, packet);
 		write_packet(stream, &parsed_packet, false);
 		obs_encoder_packet_release(&parsed_packet);

plugins/obs-outputs/rtmp-stream.c

@@ -347,7 +347,8 @@ static int send_packet(struct rtmp_stream *stream,
 		}
 	}
 
-	flv_packet_mux(packet, &data, &size, is_header);
+	flv_packet_mux(packet, is_header ? 0 : stream->start_dts_offset,
+			&data, &size, is_header);
 
 #ifdef TEST_FRAMEDROPS
 	droptest_cap_data_rate(stream, size);
@@ -876,6 +877,7 @@ static bool init_connect(struct rtmp_stream *stream)
 	stream->total_bytes_sent = 0;
 	stream->dropped_frames   = 0;
 	stream->min_priority     = 0;
+	stream->got_first_video  = false;
 
 	settings = obs_output_get_settings(stream->output);
 	dstr_copy(&stream->path,     obs_service_get_url(service));
@@ -1089,10 +1091,17 @@ static void rtmp_stream_data(void *data, struct encoder_packet *packet)
 	if (disconnected(stream) || !active(stream))
 		return;
 
-	if (packet->type == OBS_ENCODER_VIDEO)
+	if (packet->type == OBS_ENCODER_VIDEO) {
+		if (!stream->got_first_video) {
+			stream->start_dts_offset =
+				get_ms_time(packet, packet->dts);
+			stream->got_first_video = true;
+		}
+
 		obs_parse_avc_packet(&new_packet, packet);
-	else
+	} else {
 		obs_encoder_packet_ref(&new_packet, packet);
+	}
 
 	pthread_mutex_lock(&stream->packets_mutex);
 

plugins/obs-outputs/rtmp-stream.h

@@ -51,6 +51,9 @@ struct rtmp_stream {
 	struct circlebuf packets;
 	bool             sent_headers;
 
+	bool             got_first_video;
+	int64_t          start_dts_offset;
+
 	volatile bool    connecting;
 	pthread_t        connect_thread;