4 This document outlines the techniques used for doing synchronised playback of
7 Synchronisation in a GstPipeline is achieved using the following 3 components:
9 - a GstClock, which is global for all elements in a GstPipeline.
10 - Timestamps on a GstBuffer.
11 - the SEGMENT event preceding the buffers.
17 This object provides a counter that represents the current time in nanoseconds.
18 This value is called the absolute_time.
20 Different sources exist for this counter:
22 - the system time (with g_get_current_time() and with microsecond accuracy)
23 - an audio device (based on number of samples played)
24 - a network source based on packets received + timestamps in those packets (a
25 typical example is an RTP source)
28 In GStreamer any element can provide a GstClock object that can be used in the
29 pipeline. The GstPipeline object will select a clock from all the providers and
30 will distribute it to all other elements (see part-gstpipeline.txt).
32 A GstClock always counts time upwards and does not necessarily start at 0.
38 After a pipeline selected a clock it will maintain the running_time based on the
39 selected clock. This running_time represents the total time spent in the PLAYING
40 state and is calculated as follows:
42 - If the pipeline is NULL/READY, the running_time is undefined.
43 - In PAUSED, the running_time remains at the time when it was last
44 PAUSED. When the stream is PAUSED for the first time, the running_time
46 - In PLAYING, the running_time is the delta between the absolute_time
47 and the base time. The base time is defined as the absolute_time minus
48 the running_time at the time when the pipeline is set to PLAYING.
49 - after a flushing seek, the running_time is set to 0 (see part-seeking.txt).
50 This is accomplished by redistributing a new base_time to the elements that
53 This algorithm captures the running_time when the pipeline is set from PLAYING
54 to PAUSED and restores this time based on the current absolute_time when going
55 back to PLAYING. This allows for both clocks that progress when in the PAUSED
56 state (systemclock) and clocks that don't (audioclock).
58 The clock and pipeline now provide a running_time to all elements that want to
59 perform synchronisation. Indeed, the running time can be observed in each
60 element (during the PLAYING state) as:
62 C.running_time = absolute_time - base_time
64 We note C.running_time as the running_time obtained by looking at the clock.
65 This value is monotonically increasing at the rate of the clock.
71 The GstBuffer timestamps and the preceeding SEGMENT event (See
72 part-streams.txt) define a transformation of the buffer timestamps to
73 running_time as follows:
75 The following notation is used:
78 - B.timestamp = buffer timestamp (GST_BUFFER_TIMESTAMP)
80 NS: SEGMENT event preceeding the buffers.
81 - NS.start: start field in the SEGMENT event
82 - NS.stop: stop field in the SEGMENT event
83 - NS.rate: rate field of SEGMENT event
84 - NS.abs_rate: absolute value of rate field of SEGMENT event
85 - NS.time: time field in the SEGMENT event
86 - NS.accum: total accumulated time of all previous SEGMENT events. This
87 field is kept in the GstSegment structure.
89 Valid buffers for synchronisation are those with B.timestamp between NS.start
90 and NS.stop. All other buffers outside this range should be dropped or clipped
91 to these boundaries (see also part-segments.txt).
93 The following transformation to running_time exist:
96 B.running_time = (B.timestamp - NS.start) / NS.abs_rate + NS.accum
98 B.running_time = (NS.stop - B.timestamp) / NS.abs_rate + NS.accum
100 We write B.running_time as the running_time obtained from the SEGMENT event
101 and the buffers of that segment.
103 The first displayable buffer will yield a value of 0 (since B.timestamp ==
104 NS.start and NS.accum == 0).
106 For NS.rate > 1.0, the timestamps will be scaled down to increase the playback
107 rate. Likewise, a rate between 0.0 and 1.0 will slow down playback.
109 For negative rates, timestamps are received stop NS.stop to NS.start so that the
110 first buffer received will be transformed into B.running_time of 0 (B.timestamp ==
111 NS.stop and NS.accum == 0).
117 As we have seen, we can get a running_time:
119 - using the clock and the element's base_time with:
121 C.running_time = absolute_time - base_time
123 - using the buffer timestamp and the preceeding SEGMENT event as (assuming
124 positive playback rate):
126 B.running_time = (B.timestamp - NS.start) / NS.abs_rate + NS.accum
128 We prefix C. and B. before the two running times to note how they were
131 The task of synchronized playback is to make sure that we play a buffer with
132 B.running_time at the moment when the clock reaches the same C.running_time.
134 Thus the following must hold:
136 B.running_time = C.running_time
140 B.running_time = absolute_time - base_time
144 absolute_time = B.running_time + base_time
146 The absolute_time when a buffer with B.running_time should be played is noted
147 with B.sync_time. Thus:
149 B.sync_time = B.running_time + base_time
151 One then waits for the clock to reach B.sync_time before rendering the buffer in
152 the sink (See also part-clocks.txt).
154 For multiple streams this means that buffers with the same running_time are to
155 be displayed at the same time.
157 A demuxer must make sure that the SEGMENT it emits on its output pads yield
158 the same running_time for buffers that should be played synchronized. This
159 usually means sending the same SEGMENT on all pads and making sure that the
160 synchronized buffers have the same timestamps.
166 The stream time is also known as the position in the stream and is a value
167 between 0 and the total duration of the media file.
169 It is the stream time that is used for:
171 - report the POSITION query in the pipeline
172 - the position used in seek events/queries
173 - the position used to synchronize controller values
175 Stream time is calculated using the buffer times and the preceeding SEGMENT
178 stream_time = (B.timestamp - NS.start) * NS.abs_applied_rate + NS.time
180 For negative rates, B.timestamp will go backwards from NS.stop to NS.start,
181 making the stream time go backwards.
183 In the PLAYING state, it is also possible to use the pipeline clock to derive
184 the current stream_time.
186 Give the two formulas above to match the clock times with buffer timestamps
187 allows us to rewrite the above formula for stream_time (and for positive rates).
189 C.running_time = absolute_time - base_time
190 B.running_time = (B.timestamp - NS.start) / NS.abs_rate + NS.accum
193 (B.timestamp - NS.start) / NS.abs_rate + NS.accum = absolute_time - base_time;
196 (B.timestamp - NS.start) / NS.abs_rate = absolute_time - base_time - NS.accum;
199 (B.timestamp - NS.start) = (absolute_time - base_time - NS.accum) * NS.abs_rate
201 filling (B.timestamp - NS.start) in the above formule for stream time
204 stream_time = (absolute_time - base_time - NS.accum) * NS.abs_rate * NS.abs_applied_rate + NS.time
206 This last formula is typically used in sinks to report the current position in
207 an accurate and efficient way.
209 Note that the stream time is never used for synchronisation against the clock.