[FFmpeg-user] syncing live streams when mosaicing

[Daniel Oberhoff]

On 2013-02-06 10:01:19 +0000, Carl Eugen Hoyos <cehoyos at ag.or.at> said:

> Daniel Oberhoff <daniel <at> danieloberhoff.de> writes:
> 
>> now we know the streams are time stamped correctly
>> using the pts (in an h264 stream). can ffmpeg
>> somehow use the pts to avoid the shift?
> 
> Could you try to verify if FFmpeg's H264
> decoder actually correctly decodes the pts?
> 
> Carl Eugen

Ok, when I run with -loglevel debug I get outputs like this (on a 
single stream):

[force CFR for input from stream 0:0 @ 0x2d88ba0] N:159 PTS:572505 
T:22900.200000 POS:nan INTERLACED:0 -> PTS:159 T:6.360000
[force CFR for input from stream 0:0 @ 0x2d88ba0] N:160 PTS:576106 
T:23044.240000 POS:nan INTERLACED:0 -> PTS:160 T:6.400000
[force CFR for input from stream 0:0 @ 0x2d88ba0] N:161 PTS:579707 
T:23188.280000 POS:nan INTERLACED:0 -> PTS:161 T:6.440000
N:162 PTS:583308 T:23332.320000 POS:nan INTERLACED:0 -> PTS:162 T:6.480000
N:163 PTS:586908 T:23476.320000 POS:nan INTERLACED:0 -> PTS:163 
T:6.520000kbits/s

and for two streams:

[force CFR for input from stream 1:0 @ 0x2c95600] N:101 PTS:392473 
T:15698.920000 POS:nan INTERLACED:0 -> PTS:101 T:4.040000
[force CFR for input from stream 0:0 @ 0x2a72de0] N:100 PTS:360068 
T:14402.720000 POS:nan INTERLACED:0 -> PTS:100 T:4.000000
[Parsed_showinfo_2 @ 0x2a719c0] n:100 pts:100 pts_time:4 pos:-1 
fmt:yuv420p sar:1/1 s:2560x800 i:P iskey:0 type:P checksum:A7BECDB7 
plane_checksum:[6B83E2B5 01BEBCE0 05052E13]
[force CFR for input from stream 1:0 @ 0x2c95600] N:102 PTS:396072 
T:15842.880000 POS:nan INTERLACED:0 -> PTS:102 T:4.080000
[force CFR for input from stream 0:0 @ 0x2a72de0] N:101 PTS:363673 
T:14546.920000 POS:nan INTERLACED:0 -> PTS:101 T:4.040000

where you can see that the pts is not aligned