[Libav-user] How to get precise frame timestamp

[YIRAN LI]

Sorry Alex,

Seems it's my fault, av_frame_get_best_effort_timestamp was moved from
libavcode to libavutil.

I'll run test again.

Thanks


2013/10/31 YIRAN LI <mrfun.china at gmail.com>

> Hi Alex,
>
> Seems I couldn't find the source code of av_frame_get_best_effort_timestamp,
> so I have to ask here, because I found that ffmpeg 1.1.3 return -1 for
> every frame of same file.
>
> So definitely
> int64_t position = av_frame_get_best_effort_timestamp(pFrame);
> if ((position != AV_NOPTS_VALUE) && (position >=0))
>
> is needed to check if best effort timestamp is valid, but I'd like to know
> if it's enough.
>
> As you can see in ffmpeg2.0.1 (with which av_frame_get_best_effort_timestamp
> returns correct value) the best effort time stamp returned equals to  frame.pkt_dts,
> that's ok.
>
> But if it's possible, for exmpale, av_frame_get_best_effort_timestamp
>  returns all 0 for every frame?
>
> So in summary I'd like to know on what condition can I totally trust time
> stamp returned by this function (so I can forget about those packet.dts,
> packet.pts, frame.pkt_dts)
>
> Thanks!
>
>
> 2013/10/30 YIRAN LI <mrfun.china at gmail.com>
>
>> Thanks Alex,
>>
>> I just had a quick check, called
>> av_frame_get_best_effort_timestamp(pFrame); and seems
>> pFrame->best_effort_timestamp was filled with desired values (1st frame
>> having 0, 2nd having 1 and so on).
>>
>> I'll read through the link you gave me later on.
>>
>> Great Thanks!
>>
>>
>> 2013/10/30 Alex Cohn <alexcohn at netvision.net.il>
>>
>>> On Wed, Oct 30, 2013 at 5:51 AM, YIRAN LI <mrfun.china at gmail.com> wrote:
>>> >
>>> > Hi friends,
>>> >
>>> > Here I have a question about AVFrame pts needs your help.
>>> >
>>> > My application uses av_read_frame to get packets and then calls
>>>  avcodec_decode_video2 to get video frames. But recently I found one
>>> specific video file generates strange pts/dts.
>>> >
>>> > Below is a part of the log:
>>> >
>>> > 00:00:17.301  MAIN  timestamp_debug:  pkt.pts = -9223372036854775808,
>>> dts = 0
>>> > 00:00:17.301  MAIN  timestamp_debug:            frame.pkt_pts =
>>> -9223372036854775808, frame.pkt_dts = -9223372036854775808, frame.pts =
>>> -9223372036854775808
>>> > 00:00:17.301  1108  FFMPEG: looks like this file was encoded with
>>> (divx4/(old)xvid/opendivx) -> forcing low_delay flag
>>> >
>>> > 00:00:17.301  MAIN  timestamp_debug:  pkt.pts = -9223372036854775808,
>>> dts = 1
>>> > 00:00:17.301  MAIN  timestamp_debug:            frame.pkt_pts =
>>> -9223372036854775808, frame.pkt_dts = -9223372036854775808, frame.pts =
>>> -9223372036854775808
>>> > 00:00:17.301  MAIN  timestamp_debug:  pkt.pts = -9223372036854775808,
>>> dts = 2
>>> > 00:00:17.301  MAIN  timestamp_debug:            frame.pkt_pts =
>>> -9223372036854775808, frame.pkt_dts = -9223372036854775808, frame.pts =
>>> -9223372036854775808
>>> > 00:00:17.301  MAIN  timestamp_debug:  pkt.pts = -9223372036854775808,
>>> dts = 3
>>> > 00:00:17.301  MAIN  timestamp_debug:            frame.pkt_pts =
>>> -9223372036854775808, frame.pkt_dts = -9223372036854775808, frame.pts =
>>> -9223372036854775808
>>> > 00:00:17.301  MAIN  timestamp_debug:  pkt.pts = -9223372036854775808,
>>> dts = 4
>>> > 00:00:17.301  MAIN  timestamp_debug:  [a full frame : frame.pkt_pts =
>>> -9223372036854775808, frame.pkt_dts = 0], frame.pts = 0
>>> > 00:00:18.393  LVid236  timestamp_debug:  pkt.pts =
>>> -9223372036854775808, dts = 5
>>> > 00:00:18.408  LVid236  timestamp_debug:  [a full frame : frame.pkt_pts
>>> = -9223372036854775808, frame.pkt_dts = 1], frame.pts = 0
>>> > 00:00:18.439  LVid236  timestamp_debug:  pkt.pts =
>>> -9223372036854775808, dts = 6
>>> > 00:00:18.439  LVid236  timestamp_debug:  [a full frame : frame.pkt_pts
>>> = -9223372036854775808, frame.pkt_dts = 2], frame.pts = 0
>>> > 00:00:18.439  LSou259  FFMPEG: first_dts 0 not matching first dts
>>> 113424 in the queue
>>> > 00:00:18.486  LVid236  timestamp_debug:  pkt.pts =
>>> -9223372036854775808, dts = 7
>>> > 00:00:18.486  LVid236  timestamp_debug:  [a full frame : frame.pkt_pts
>>> = -9223372036854775808, frame.pkt_dts = 3], frame.pts = 0
>>> > 00:00:18.517  LVid236  timestamp_debug:  pkt.pts =
>>> -9223372036854775808, dts = 8
>>> > 00:00:18.517  LVid236  timestamp_debug:  [a full frame : frame.pkt_pts
>>> = -9223372036854775808, frame.pkt_dts = 4], frame.pts = 0
>>> > 00:00:18.549  LSou259  FFMPEG: first_dts 0 not matching first dts
>>> 114675 in the queue
>>> >
>>> > We can see that, it was until 4th packet is sent to decode function
>>> that we got the first full frame. and all frames had pts = 0
>>> >
>>> > So here I want to know, frame.pts, frame.pkt_dts, packet,dts. which is
>>> best to represent the time stamp of a video frame (I mean pts, but in this
>>> case, no pts is valid).
>>> >
>>> > For example, in this case, for the 1st frame, should I use
>>> frame.pkt_dts (that's 0), or pkt.pts (the last packet that generates this
>>> frame, in this case 4) as the time  position
>>> > of the frame?
>>> >
>>> > The file I used to test is here
>>> https://dl.dropboxusercontent.com/u/89678527/av_divx_24_yuv420p_mp3_44100_2_1.avi
>>> >
>>> > Thanks
>>>
>>>
>>> That's what best_effort_timestamp is for. See
>>> http://dranger.com/ffmpeg/tutorial05.html about different values of
>>> pts/dts.
>>>
>>> BR,
>>> Alex Cohn
>>> _______________________________________________
>>> Libav-user mailing list
>>> Libav-user at ffmpeg.org
>>> http://ffmpeg.org/mailman/listinfo/libav-user
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://ffmpeg.org/pipermail/libav-user/attachments/20131031/372258cc/attachment.html>