[Ffmpeg-devel-irc] ffmpeg.log.20190829

burek burek at teamnet.rs
Fri Aug 30 03:05:03 EEST 2019


[00:00:20 CEST] <dastan> With this one i get the speed of the process
[00:00:20 CEST] <dastan> tail -f "log" | grep -o -E "speed=[0-9]+.[0-9]+++[a-z]+|speed=[0-9]+.[0-9]++[a-z]+|speed=[0-9]+.[0-9]+[a-z]+|speed=[0-9]+[a-z]+"
[00:01:42 CEST] <dastan> With this one i get the file being opened at the moment
[00:01:42 CEST] <dastan> tail -f "log" | grep -o -E "Opening '/home/build/MSS-ONE/temp/cache/TN-27153d00/TN[0-9]+.ts|TN[0-9]++.ts|TN[0-9]+++.ts' for writing"
[00:01:52 CEST] <dastan> i need to show them together
[00:06:47 CEST] <klaxa> you can simply combine those to: grep -o -E "(speed=[0-9]+.[0-9]+++[a-z]+|speed=[0-9]+.[0-9]++[a-z]+|speed=[0-9]+.[0-9]+[a-z]+|speed=[0-9]+[a-z]+|Opening '/home/build/MSS-ONE/temp/cache/TN-27153d00/TN[0-9]+.ts|TN[0-9]++.ts|TN[0-9]+++.ts' for writing)" or can you not?
[00:12:56 CEST] <dastan> lets try
[00:14:23 CEST] <dastan> its not in the same line but its a good improvement
[00:14:53 CEST] <dastan> i get this speed=1.01x
[00:14:53 CEST] <dastan> TN829.ts' for writing
[00:15:47 CEST] <dastan> and i dea to show the info in two columns?
[06:17:01 CEST] <mr_lou> Does ffmpeg have any fancy way of detecting and removing socalled flash frames? I.e. frame in the video overexposed due to flash photography.
[06:28:14 CEST] <VikeStep> just wanted a quick sanity check, I've written a program that does the avcodec_send_packet/avcodec_receive_frame loop for every frame in a video. for a 1h20m 1080p h264 mp4 video at 30fps, it takes 4m32s to parse the entire video. Does that sound like a reasonable amount? Or would you expect it to be much faster?
[06:37:02 CEST] <furq> VikeStep: time ffmpeg -i foo.mp4 -f null -
[06:37:13 CEST] <furq> should at least give you something to compare it against
[06:37:29 CEST] <VikeStep> ah, I've never actually used the ffmpeg tool before so forgot that existed
[06:37:54 CEST] <VikeStep> will see how that goes
[06:44:25 CEST] <VikeStep> it gives pretty much the exact same speed, cheers
[07:54:37 CEST] <lain98> whats a good heuristic to check for vfr. i know vfrdet does this but i dont want to run a filter over the entire video
[07:59:12 CEST] <furq> https://ffmpeg.org/doxygen/trunk/structAVPacket.html#a622e758be29fd500aed0ffdc069550f7
[07:59:15 CEST] <furq> probably that
[08:00:02 CEST] <furq> ffprobe -show_packets if you're not using the spi
[08:00:03 CEST] <furq> api
[08:09:30 CEST] <lain98> furq: so AVPacket::duration
[08:09:32 CEST] <lain98> will be unknown(0) ?
[08:14:42 CEST] <furq> if it's cfr then it should be the same for every packet
[08:15:21 CEST] <lain98> but that still means i have to iterate over packets :D
[08:15:40 CEST] <furq> sure but you don't have to decode the video
[08:15:40 CEST] <lain98> hmm
[08:15:42 CEST] <lain98> yes thats true
[08:15:42 CEST] <furq> you can also break early if it is vfr
[08:16:51 CEST] <lain98> right now i just check avg_frame_rate.num==1 in stream
[08:16:51 CEST] <lain98> thanks furq
[08:17:43 CEST] <furq> there might be a better way, i've not used the api very much
[08:26:35 CEST] <JEEB> lain98: if you really want to know there's no other way than indexing :P
[08:26:40 CEST] <JEEB> you can try and attempt to hack around it
[08:26:53 CEST] <JEEB> but you have asked a technical question and technically there is just one answer to it
[08:26:59 CEST] <JEEB> I am sorry to bring it to you :P
[08:32:04 CEST] <lain98> JEEB: looks like most of my problems could be solved by indexing
[08:32:06 CEST] <lain98> :)
[08:33:30 CEST] <JEEB> lain98: the best stuff of course is when you have a container with non-exact timestamps like MPEG-TS or FLV. then you also have to have logic (like ffms2 has I think) which looks at some difference to constant frame rate and sets the content as CFR if the timestamps seem to be just oscillating because of time base being inexact
[08:34:09 CEST] <lain98> *mind blow*
[08:34:47 CEST] <lain98> codecs and containers are as varied as animals in the jungle
[08:53:56 CEST] <Mavrik> It is easier if you have a limited number of input options
[08:58:10 CEST] <lain98> got me thinking if i have a list of some non-standard fps like 29.97 and stuff and then combine the check with or the fps is a whole number like 1/x where i just check the avg_frame_rate.num==1
[09:07:41 CEST] <lain98> i should be fine ?
[13:27:52 CEST] <kepstin> note that 29.97 (actually exactly 30000/1001) is a standard fps
[13:31:41 CEST] <lain98> okay, so i was checking out a video. and the frames have duration of 1024 and 512 alternatively. but this is constant frame rate how ?
[13:32:44 CEST] <lain98> oh shit its an audio packet
[13:41:56 CEST] <pomaranc> does anyone have an example HLS stream with webvtt subtitles?
[13:42:07 CEST] <termos> hmm, I'm trying to init a bsf like so av_bitstream_filter_init("filter_units=remove_types=35") but it just returns nullptr, it's not null if I just try to specify the filter name "filter_units". Is there another way to specify the arguments?
[13:42:24 CEST] <termos> I should probably switch to the non-legacy API, but just doing this as a quick test
[16:30:04 CEST] <classsic> Hi somebody can explain what means the parameters "speed=1.01x" at the end of logs?
[16:30:51 CEST] <FooNess> classsic, the speed of transcoding relative to the actual length of the media on the file.
[16:31:18 CEST] <FooNess> classsic, so if you see, for example, "speed=1x" for a movie that lasts 90 minutes, the transcode took 90 minutes.
[16:31:38 CEST] <FooNess> If you saw "speed=2.00x", then it took 45 minutes to transcode it.
[16:31:44 CEST] <classsic> but, this appear when stream with vcodec copy to rtmp server, some cases under 1x
[16:32:21 CEST] <FooNess> Right; it could be faster or slower than the original time.
[16:32:26 CEST] <durandal_1707> yes, its normal
[16:32:28 CEST] <FooNess> Depending on a lot of things.
[16:32:38 CEST] <FooNess> Network latency/bandwidth, processing power, etc.
[16:49:45 CEST] <classsic> but that not means something is wrong, right?
[16:52:06 CEST] <kepstin> are you streaming from a file with -re, or restreaming another live source?
[17:19:44 CEST] <classsic> kepstin  restreaming frome another live source
[17:34:51 CEST] <dastan> hello people
[17:35:39 CEST] <dastan> i am redirecting the std out of a ffmpeg command to create a log
[17:36:25 CEST] <dastan> and i want to read this log but i want to only show the speed, can someone help me with the scripting
[17:37:12 CEST] <dastan> with ths next command i get speed=1.01 or speed=1.02
[17:37:44 CEST] <dastan> tail -f /home/build/MSS-ONE/temp/logs/TN-f9aa8baa/TN-f9aa8baa-cache.log | awk '/speed=/ { speed=$10 ; print speed ; }'
[17:38:53 CEST] <dastan> how can i save the number 1.01 or 1.04 or 1 in the variable speed?
[17:39:19 CEST] <pink_mist> since you're using awk, ask in an awk channel
[17:39:21 CEST] <pink_mist> this is #ffmpeg
[17:39:35 CEST] <dastan> there is a awk chnnel?
[17:39:39 CEST] <pink_mist> no clue
[17:40:20 CEST] <dastan> yep
[17:40:21 CEST] <dastan> #awk
[17:40:21 CEST] <pink_mist> #ffmpeg is not the awk channel however
[17:40:36 CEST] <dastan> i am workiung with ffmpeg
[17:40:39 CEST] <BtbN> You also asked the exact same question yesterday, and got a pretty good answer?
[17:41:09 CEST] <pink_mist> you're working with text data. the fact that it's ffmpeg producing it is completely irrelevant
[17:41:15 CEST] <dastan> we didnt get a solution with klaxa
[17:41:55 CEST] <dastan> he was more friendly than you
[17:42:20 CEST] <dastan> but dont worry, i will check in the awk channel
[18:15:26 CEST] <durandal_1707> i want working solution now and for free
[18:16:11 CEST] <durandal_1707> also ffmpeg developers do free consulting because it is free
[18:46:34 CEST] <classsic> talking about logs output, ffmpeg don`t write logs line by line right?
[18:47:12 CEST] <classsic> I get a looooooong line with speed, and other data.
[19:00:46 CEST] <durandal_1707> you do something wrong
[19:00:55 CEST] <durandal_1707> log output is fine
[19:02:24 CEST] <classsic> ok, I execute ffmpeg inside docker, maybe that`s the problem
[19:04:11 CEST] <kepstin> the stats line is terminated with \r instead of \n (that causes it to overwrite the same line on a terminal)
[19:04:32 CEST] <kepstin> you need to account for that when splitting lines
[19:06:24 CEST] <classsic> but I user tail .....ffmpeg.log, in bash env
[19:07:04 CEST] <classsic> *i use
[23:10:22 CEST] <TheDrode> Good afternoon, i am trying to re-stream an hls source to a Colable 5011 device using the command related in my pastebin (available in a second) I am getting broken video full of artifacts
[23:10:26 CEST] <TheDrode> what am i doing wrong?
[23:10:27 CEST] <TheDrode> http://paste.ubuntu.com/p/y42pzmXnyZ/
[23:10:37 CEST] <TheDrode> any help will be really appreciated
[23:18:17 CEST] <DHE> is anyone else getting a broken paste?
[23:18:47 CEST] <DHE> n/m
[23:19:05 CEST] <TheDrode> tIt takes a little while to load sir DHE
[23:25:38 CEST] <TheDrode> https://anonymousfiles.io/5b9fLcOa/
[23:25:42 CEST] <TheDrode> The Video looks like this:
[23:26:03 CEST] <TheDrode> and like this: https://anonymousfiles.io/RhiiUYUE/
[23:26:32 CEST] <TheDrode> i checked all the NICs, replaced them, just bought a more powerful CPU and still can't get it working
[23:48:47 CEST] <kepstin> wow, what a terrible pastebin site. the show raw paste link takes you to an ubuntu login page :/
[23:54:11 CEST] <TheDrode> kepstin: sorry, will take my log to another site right now :D
[23:56:02 CEST] <furq> could you run that with FFREPORT=level=24 ffmpeg ... instead of ffmpeg -report
[23:56:12 CEST] <furq> so we don't have to wade through 85k lines of debug messages
[23:57:44 CEST] <TheDrode> Sorry :D
[23:59:54 CEST] <TheDrode> furq: this is what i get with level=24
[23:59:55 CEST] <TheDrode> https://paste.laravel.io/694c6ebe-307d-4853-b6d0-02062fb00af7
[00:00:00 CEST] --- Fri Aug 30 2019


More information about the Ffmpeg-devel-irc mailing list