[Ffmpeg-devel-irc] ffmpeg.log.20190517

[burek]

[02:13:09 CEST] <johnharveybc> For anyone keeping score at home, `ffmpeg -i input.mov -copyts -output_ts_offset 30` moves the time offset 30 seconds in the future.
[03:51:05 CEST] <kepstin> johnnyhostile: not sure what you mean - the first frame of the video is the first frame. If you want it to be frame N, you have to add more frames before it
[03:51:29 CEST] <kepstin> er, nvm, johnharveybc is gone
[05:33:11 CEST] <thebombzen> did vf_crop change recently? I'm getting a segfault with it
[05:34:45 CEST] <kepstin> if you're getting a segfault with vf_crop, it's usually because you've cropped to an unaligned size, and the *next* filter can't handle the unalignment.
[05:35:54 CEST] <kepstin> crop itself doesn't touch the frame data/buffers, it just adjusts pointers and line length values.
[05:37:05 CEST] <kepstin> the crop filter hasn't been touched in over a year, fwiw
[05:43:03 CEST] <kepstin> if you're getting a sefgault in the crop filter...? i can't think of any way that could happen using ffmpeg cli. If you're using api and pass it an invalid avframe i guess bad things will happen.
[05:48:58 CEST] <thebombzen> kepstin: well I'm getting a segfault for taking 2160x1080 video and changing it to 1920x1080, with crop=1920:1080:120:0
[05:49:22 CEST] <thebombzen> this doesn't happen with ffmpeg 4.1.3, it only happens with git master
[05:49:39 CEST] <thebombzen> the filter *right afterward* is setpts
[05:49:44 CEST] <thebombzen> possibly relevant
[05:49:55 CEST] <kepstin> complete fmpeg command line and output would be helpful
[05:49:55 CEST] <thebombzen> also, yes, CLI
[05:50:07 CEST] <thebombzen> sure, lemme simplify it so it still reproduces, then drop it here
[05:50:47 CEST] <kepstin> and if you're building your own ffmpeg, you could consider running ffmpeg_g in gdb and getting a backtrace.
[05:54:20 CEST] <Mysterytrain> ok, I have these video files, and separate .sub files, which I find a bit messy. how can I combine these into one file?
[05:54:49 CEST] <kepstin> Mysterytrain: ffmpeg -i video.mkv -i video.sub -map 0 -map 1 -c copy combined.mkv
[05:55:12 CEST] <kepstin> probably, depends on the exact sub format if that'll work
[05:55:32 CEST] <Mysterytrain> ffprobe reports "dvd subtitle" if that changes something
[05:56:08 CEST] <cawk>  -map 0 -map 1 ?
[05:56:26 CEST] <cawk> how can both exist
[05:56:33 CEST] <another> two inputs
[05:56:37 CEST] <kepstin> cawk: that ensures that all streams are copied from the input files (if e.g. the source has multiple audio tracks)
[05:56:54 CEST] <cawk> ok
[05:57:04 CEST] <kepstin> cawk: by default, ffmpeg picks just one video and one audio from each input, this is an override to select all
[05:57:18 CEST] <cawk> kepstin i see
[05:57:59 CEST] <thebombzen> kepstin: issue with zscale, I've pinned it down
[05:58:00 CEST] <thebombzen> ffmpeg -v 9 -loglevel 99 -y -i Voxel-Recording/2019_05_16_23_12_39.mp4 -map v -vf crop=1920:1080:120:0,zscale,format=yuv420p10le -c rawvideo -f nut /dev/null
[05:58:01 CEST] <cawk> Mysterytrain  what subtitle format?  because there are formats that ffmpeg does NOT support
[05:58:21 CEST] <thebombzen> https://0x0.st/zA7A.log
[05:58:21 CEST] <furq> thebombzen: zscale segfaults all the time for me
[05:58:33 CEST] <kepstin> Mysterytrain: cawk dvd subtitles should be fine, and they can be stored in mkv iirc.
[05:58:39 CEST] <furq> i don't think it's an ffmpeg issue
[05:58:49 CEST] <thebombzen> removing the crop filter makes it not segfault
[05:58:58 CEST] <thebombzen> which leads me to believe it might be an issue with the interoperability
[05:59:11 CEST] <cawk> kepstin  dvd subtitles use  bmp format  not text format
[05:59:14 CEST] <furq> iirc i've had it before with non-mod16 stride or something
[05:59:23 CEST] <furq> you should be able to crop after converting
[05:59:25 CEST] <Mysterytrain> it just says:
[05:59:28 CEST] <Mysterytrain>     Stream #0:0[0x20]: Subtitle: dvd_subtitle
[05:59:39 CEST] <thebombzen> in theory yes but doesn't change hte fact that this is a bug
[06:00:26 CEST] <kepstin> cawk: yes, and? that's not a problem. ffmpeg can copy bitmap subtitles between files fine when using -c:s copy.
[06:00:26 CEST] <furq> i googled this and the top match is an irc log of me mentioning it in here
[06:00:28 CEST] <furq> so that's helpful
[06:00:39 CEST] <thebombzen> lul
[06:00:43 CEST] <thebombzen> I love irc logs
[06:00:51 CEST] <cawk> kepstin but is ffmpeg smart enough to do  OCR ?
[06:00:57 CEST] <kepstin> cawk: lol no
[06:01:08 CEST] <cawk> i didn't think so
[06:01:08 CEST] <furq> is ffmpeg smart enough to ride a bicycle
[06:01:16 CEST] <thebombzen> cawk: tesseract is a filter
[06:01:20 CEST] <thebombzen> in libavfilter
[06:01:22 CEST] <Mysterytrain> this .sub file is 4mb, so i'm betting it's images
[06:01:28 CEST] <thebombzen> it performs OCR and stores it as frame metadata
[06:01:31 CEST] <kepstin> cawk: if you attempt to transcode bitmap subtitles to e.g. srt, ffmpeg will give you an error
[06:01:33 CEST] <Mysterytrain> sigh. that wont work for me. I need text subs lol
[06:01:35 CEST] <furq> Mysterytrain: idx/sub is vobsub which is image based
[06:01:47 CEST] <Mysterytrain> oh well
[06:01:52 CEST] <thebombzen> Mysterytrain: you can run it through the tesseract OCR filter in libavfilter
[06:01:56 CEST] <furq> you'll want to use a visual ocr tool to convert it anyway
[06:01:58 CEST] <thebombzen> which stores the OCRed data as frame metadata
[06:02:02 CEST] <furq> the automatic ones aren't very good
[06:02:04 CEST] <thebombzen> you'll have to edit it anyway
[06:02:07 CEST] <Mysterytrain> yep
[06:02:11 CEST] <thebombzen> it'll also dump it into a weird format
[06:02:23 CEST] <thebombzen> but I wrote a perl script to convert it to webvtt if you're interested
[06:02:54 CEST] <kepstin> huh, i didn't think libavfilter could filter subtitle tracks at all. what are you doing, burning them onto a blank video to ocr them?
[06:03:30 CEST] <cawk> ffmpeg -f lavfi -i "movie=test.mkv[out+subcc]" -map 0:1 output.srt    extracts all the text but timing is all wrong:   should i report this as a bug?
[06:04:18 CEST] <kepstin> back when i was doing stuff with dvd subs, i had to do fun things like manually change the palette colours to remove the outlines to improve recognition :/
[06:05:23 CEST] <cawk> what is the difference between idx/sub and .sup
[06:06:47 CEST] <kepstin> 'sup' appears to be some sort of hi-def video subtitle/caption format? maybe used in broadcast tv?
[06:07:10 CEST] <kepstin> or maybe that's the bluray format, i can't remember
[06:07:24 CEST] <Mysterytrain> i love .srt because it's text. I like to be able to follow along on my braille display as the video plays. but .sub obviously I can't do that. heh
[06:07:34 CEST] <cawk> kepstin are you ffmpeg dev?
[06:08:24 CEST] <kepstin> I am a developer, and some of the code I've written has been merged into ffmpeg.
[06:08:35 CEST] <cawk> <JEEB> michael does random stuff, paul does filters, and then you have everyone else doing what they care about
[06:08:44 CEST] <cawk> who is michael?
[06:13:18 CEST] <cawk> thebombzen  why webvtt format?
[06:13:30 CEST] <kepstin> cawk: Michael Niedermayer was the leader of ffmpeg development for a significant period of time, and continues to be involved in developing/maintaining ffmpeg.
[06:14:00 CEST] <cawk> i see, i hoping an answer of freenode name
[06:14:22 CEST] <kepstin> cawk: he should sort to the top of the list, due to having channel operator status ;)
[06:14:45 CEST] <cawk> kepstin i have a feeling you want to be in that top of the list
[06:14:51 CEST] <kepstin> no, not really
[06:15:13 CEST] Action: kepstin had enough of channel operator status back in the bad old days of anime fansub communities ;)
[06:15:20 CEST] <cawk> kepstin why not? you contribute nicely to this #ffmpeg community
[06:15:54 CEST] <cawk> kepstin i have a feeling furq/another wants to be in that top of the list
[06:17:32 CEST] <kepstin> i still consider myself mostly an ffmpeg user, i just use it enough that i can answer a lot of questions other people have, and enough that i've fixed some issues I ran into in my use.
[06:18:18 CEST] <cawk> kepstin are you hardcore user enough to  add   libaac_fdk   on your own  ffmpeg binary?
[06:19:12 CEST] <kepstin> that's kind of a trivial thing to do assuming you're building your own ffmpeg in the first place
[06:19:28 CEST] <kepstin> hilariously i never actually use it, since i use opus for most stuff nowadays
[06:19:41 CEST] <cawk> i see
[06:19:52 CEST] <cawk> not sure if that is yes or no though
[06:20:13 CEST] <cawk> kepstin  what do you think about  Furq and Another  user ?
[06:21:22 CEST] <kepstin> furq has seemed generally knowledgeable about stuff when I've seen them answering questions in chat. I don't interact much with another.
[06:21:36 CEST] <cawk> i see
[06:21:48 CEST] <cawk> is "another" not very friendly to you or something?
[06:22:09 CEST] <another> kepstin: same here regarding fdkaac. i explicitly build it but actually using it? naaah
[06:22:09 CEST] <kepstin> probably mostly time zone differences? lol
[06:22:38 CEST] <cawk> kepstin i see,  well another is here now
[06:22:44 CEST] <another> i
[06:22:52 CEST] <kepstin> time for me to disappear then, g'night ;)
[06:22:59 CEST] <another> i'm actually on my way out
[06:23:07 CEST] <another> o/
[06:23:21 CEST] <cawk> dong are you here?
[07:10:04 CEST] <cawk> what is better for  doing video/ffmpeg?  intel or amd?
[07:27:47 CEST] <DHE> that's not really a fair question. if you're really on the bleeding edge, intel has avx512 which should help a bit. but AMD chips tend to be good prices for lots of cores.
[07:27:56 CEST] <DHE> you can make arguments both ways
[07:29:06 CEST] <cawk> avx512 feature help with ffmpeg ?
[07:29:14 CEST] <cawk> DHE
[07:29:28 CEST] <DHE> the current version of x264 has some avx512 assembly
[07:29:44 CEST] <cawk> i see
[07:30:00 CEST] <DHE> but we're talking percentage points, not miracles
[07:30:04 CEST] <cawk> what avx version does amd support  now?
[07:30:22 CEST] <DHE> ryzen it has avx2
[07:30:38 CEST] <DHE> (and by extension threadripper and epyc)
[07:30:41 CEST] <cawk> is there a avx version  between  avx2 and avx512 ?
[07:32:30 CEST] <DHE> not that I'm aware of, but avx512 is a collection of instruction set sub-families.
[07:46:53 CEST] <bamdads> oin #firefox
[07:47:34 CEST] <cawk> <hmir> is there a parameter for ffmpeg that forces it to use the iGPU for encoding?
[08:38:09 CEST] <cawk> why is it so hard to record  MIC and whatyouhear at the same time in windows OS?  is this windows limitation? or does linux have same problem?
[09:44:24 CEST] <dongs> cawk, lunix has the problem of audio not working there
[09:44:35 CEST] <cawk> huh
[09:44:51 CEST] <dongs> also it lacks kernel audio mixer
[09:44:58 CEST] <dongs> something thats been standard in windows since uh... '98 or so
[09:45:04 CEST] <dongs> so you can't play multiple sounds at once
[09:45:11 CEST] <cawk> what OS do you use now
[09:45:12 CEST] <dongs> without running some userspace garbage written in C++
[09:45:14 CEST] <dongs> windows 10
[09:45:28 CEST] <cawk> you are running irssi on windows 10 ?
[09:45:39 CEST] <dongs> yes, you have a problem with that?
[09:45:47 CEST] <cawk> yes, that's weird
[09:46:18 CEST] <dongs> but yeah to answer your qestion, lunix is not suitable for multimedia work of any kind
[09:46:53 CEST] <cawk> windows is not any better
[09:46:53 CEST] <dongs> no idea about recording 'mic and whatever you hear' in windows, that doesn't sound like it would work anywhere unelss you fed all the outputs and mic(s) into a proper audio interface
[09:46:53 CEST] <kuznetsss> Hello. Is it possible to obtain timestamp for each frame while demuxing rtsp video streaming?
[09:52:10 CEST] <cawk> dongs   okay  record something on the mic and  play audio from youtube right now and record
[09:58:25 CEST] <dongs> cawk, what "mic"?
[09:58:38 CEST] <dongs> i'm a professional youtuber, i don't have a fucking garbage mic plugged into PC
[09:58:48 CEST] <dongs> i have an audio interface / mixer where all the inputs go to
[09:58:50 CEST] <cawk> your youtube channel then
[09:59:46 CEST] <cawk> you still need PC to record
[09:59:50 CEST] <dongs> yes
[09:59:52 CEST] <dongs> and?
[10:00:06 CEST] <cawk> link to your youtube channel then
[10:00:44 CEST] <dongs> in the unlikely scenario when I need to record something PC is playing (and I cannot get it into the mix by another way), i can just feed line out from PC to the mixer and work with it from there.
[10:01:19 CEST] <dongs> but normally i just have a dedicated output channel for pc "audio" thats part of the interface
[10:01:20 CEST] <cawk> link to your youtube channel then
[10:01:32 CEST] <dongs> and that can get mixed to stuff.
[10:01:37 CEST] <dongs> no, fuck off mr.cawk
[10:01:47 CEST] <cawk> you said you are pro youtuber
[10:01:58 CEST] <dongs> yes
[10:01:59 CEST] <cawk> why is it so hard to share your youtube channel
[10:02:13 CEST] <dongs> because you're not the target audience
[10:02:18 CEST] <cawk> lol
[10:02:25 CEST] <cawk> and you would know that how?
[10:03:37 CEST] <cawk> why is it so hard to share your channel
[10:04:30 CEST] <cawk> dongs i don't even care if it's  NSFW
[10:23:16 CEST] <cawk> dongs i exposed you
[10:23:49 CEST] <dongs> "exposed"?
[10:23:55 CEST] <cawk> why is it so hard to share your channel
[10:24:01 CEST] <dongs> < dongs> because you're not the target audience
[10:24:09 CEST] <cawk> you don't know that
[10:24:19 CEST] <cawk> i could speak Japanese for all you know
[10:24:44 CEST] <dongs> yeah, im sure
[10:24:57 CEST] <dongs> why don't you go pump somewhere else?
[10:25:05 CEST] <cawk> pump?
[10:25:15 CEST] <dongs> yeah, what you're doing now
[10:25:29 CEST] <cawk> all you have to do is share your channel
[10:25:36 CEST] <dongs> no, i don't have to do anything
[10:25:45 CEST] <cawk> what are you hiding
[10:27:00 CEST] <cawk> dongs you sure love to bragg that you are a pro youtubers
[10:27:25 CEST] <cawk> what kind of pro youtuber is secretive about their youtube channel
[10:28:18 CEST] <dongs> keep "exposing" me, you're doing great
[10:28:31 CEST] <cawk> and who uses irssi on windows
[10:29:07 CEST] <cawk> dongs point of youtuber is get more and more people in their site
[10:29:25 CEST] <dongs> yes, but you can hardly be considered "people"
[10:29:38 CEST] <cawk> why not
[10:46:55 CEST] <th3_v0ice> Can anyone tell me how are the SwrContext->in_ch_layout (63) and AVFrame->channel_layout (1551), related? In my case they seem to represent the same channel layout, but the values are different. How can I test if the channel layout is the same between the AVFrame and SwrContext?
[10:48:13 CEST] <durandal11707> it is number, compare numbers
[10:51:14 CEST] <th3_v0ice> Maybe I didnt make my self clear. This channel layout is 5.1 audio. And there are different values for 5.1 audio in AVFrame and SwrContext.
[10:52:07 CEST] <th3_v0ice> I can hardcode these two values, but what happens when I get stereo sound is I will have different values from 1551 and 63.
[10:52:32 CEST] <durandal11707> different values, different layouts
[10:55:17 CEST] <th3_v0ice> How so? This AVFrame is actually 5.1, and SwrContext is initialized with a decoder that decodes 5.1 audio
[10:55:54 CEST] <durandal11707> 63 != 1551
[10:56:36 CEST] <th3_v0ice> I totally agree on that. There must be something else that I am missing...
[11:01:14 CEST] <th3_v0ice> DecoderContext->channel_layout 1551. EncoderContext->channel_layout 3, SwrContext->in_ch_layout 63, SwrContext->out_ch_layout 3. How did they get the 63 value?
[11:02:06 CEST] <DHE> the value is a bitfield. the number of bits is identical. could be the difference between 5.1(side) vs 5.1(rear)
[11:02:47 CEST] <DHE> I've seen them sometimes treated as identical for the purposes of resampling. not sure if that's mathematically correct
[11:10:43 CEST] <th3_v0ice> DHE: Are you saying that the number of bits set to 1 is identical between the two?
[11:11:06 CEST] <DHE> yes
[11:11:40 CEST] <th3_v0ice> So that is the only connection between the two? What did You mean by "not sure if that's mathematically correct"?
[11:12:00 CEST] <DHE> 5.1 where the rear speakers are actually behind the listened vs 5.1 where the rear speakers are more to the side of the user
[11:12:06 CEST] <furq> whether it's correct for the resampler to treat 5.1 and 5.1 (side) the same when downmixing
[11:12:07 CEST] <DHE> that's (probbaly?) the difference between the two
[11:12:16 CEST] <DHE> *listener
[11:12:24 CEST] Action: DHE cannot type this morning
[11:13:51 CEST] <th3_v0ice> Ah, ok. Thanks guys for clarifying this.
[11:14:43 CEST] <DHE> yeah audio downmixing was the bane of my life a few years ago... learned more than I thought I'd need to know...
[11:20:45 CEST] <furq> so i checked
[11:20:50 CEST] <furq> 1551 is 5.1 (side) and 63 is 5.1
[11:24:45 CEST] <th3_v0ice> Where did You check this?
[11:26:37 CEST] <furq> https://ffmpeg.org/doxygen/trunk/group__channel__mask__c.html#ga99d0b5bb80534d13a6cc96336cf9076a
[11:27:15 CEST] <furq> i guess i could have also worked my way back through all the nested definitions at the top of the page but this seemed faster
[17:03:26 CEST] <JordiGH> Anyone have a clue why Firefox would say an mp4 file is corrupt and refuses to play it? VLC can handle it fine.
[17:04:09 CEST] <JordiGH> This file: http://inversethought.com/jordi/tmp/huntara.mp4
[17:06:43 CEST] <ritsuka> probably firefox doesn't handle High 4:4:4 Predictive at L4 profile
[17:08:23 CEST] <JordiGH> Recommendation on how to fix it?
[17:08:36 CEST] <ritsuka> use the good old 4:2:0
[17:12:18 CEST] <JordiGH> Thanks, that fixed it!
[17:39:09 CEST] <hans_> ffmpeg -i text000.png -vf scale=320:240 -f png -
[17:39:15 CEST] <hans_> [NULL @ 0000000000492e80] Requested output format 'png' is not a suitable output format
[17:39:33 CEST] <hans_> .. how can i tell it to print the resulting png to stdout ?
[17:40:27 CEST] <cehoyos> -f rawvideo -vcodec png
[17:40:34 CEST] <hans_> right, -f image2pipe
[17:40:40 CEST] <hans_> hmm, or that i guess
[17:41:01 CEST] <hans_> that works too, thanks
[17:49:10 CEST] <hans_> ending ffmpeg with `-` is that equivalent to ending it with `pipe:1` or is there some difference?
[18:09:13 CEST] <cehoyos> I'd say that "-" is the same as "pipe:"
[18:16:35 CEST] <scientes> e-mail verifyications for trac accounts are not being sent out
[18:26:30 CEST] <durandal11707> check spam
[18:27:56 CEST] <scientes> I did
[18:37:47 CEST] <cartman412> Hey everyone, I'm trying to extract keyframes from a video to populate a ranged slider (used to selected start/end of video for trimming). The call works fine on my mac but it fails on both ios and android. Here's the command + output: https://pastebin.com/BtZLBwKv
[18:38:27 CEST] <cartman412> seems -skip_frame nokey is not recognized but I'm not sure why
[18:51:10 CEST] <another> cartman412: your version is pretty old. maybe try a new version
[18:52:00 CEST] <cartman412> yeah its what I feared
[18:59:05 CEST] <another> what video codec do you have in there?
[18:59:52 CEST] <cartman412> you mean this? libavcodec     57. 24.102 / 57. 24.102
[19:00:27 CEST] <cartman412> tbh I'm using a plugin in Unity to run these commands so I'm a bit limited in terms of upgrades and changing stuff in general
[19:00:36 CEST] <another> no. the video codec in your file
[19:14:48 CEST] <cehoyos> skip_frame nokey works with FFmpeg 0.6 here
[19:20:00 CEST] <another> i've seen some commit adding it to hevc in 2017
[19:29:18 CEST] <cartman412> another: I believe its h264 is that what you're asking?
[19:30:34 CEST] <cartman412> "     Stream #0:0(eng), 1, 1/90000: Video: h264 (High), 1 reference frame (avc1 / 0x31637661), yuvj420p(pc, smpte170m, left), 1920x1080 (1920x1088), 1/180000, 21862 kb/s, SAR 1:1 DAR 16:9, 29.93 fps, 120 tbr, 90k tbn, 180k tbc (default)"
[21:04:21 CEST] <tpaper> Hi all! I am using ffmpeg to transcode (and save to disk) an http video stream. I have a small opencv c++ software that detects movements in the stream. My goal is to "pause" the ffmpeg recording when there is no movement... I thougt about sending some POSIX signal to the process but SIGSTOP and SIGCON does not seems to work fine
[21:52:14 CEST] <DHE> that will stop the process, but it will also prevent it from reading from the HTTP stream which will quickly cause unhappy TCP
[00:00:00 CEST] --- Sat May 18 2019