[Ffmpeg-devel-irc] ffmpeg.log.20160719

burek burek021 at gmail.com
Wed Jul 20 02:05:02 CEST 2016


[00:01:12 CEST] <whomp> or, how can i convert a video from 29.97 fps to 30 fps? i believe this will solve the problem, as 44100 is divisible by 30. i just want to slow the video down about .1%
[00:01:47 CEST] <whomp> i see a command to change the speed, but only by a multiplier, which for a repeating fraction like 29.97 / 30, doesn't seem like it'll work
[00:03:34 CEST] <furq> whomp: you can use -shortest to truncate the output file to the shortest input file length
[00:04:28 CEST] <whomp> furq, so i would split this up into its audio and video components, then re-merge with shortest?
[00:05:29 CEST] <furq> the docs say input stream, so it might work with one file
[00:05:35 CEST] <furq> i've only ever used it with separate files
[00:06:20 CEST] <whomp> furq, won't i still have the issue of 44100 not being divisible by 29.97?
[00:06:30 CEST] <whomp> how would one ever get the audio and video to be the same length?
[00:09:26 CEST] <furq> those two things aren't connected at all
[00:11:24 CEST] <whomp> when they aren't the same length, ios gets significant performance issues
[00:28:38 CEST] <esoul> Hello everybody. I am experimenting with screen capture and want to record rawvideo(like yuv420p or rgb5555l2) . It works with tools like Potplayer. BUt I really want to use ffmpeg. But I am getting the "real-time buffer [screen-capture-recorder] [video input] too full" message pretty quick. I cannot find a solution anywhere.
[00:29:08 CEST] <esoul> Is there a dummy solution ?
[00:29:28 CEST] <furq> increase -rtbufsize
[00:29:46 CEST] <esoul> set to 1500M. But is full really quick. Raw video is huge
[00:30:17 CEST] <furq> is your disk keeping up
[00:31:26 CEST] <furq> there's no workaround if the buffer is getting data faster than it can evict it
[00:33:01 CEST] <esoul> ok
[00:33:47 CEST] <esoul> I make some more test
[00:36:30 CEST] <esoul> One more thing. Is there a way to flush the buffer? It flushes itself somehow, but this takes soe time.
[00:37:29 CEST] <c_14> It'll flush itself as fast as your output processes.
[00:42:32 CEST] <esoul> Lets say the bufer is full. I get this message "real-time buffer [screen-capture-recorder] [video input] too full or near too full (76% of size: 1500000000 [rtbufsize parameter])! frame dropped!". Now I press STRG+C to stop the capture because pressing "q" doesn't work as soon as I get this message. After that I cannot start the capture again. I immediately get the "BUffer full" message.
[00:42:55 CEST] <esoul> I always wait some time and suddenly I can capture again.
[00:56:38 CEST] <vbgunz> no matter what I try, one instance of ffmpeg and 2 inputs, 2 outputs OR 2 ffmpeg instances of 1 input and 1 output, I only randomly ever get the microphone and monitor to rarely sync up to be the same length. otherwish I usually have to truncate the monitor recording. surely theres a better way?
[00:57:22 CEST] <vbgunz> basically, I'm trying to record the microphone and monitor, 2 recordings, but they're almost always a different length from one another.
[01:07:19 CEST] <sachin65> Hi
[01:07:46 CEST] <sachin65> I am trying to remove a certain section in a audio file (mp3)
[01:08:24 CEST] <sachin65> like remove from 1.33333 sec to 3.4444 sec
[01:12:03 CEST] <sachin65> any resource or help ??
[03:06:30 CEST] <freaksed> hi all, having a rather perplexing problem with ffmpeg just stopping on my new render boxes
[03:07:24 CEST] <freaksed> we just got 4 rack mount systems from Boxx, running Windows 10 Pro, but when I run a ffmpeg command on them they just stop midway through and hang there
[03:07:52 CEST] <freaksed> the command im running is "ffmpeg -i <input> -r 80 -b 64000k <output>"
[03:08:37 CEST] <freaksed> there is no error produced from this, it just runs, and then at some point it stops updating, the time output stops going up
[03:09:14 CEST] <freaksed> note we are accessing files from a nas, but this has caused no issues on our other systems
[03:13:42 CEST] <freaksed> hi all, having a rather perplexing problem with ffmpeg just stopping on my new render boxes
[03:13:42 CEST] <freaksed> we just got 4 rack mount systems from Boxx, running Windows 10 Pro, but when I run a ffmpeg command on them they just stop midway through and hang there
[03:13:42 CEST] <freaksed> the command im running is "ffmpeg -i <input> -r 80 -b 64000k <output>"
[03:13:42 CEST] <freaksed> there is no error produced from this, it just runs, and then at some point it stops updating, the time output stops going up, i even tried a full trace output but same thing, it just hangs there
[03:13:42 CEST] <freaksed> note we are accessing files from a nas, but this has caused no issues on our other systems
[03:17:59 CEST] <freaksed> aving a really weird problem with ffmpeg on our new render boxes, running windows 10 pro the command is "ffmpeg -i <input> -r 80 -b 64000k <output>" we can run this fine on our desktops but on our new rack systems the process just hangs, cpu goes to 0% and the log shows no errors, i even tried a full trace output but the log just stops updating
[03:47:04 CEST] <fluffrabbit> can someone explain the gradual slowdown that occurs during video encoding?
[03:47:41 CEST] <limbo> cpu heating up, maybe.
[03:48:05 CEST] <fluffrabbit> have you mitigated the problem?
[03:48:44 CEST] <fluffrabbit> I'm using a laptop. ubuntu
[03:49:19 CEST] <fluffrabbit> I guess I could get like an ice pack
[03:50:03 CEST] <fluffrabbit2> woah got kicked off the channel there
[03:51:01 CEST] <fluffrabbit2> test
[03:52:09 CEST] <fluffrabbit> my laptop's underside is only lukewarm to the touch
[03:52:55 CEST] <fluffrabbit> putting it on a platform with an atrium does not seem to improve encoding speed at all
[03:56:02 CEST] <fluffrabbit> this seems to occur with h.264, mpeg-2, and vp9, as well as other codecs I'm sure. with mjpeg it seems to stay fairly consistently fast though and never falls below 11 fps for 1920x1080
[03:57:38 CEST] <fluffrabbit> in other news, CVS caffeine pills are duds
[04:02:17 CEST] <fluffrabbit> I get the feeling that there is a long-standing memory leak that causes this. it's not just ffmpeg, it's a problem in libav as well
[04:09:37 CEST] <fluffrabbit> interestingly, along with the time to encode each frame, the bitrate is also steadily increasing.
[04:13:04 CEST] <fluffrabbit> a solution would be to cap the bitrate, but the wiki recommends the crf option as a surefire way toget good quality, which is what I'm using. afaik crf stands for "CONSTANT rate factor", not "cumulative rate factor"
[04:20:22 CEST] <fluffrabbit> it's a great way to improve the viewing experience of videos: the quality gets better towards the end /sarcasm
[04:43:01 CEST] <fluffrabbit> conspiracy?
[05:48:40 CEST] <kiko> hi
[05:49:29 CEST] <kiko> This've been a long time in IRC chat.
[05:50:23 CEST] <kiko> Is anyone here?
[06:11:40 CEST] <kriskropd> whats the easiest way for an idiot to make a stereo mp3 mono?
[06:14:56 CEST] <kriskropd> i should mention - without muting one side of the stereo
[06:15:22 CEST] <kriskropd> im trying to merge both stereo sides into a single channel - but my eyes are glazing over trying to figure it out from man
[06:16:23 CEST] <kriskropd> nvm - figured it out by googling "mixdown"
[06:16:25 CEST] <kriskropd> -ac 1
[06:16:49 CEST] <kriskropd> just the idiot solution i needed :D thanks guys - you are awesome - pat yourselves on the back and have a hug
[08:27:33 CEST] <CoJaBo> lol.
[11:03:58 CEST] <shams> Hello all
[11:04:27 CEST] <shams> i'm having an issue running with ffmpeg
[11:05:42 CEST] <shams> i have its binary file in /opt/ffmpeg/bin/ffmpeg but when i execute it with shell_exec() it throws an error permission denie for /opt/ffmpeg/bin/
[11:06:05 CEST] <shams> any here have idea?
[11:10:23 CEST] <durandal_1707> missing permissions?
[11:11:04 CEST] <shams> i tried to set its permission to chmod 755 /opt/ffmpeg/bin/
[11:11:11 CEST] <shams> but still got the same error
[11:12:54 CEST] <iive> shams: make sure /opt is not on volume mounted with noexec.
[11:13:13 CEST] <iive> strace might help you to find the exact operation that returns the error.
[11:13:25 CEST] <shams> iive: what do you mean?
[11:13:40 CEST] <shams> can you guide how can i check it?
[11:14:32 CEST] <iive> `mount` might report it, or `cat /proc/mounts`
[11:16:29 CEST] <shams> iive: no, not listed under cat /proc/mounts
[11:17:18 CEST] <iive> your '/' root might be mounted with noexec
[11:17:55 CEST] <__jack__> (which is creepy)
[11:18:35 CEST] <shams> what's the solution?
[11:23:47 CEST] <iive> shams: do you see `noexec` anywhere in the mounts?
[11:24:41 CEST] <shams> iive: yes
[11:24:53 CEST] <shams> sysfs /sys sysfs rw,nosuid,nodev,noexec,relatime 0 0
[11:24:56 CEST] <shams> proc /proc proc rw,nosuid,nodev,noexec,relatime 0 0
[11:25:15 CEST] <shams> devpts /dev/pts devpts rw,nosuid,noexec,relatime,gid=5,mode=620,ptmxmode=000 0 0
[11:25:28 CEST] <shams> tmpfs /run tmpfs rw,nosuid,noexec,relatime,size=5359196k,mode=755 0 0
[11:25:39 CEST] <shams> on above 4 places
[11:26:14 CEST] <iive> these are all virtual filesystems, and none is root for /opt/
[11:26:20 CEST] <iive> so the problem is different.
[11:26:41 CEST] <shams> yep
[11:26:58 CEST] <__jack__> chmod 755 /opt/ffmpeg/bin/ -> chmod -R 755 /opt/ffmpeg/bin/ (or chmod 755 /opt/ffmpeg/bin/ffmpeg)
[11:27:26 CEST] <shams> already did it
[11:30:56 CEST] <shams> Hello Fletcher
[11:31:17 CEST] <shams> i hope you can help me with an issue
[11:31:46 CEST] <shams> i'm having an issue running with ffmpeg
[11:31:52 CEST] <shams> i have its binary file in /opt/ffmpeg/bin/ffmpeg but when i execute it with shell_exec() it throws an error permission denie for /opt/ffmpeg/bin/
[11:32:17 CEST] <iive> shams: does it work if you start it from a shell ?
[11:32:30 CEST] <iive> i mean, text console with bash
[11:32:42 CEST] <shams> i'm trying it with shell at this time
[11:32:49 CEST] <shams> yes, doing it with SSH
[11:32:51 CEST] <__jack__> shams: namei -lo /opt/ffmpeg/bin/
[11:34:19 CEST] <shams> @jack: result
[11:34:26 CEST] <shams> f: /opt/ffmpeg/bin/ drwxr-xr-x
[11:34:32 CEST] <shams> root root / drwxr-xr-x
[11:34:40 CEST] <shams> root root opt drwxr-xr-x
[11:34:45 CEST] <shams> root root ffmpeg drwxr-xr-x r
[11:34:47 CEST] <shams> oot root bin
[11:35:07 CEST] <shams> this is the result of namei comm.
[11:42:07 CEST] <__jack__> shams: you may try strace: strace -ft /opt/ffmpeg/bin/ffmpeg
[11:52:27 CEST] <iive> i suspect SE linux might be involved... no idea how to fight that.
[11:53:10 CEST] <iive> maybe `dmesg | tail` would contain something useful
[11:57:53 CEST] <treaki__> hi
[11:58:24 CEST] <treaki__> once i have stumpled by accident accross a page of fmmpeg with a lot of sample files
[11:58:32 CEST] <treaki__> to test with (manny formats)
[11:59:01 CEST] <treaki__> now i am looking for a .sac audio file to test with, and cant find this page anymore
[11:59:05 CEST] <treaki__> can you help me out?
[11:59:20 CEST] <c_14> samples.ffmpeg.org probably
[11:59:48 CEST] <treaki__> thanks
[11:59:55 CEST] <treaki__> wasnt able to find that by google
[12:00:19 CEST] <BtbN> It's the first google hit for ffmpeg samples
[12:01:31 CEST] <treaki__> i have fund manny thinks about setting sample rate, maybe i had a typo or my search request was to specific...
[12:01:34 CEST] <treaki__> thanks for the help
[12:02:36 CEST] <treaki__> but looks like no sac;(
[12:02:53 CEST] <treaki__> https://www.google.de/search?q=.sac+site:https://samples.ffmpeg.org/&ie=utf-8&oe=utf-8
[12:04:07 CEST] <c_14> treaki__: ./archive/audio/sac3 ?
[12:05:38 CEST] <treaki__> ah, thanks
[12:39:57 CEST] <haarts> I'm using http://sprunge.us/gagT to convert a MKV to webM. But the resulting file is 20x larger than the original (1:00 of video, 3.2M). How come?
[12:40:25 CEST] <haarts> Ultimately I want to convert whatever format I have into WebM.
[12:40:35 CEST] <c_14> because you're encoding it losslessly
[12:41:03 CEST] <c_14> eh, I don't think you can use -lossless and -b:v...
[12:41:19 CEST] <c_14> I think you meant -b:a for the audio stream
[12:42:57 CEST] <haarts> aha! That might be the bug which has been plagueing me for so long. I got such shitty quality earilier.
[12:47:17 CEST] <haarts> hmm, small file now! That's good. But shitty quality. Do these settings make sense, I want to preserve quality as much as possible? http://sprunge.us/Begc
[12:47:46 CEST] <c_14> set a crf
[12:47:54 CEST] <c_14> you didn't which is why the quality is low
[12:48:28 CEST] <furq> it says -crf 33
[12:48:52 CEST] <c_14> Ah, I'm just blind then
[12:48:59 CEST] <furq> it doesn't in the first one
[12:49:08 CEST] <__jack__> 33 is high
[12:49:11 CEST] <c_14> If you want constant quality, you need -b:v 0 though
[12:49:13 CEST] <__jack__> doc said: The CRF value can be from 063. Lower values mean better quality.
[12:49:27 CEST] <c_14> Otherwise I think it'll default to constrained quality with a relatively low bitrate
[12:49:31 CEST] <furq> it has that too
[12:49:42 CEST] <c_14> Well
[12:49:47 CEST] <furq> haarts: reduce the crf value until it's good enough
[12:49:51 CEST] <c_14> Maybe I should learn to read.
[12:50:12 CEST] <haarts> furq: trying 3 now. (binary search :))
[12:50:24 CEST] <furq> that's almost certainly going to be too high but ok
[12:50:49 CEST] <furq> surely a binary search would lead you to 17
[12:50:53 CEST] <haarts> file gets big again. 3.2 orig vs 36 recoded. Too high for sure. :)
[12:51:07 CEST] <haarts> that would be the next step.
[12:51:50 CEST] <haarts> is that the only truly relevant param to tweak?
[12:51:59 CEST] <furq> that and -speed
[12:52:22 CEST] <furq> i assume -speed will reduce filesize and/or increase quality at the same crf
[12:52:58 CEST] <furq> also libvpx's multithreading will reduce quality a little bit, but chances are it's not doing anything anyway
[12:53:35 CEST] <__jack__> increase the crf, there are many values between 3 and 33 :p
[12:54:24 CEST] <furq> you can also try dropping to 1-pass if it's too slow
[12:54:27 CEST] <c_14> haarts: only use -frame-parallel if you're planning on decoding the video with libvpx, it'll actually only hurt quality and decoding speed when decoded by ffmpeg
[12:54:42 CEST] <furq> frame-parallel isn't needed for multithreading any more is it
[12:54:53 CEST] <furq> i think you just need tile-columns and threads now
[12:55:09 CEST] <haarts> 17 looks good. file doubled in size. Would I have to do this tweaking with every different input file type?
[12:55:17 CEST] <c_14> It does something to make it easier to decode with libvpx
[12:55:21 CEST] <furq> 17 is nominally constant quality, so no
[12:55:32 CEST] <furq> pick a value that looks good and it should be the right choice for most inputs
[12:55:51 CEST] <furq> s/17/crf/
[12:55:55 CEST] <haarts> excellent
[12:56:15 CEST] <furq> or at least it is with x264, i've still not got used to the different mappings with libvpx
[12:56:35 CEST] <furq> given that 2-pass crf actually has an effect with libvpx, it's probably not safe to assume much is the same
[12:58:07 CEST] <haarts> it is expected that the file size gets significantly bigger? I'm lacking a frame of reference
[13:14:34 CEST] <Primal> hello! can anyone please explain how to update ffmpeg? we currently have v.2.4.3 installed. running on apache. access through putty (ssh).
[13:15:33 CEST] <__jack__> huh ? what's the relation between apache, ffmpeg and ssh ?
[13:16:48 CEST] <Primal> i thought that info would be necessary.
[13:17:02 CEST] <Primal> i am assuming i was mistaken :)
[13:30:03 CEST] <iuppiter> wow
[13:30:20 CEST] <iuppiter> AMD ZEN will have 10bit decoding!!
[13:32:05 CEST] <haarts> Primal: what operating system?
[13:34:14 CEST] <Primal> for me, windows
[13:34:42 CEST] <haarts> That I gathered from having Putty. But the system you SSH into.
[13:35:30 CEST] <Primal> linux
[13:35:40 CEST] <haarts> what kind of linux?
[13:37:06 CEST] <Primal> 2.6.32-042stab108.2 GNU/Linux
[13:37:40 CEST] <haarts> close. and what variant? Ubuntu? CentOS? Try 'cat /etc/lsb-release'
[13:38:18 CEST] <Primal> haha, ubuntu 14.04
[13:38:31 CEST] <Primal> Ubuntu 14.04.4 LTS
[13:39:03 CEST] <furq> http://johnvansickle.com/ffmpeg/
[13:39:12 CEST] <furq> extract that to /usr/local/bin
[13:39:52 CEST] <haarts> should he just use apt?
[13:40:04 CEST] <furq> there is no ffmpeg in 14.04 or in backports
[13:40:11 CEST] <haarts> not sure what the latest and greatest is on Ubuntu.
[13:40:14 CEST] <haarts> Wait. Wat?
[13:40:21 CEST] <furq> you'd have to use a ppa and those all look crap
[13:40:34 CEST] <furq> the most notable one was violating gpl until very recently, if not still
[13:40:45 CEST] <haarts> omg
[13:40:52 CEST] <haarts> OK, what furq said
[13:41:17 CEST] <furq> ffmpeg was gone from ubuntu/debian for a couple of years because of the whole libav situation
[13:41:37 CEST] <furq> fortunately no ubuntu or debian user is using a distro that's several years old
[13:41:53 CEST] <furq> he said, before bursting into tears
[13:41:59 CEST] <haarts> lol
[13:42:25 CEST] <haarts> what are you running on your server (if any)? Me Ubuntu 16.04.
[13:42:30 CEST] <furq> debian testing
[13:43:13 CEST] <haarts> in 16.04 ffmpeg is there, v2.8
[13:43:32 CEST] <furq> it sure is
[13:43:53 CEST] <furq> but people on LTS releases usually don't take kindly to you suggesting they upgrade four times
[13:43:59 CEST] <furq> not least because it usually doesn't work
[13:47:05 CEST] <Primal> so is what you're saying, for a layman like me, that i might as well not bother upgrading?
[13:47:36 CEST] <furq> no
[13:47:37 CEST] <furq> 12:39:03 ( furq) http://johnvansickle.com/ffmpeg/
[13:47:37 CEST] <furq> 12:39:12 ( furq) extract that to /usr/local/bin
[13:47:41 CEST] <furq> that's what i'm saying
[13:47:50 CEST] <Primal> haha, ok :)
[13:48:22 CEST] <drastik> ffmpeg -i https://streamsite/playlist.m3u8 -c copy -bsf:a aac_adtstoasc -f mp4 out.mp4
[13:48:25 CEST] <furq> you probably also want to remove the ffmpeg you already have installed so it doesn't conflict
[13:48:40 CEST] <furq> i guess you installed it from a ppa, in which case `apt-get purge ffmpeg`
[13:49:41 CEST] <drastik> hi everyone i am running that command above to download a stream from a website. The problem is that ffmpeg doesnt stop when the ppl on the site stop streaming, but instead it waits for me to type q to quit
[13:50:03 CEST] <drastik> anyway idea how to make ffmpeg stop on it own?
[13:56:23 CEST] <drastik> anyone please?
[14:02:06 CEST] <iuppiter> I've got a video and want to trim the black borders, is this doable?
[14:03:57 CEST] <furq> https://ffmpeg.org/ffmpeg-filters.html#crop
[14:04:20 CEST] <drastik> any idea for my problem?
[14:04:31 CEST] <furq> drastik: which ffmpeg version
[14:04:40 CEST] <drastik> lemme check
[14:05:03 CEST] <furq> i just tested and it stops for me when the m3u8 is no longer available
[14:05:12 CEST] <furq> if the streaming server you're using doesn't do that then i have no idea
[14:05:25 CEST] <drastik> wait
[14:05:38 CEST] <furq> you could maybe try -timeout but that didn't seem to have any effect for me
[14:05:56 CEST] <drastik> ffmpeg version 3.0.2-4~bpo8+1
[14:06:05 CEST] <furq> probably not that then
[14:06:22 CEST] <drastik> it stay in this state for the whole night
[14:06:32 CEST] <drastik> i had to manually stop it
[14:06:54 CEST] <furq> i suspect the hls muxer isn't smart enough to figure it out
[14:07:06 CEST] <drastik> but how to fix it?
[14:07:18 CEST] <drastik> what would the timeout do?
[14:07:29 CEST] <furq> you can try -timeout 10 before -i
[14:07:34 CEST] <furq> but like i said, i don't think it does anything
[14:07:41 CEST] <drastik> i c
[14:07:48 CEST] <drastik> so any other suggestions?
[14:08:00 CEST] <furq> none better than filing a feature request
[14:08:08 CEST] <drastik> i c ty
[15:29:53 CEST] <zyme> Dumb question, you have a mkv file, DD5.1 H.264 4.1 720P; that always freezes on the exact same scene, not on all players but one you need to use it with and several others. You can skip past this point and resume right after if you watch closely.
[15:31:47 CEST] <zyme> Are there any programs around to check the file for encoding errors and easily find/ attempt to fix the file or cut the bad frames out?
[15:33:18 CEST] <DHE> doesn't work that way. if there are corrupted video frames that can't be repaired (and very few people use FEC) then you have to drop everything up to the next keyframe
[15:46:32 CEST] <zyme> Is there a way to check for corrupt video frames or find out if it's just the decoder?
[15:48:31 CEST] <xx_pk_xx> Hey guys. I'm trying to encode bitmap subtitles. However, I'm not sure what data I need as input for AVSubtitle.data. Let's say I have an image of the subtitle. To which pix format should I transform it? I will be greatfull for any advice
[16:22:42 CEST] <abumusa> where can I find an example of code that turn a sequence of images into webm file ?
[17:27:48 CEST] <xrmx> hello, does anyone have a clue about how to debug Assertion ret >= 0 failed at src/libavcodec/options.c:142 ? http://paste.debian.net/783181/
[17:47:43 CEST] <DHE> xrmx: I'm guessing you specified "1.0" in a field that only accepts integer inputs
[18:07:16 CEST] <xrmx> DHE, that's me trying to play a video from a video player, like i have no clue where this "1.0" is :)
[18:15:01 CEST] <roxlu> Hi, I just noticed that when i stream a MP4 file (video only) to RTP, that ffmpeg does not send the PPS/SPS as the first packet. Is there a way to tell ffmpeg to first send the PPS/SPS before any other packet?
[18:18:19 CEST] <roxlu> When I use the testsrc filter, I do receive a pps/sps as the first rtp-packet
[18:28:02 CEST] <DHE> testsrc isn't an encoded video
[18:45:49 CEST] Action: DelphiWorld neighboring kepstin somewhere!
[18:46:09 CEST] Action: kepstin waves
[18:49:32 CEST] <DelphiWorld> kepstin: LOL. got me? :P
[19:23:17 CEST] <blb> how does one deal with the VBI lines in IMX MXF files? I know the MXF container has a vertical offset field, but it appears to be ignored
[19:27:34 CEST] <xx_pk_xx> Hey guys. I'm trying to encode bitmap subtitles. However, I'm not sure what data I need as input for AVSubtitle.data. Let's say I have an image of the subtitle. To which pix format should I transform it? I will be greatfull for any advice
[19:47:31 CEST] <JEEB> blb: oh there's a field for that?
[19:47:47 CEST] <JEEB> I think all of the samples I've seen with the non-visual data on top didn't have anything
[19:48:26 CEST] <JEEB> container cropping is a thing now so if it's sane it could be supported if it exists
[19:50:47 CEST] <JEEB> oh, YOsiz
[19:50:49 CEST] <blb> JEEB: using MediaInfo, I see 0x320A for horizontal offset and 0x320B for vertical offset. with this particular file, the offset being 16
[19:51:58 CEST] <blb> 0x320A and 0x320B are ignored by mxf demuxer, afaik
[19:52:59 CEST] <JEEB> I might poke at it in the nearby future if I have the time. need to check the samples I have of that painful format
[19:53:10 CEST] <blb> maybe i haven't seen how a container can provide cropping info to the upper api
[19:58:10 CEST] <JEEB> I will have to look into it a bit more but there's at the very least the difference between codecpar's width/height and coded_width/height
[19:58:51 CEST] <blb> JEEB: neat. i appreciate it
[19:59:35 CEST] <JEEB> mostly I'm poking the guy behind cropping if there's a definition for the cropping box
[19:59:40 CEST] <JEEB> outside of the H.264 decoder that is
[20:03:22 CEST] <blb> yes. that's the one part i'm curious about. should the decoder pay attention to that or should the user?
[20:03:46 CEST] <JEEB> lemme post a short discussion I just had
[20:04:10 CEST] <JEEB> http://up-cat.net/p/7c10f9cd
[20:04:23 CEST] <JEEB> unfortunately container cropping is currently limited to the bottom
[20:05:51 CEST] <blb> dang :/
[20:06:24 CEST] <JEEB> then there's the whole side data business but then support for that would have to be added to ffmpeg cli as well
[20:06:45 CEST] <JEEB> so you add cropping side data and then ffmpeg cli would read the values there
[20:06:51 CEST] <JEEB> and crop accordingly
[20:06:55 CEST] <JEEB> or another app like player
[20:07:22 CEST] <JEEB> ...or you look at how the AVC and HEVC decoders do it
[20:07:33 CEST] <JEEB> although at the demuxer level you don't have the decoded picture yet
[20:07:34 CEST] <JEEB> meh
[20:09:05 CEST] <blb> i'm technically using the ffmpeg api. i can decode the full 720x608 frame. after the decoding, if i knew that i should ignore the top 32 lines, i'd be fine, but that info isn't made available from the demuxer
[20:09:33 CEST] <blb> the mxf field technically says 16, but it's per-field
[20:09:39 CEST] <JEEB> hah! :D
[20:09:55 CEST] <JEEB> but yeah, then side data would be OK for you as well
[20:10:28 CEST] <blb> if the decoder did it, that'd be awesome, if not, that's ok. just as long as i could access it
[20:10:56 CEST] <JEEB> seems like adding a type of side data would be the least bad way currently
[20:11:17 CEST] <JEEB> but the Father Of Side Data says it might be more work than I'd be willing to do for it
[20:11:28 CEST] <JEEB> http://up-cat.net/p/6c7dade3
[20:12:11 CEST] <blb> mxf is one of those containers that really likes to override what the codec says
[20:12:19 CEST] <JEEB> tell me about it :P
[20:12:20 CEST] <blb> i've seen aspect ratio and frame rate overrides
[20:12:23 CEST] <blb> argh!
[20:12:31 CEST] <JEEB> oh, I was just about to ask about aspect ratio
[20:12:43 CEST] <JEEB> because I had a case of some vendor always outputting 4:3 AR'd stuff
[20:12:51 CEST] <JEEB> while the content could randomly be 4:3 or 16:9
[20:12:52 CEST] <blb> aspect ratio is correct i believe, the container reads that field
[20:13:25 CEST] <JEEB> ok, then it's just a case of even crappier stuff :P because the AR flag in the video stream at least was *always* 4:3
[20:14:11 CEST] <JEEB> and the aspect ratio was meant to be applied *after* the 32 samples were cropped off
[20:14:49 CEST] <blb> btw, if you want to see a file like that i'm talking about, here's one https://gstreamer.freedesktop.org/data/media/incoming/mxf/opencubetech.com/xdcam-pal-d10-imx50.mxf
[20:15:09 CEST] <JEEB> cheers
[20:15:14 CEST] <blb> media info shows it should be 720x576 with the original height being 608
[20:15:27 CEST] <JEEB> yeah, that sounds like the stuff I had seen as well
[20:15:45 CEST] <JEEB> the aspect ratio part intrigues me because I distinctly remember that thing not being in the video or container correctly
[20:15:45 CEST] <blb> vlc plays it with the VBI lines showing
[20:16:24 CEST] <blb> i do have a file where mxf says it should be 16:9 but the video data said 4:3
[20:17:08 CEST] <JEEB> oh, and let's not forget about audio mappings
[20:17:18 CEST] <blb> :S nothx
[20:17:47 CEST] <JEEB> yeah, this looks like what I saw
[20:17:53 CEST] <JEEB> doesn't seem to have AR data on MXF side
[20:17:58 CEST] <JEEB> and 4:3 AR on video
[20:19:54 CEST] <JEEB> but yeah, audio seems to be a whole separate mess. so far I've just done selection under "if MXF and either channel count is something or stream count is something"
[20:20:50 CEST] <blb> separating the channels into streams... argh
[20:21:25 CEST] <JEEB> and then I'm not even sure if there's metadata in there which would tell me how many channels are actually used :P
[20:21:41 CEST] <JEEB> currently it's pretty much case-by-case kind of thing for me
[20:25:04 CEST] <madprops> hi
[20:25:18 CEST] <madprops> im trying to create a single big image from 6 long images
[20:25:22 CEST] <madprops> i want to stack them vertically
[20:25:34 CEST] <madprops> this is what im trying
[20:25:35 CEST] <madprops> ffmpeg -i 1.png -i 2.png -i 3.png -i 4.png -i 5.png -i 6.png -filter_complex vstack output.jpg
[20:25:44 CEST] <madprops> but it only seems to merge the first 2 and ignores the rest
[20:25:47 CEST] <madprops> how can I do this?
[20:26:51 CEST] <DHE> vstack=inputs=6
[20:26:57 CEST] <DHE> instead of just "vstack"
[20:34:19 CEST] <madprops> DHE, I tried ffmpeg -i 1.png -i 2.png -i 3.png -i 4.png -i 5.png -i 6.png -filter_complex vstack=inputs=6 output.jpg
[20:34:30 CEST] <madprops> I got Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height
[20:34:58 CEST] <madprops> the last image has a different height than the other ones
[20:35:11 CEST] <madprops> don't know if that has something to do
[20:39:11 CEST] <furq> it shouldn't
[20:39:17 CEST] <furq> All streams must be of same pixel format and of same width.
[20:39:42 CEST] <madprops> they are
[20:40:08 CEST] <furq> you might need -pix_fmt yuvj420p
[20:40:28 CEST] <furq> actually if it worked with two images then probably not
[20:50:27 CEST] <kepstin> madprops: can you pastebin the *complete* ffmpeg output please?
[20:50:58 CEST] <kepstin> (when ffmpeg prints that 'maybe incorrect parameters' message, it usually has printed earlier what the *actual* error is)
[20:51:09 CEST] <madprops> sure
[20:51:47 CEST] <madprops> http://gsick.com/paste/110/
[20:52:18 CEST] <furq> [mjpeg @ 0000000000567400] [IMGUTILS @ 000000000022f000] Picture size 1920x156476 is invalid
[20:52:21 CEST] <furq> nice
[20:52:52 CEST] <furq> the max resolution for jpeg is 65,535
[20:53:48 CEST] <madprops> oh...
[21:24:23 CEST] <DHE> they're actually 30,000 pixels tall? didn't realize that..
[21:55:18 CEST] <shincodex> full human body image to scale maybe
[00:00:00 CEST] --- Wed Jul 20 2016


More information about the Ffmpeg-devel-irc mailing list