[Ffmpeg-devel-irc] ffmpeg.log.20161208

burek burek021 at gmail.com
Fri Dec 9 03:05:01 EET 2016


[00:21:45 CET] <Phi> well, I'm gonna sleep, anywho
[00:21:49 CET] <Phi> ta ta
[02:41:23 CET] <autofsckk> i want to put a png file over an animated gif, but the png is very big, is there a way to resize it while doing this?  i did this but i get this jjajaja ffmpeg -i dreams-reality.gif -i ~/00ttnk/logo/logo.png -filter_complex "overlay=5:main_h-overlay_h" -codec:a copy dreams-reality-output.gif i would like to put the png in the lower right cornet, any ideas?
[02:42:02 CET] <autofsckk> and i see that it changed the quality of the gif too, it doesnt look as sharp as the original
[02:44:45 CET] <autofsckk> this one is the orignal http://imgur.com/a/rQNBq and this one http://imgur.com/a/4ugW6  what i just did
[03:26:14 CET] <CoJaBo> autofsckk: Any specific reason it has to be GIF? It's a very poor format for animations
[03:27:17 CET] <autofsckk> CoJaBo:  im just expermienting with some images used at social pages
[03:28:02 CET] <autofsckk> but really i would like to really learn to use ffmpeg, its so amazing but i havent had time to learn it
[03:28:48 CET] <CoJaBo> See if it accepts webm. You can't really avoid significant generation loss with GIFs
[03:29:51 CET] <autofsckk> ok, but how do you resize a png before making the output file?
[03:32:13 CET] <autofsckk> this is great :D http://www.youtube.com/watch?v=MPV7JXTWPWI
[03:33:13 CET] <autofsckk> for begginers like me, of course :D
[05:28:04 CET] <chocolate-elvis> hey! what's the latest with ST-2084, Dolby Vision HDR?  any support?
[05:36:43 CET] <TheMaskedMeowth> hi hi, i've got a question
[05:37:37 CET] <TheMaskedMeowth> i am kinda not that familiar with this program but i tried setting the path to /bin, but there is no /bin folder in the thing i downloaded
[05:38:54 CET] <TheMaskedMeowth> in fact now that i look at it
[05:39:08 CET] <TheMaskedMeowth> the thing i downloaded doesn't even have an .exe file to run
[05:40:22 CET] <DHE> so what did you download?
[05:41:07 CET] <TheMaskedMeowth> https://ffmpeg.org/download.html the big ol' button on this page
[05:42:04 CET] <DHE> that would be source code. you can run "sh configure" in the directory to prepare it, then "make" to build it.
[05:42:22 CET] <DHE> it's pretty common for unix software
[05:43:00 CET] <DHE> more info is highly googlable
[06:04:23 CET] <TheMaskedMeowth> i'm still quite confused
[06:04:28 CET] <TheMaskedMeowth> i'm sorry if i'm sounding dumb
[06:04:40 CET] <TheMaskedMeowth> but it's telling me configure's not a recognized file
[06:04:54 CET] <TheMaskedMeowth> err rather
[06:04:55 CET] <TheMaskedMeowth> sh isn't
[06:24:49 CET] <TheMaskedMeowth> sorry i left
[07:53:05 CET] <cnote> is there a freenode channel for mkvtoolnix
[08:21:17 CET] <llamapixel> cnote: https://mkvtoolnix.download/contact.html  you could of googled that one mate
[08:51:07 CET] <thebombzen> autofsckk: you need to use the scale2ref filter.
[08:57:39 CET] <thebombzen> such as, ffmpeg -i input_video -i overlay.png -lavfi "[0:v] [1:v] scale2ref [b][a]; [a][b] overlay" output_options
[08:58:28 CET] <thebombzen> oops, my bad
[08:59:06 CET] <thebombzen> I mean ffmpeg -i input_video -i overlay.png -lavfi "[1:v] [0:v] scale2ref [b][a]; [a][b] overlay" output_options
[08:59:32 CET] <thebombzen> the first thing you feed to scale2ref is the thing you're scaling and the second input is the "reference"
[09:00:15 CET] <thebombzen> the reference is passed through unchanged the base is scaled. then it outputs [b] as the scaled base and [a] as the reference (because that's the order used). then it overlays [b] onto [a]
[11:11:26 CET] <booterror> How do you audio stream mp3 files?
[11:31:06 CET] <azatuni> Hello, I have a problem with  ffmpeg, I am pushing stream by udp to my server and trying to encode and sent it to media server. I can see stream by vlc, but on the server ffmpeg freezs
[11:31:29 CET] <azatuni> when I am trying debug mode
[11:32:55 CET] <azatuni> http://pastebin.com/5xtQ9L3z
[11:35:12 CET] <azatuni> http://pastebin.com/Lb0Td4yV
[12:10:58 CET] <l9> what cpu stats do i need too transcode streams, i have a i7 930 cpu running on a freebsd host.
[12:11:55 CET] <superware> can someone please (please, please) take a look at https://trac.ffmpeg.org/ticket/5615 ? it's easy to reproduce so maybe it's easy to fix.
[14:01:57 CET] <JoshX> Hi! I don't get it... could anybody look at this:
[14:01:58 CET] <JoshX> http://townsville.nl/ffmpeg.txt
[14:02:31 CET] <JoshX> it shows a file i have, it should be 900 seconds at 30 fps..
[14:02:44 CET] <JoshX> it dropped some frames, but okay...
[14:03:15 CET] <JoshX> then when i demux the h264 stream from it using the command in the file.. the framerate of the h264 file is 25fps
[14:03:20 CET] <JoshX> why??
[14:06:12 CET] <furq> JoshX: h264 elementary streams don't have a framerate
[14:06:17 CET] <furq> set -r 30 before -i
[14:06:45 CET] <JoshX> lets try that
[14:07:26 CET] <JoshX> then too, could i get ffmpeg to recode my file to be lets say 25fps and exactly 22500 frames long from this h264 source?
[14:07:38 CET] <furq> i'm actually a bit surprised that works, i always thought you needed an external tool to do that
[14:07:45 CET] <furq> so maybe it does something wrong i'm not seeing
[14:08:31 CET] <JoshX> ffmpeg -r 30 -i in.mp4 -vcodec copy -an -f h264 out.h264 will give a h264 file that reports 25 fps as well
[14:08:53 CET] <furq> -r 30 -i in.h264
[14:09:13 CET] <furq> it won't work if the stream is already in a container
[14:09:20 CET] <JoshX> okay that gives me a mp4 file that is longer
[14:10:54 CET] <JoshX> okay now I have an mp4 file with the correct length and framerate so it seems
[14:11:13 CET] <JoshX> let me test
[14:37:47 CET] <JoshX> hmm I think i can detect where my gaps are in the files, could i fill those 'missing' frames with black frames?
[14:37:59 CET] <JoshX> or would that be quite impossible?
[14:38:25 CET] <JoshX> boxdumper tells me
[14:38:26 CET] <JoshX>   sample_delta = 3010
[14:38:26 CET] <JoshX>                                 sample_delta = 51002
[14:38:26 CET] <JoshX>                                 sample_delta = 12000
[14:38:26 CET] <JoshX>                                 sample_delta = 5995
[14:38:29 CET] <JoshX>                                 sample_delta = 3012
[14:39:29 CET] <JoshX> these are the odd ones, there you can see i loose about 20 frames
[14:39:58 CET] <JoshX> (16 + 3 + 1 'too much')
[14:41:32 CET] <JoshX> so i know i need to insert 16 blank frames at
[14:41:33 CET] <JoshX> 31589                             entry[10468]
[14:41:33 CET] <JoshX> 31590                                 sample_count = 1
[14:41:33 CET] <JoshX> 31591                                 sample_delta = 51002
[14:41:44 CET] <JoshX> frame offset 10468
[14:42:02 CET] <JoshX> but would that be doable and not taking a very long time?
[14:42:53 CET] <JoshX> in an ideal world I need all my files to be perfect, all 900seconds * 30 frames per second = 27000 frames per file
[14:43:08 CET] <JoshX> so i can overlay them on top of eachother frame by frame
[14:47:00 CET] <JoshX> perhaps the timelineeditor from l-smash could help but documentation is virtually non-existent
[16:52:45 CET] <kerio> mpv started dropping frames :\
[16:53:14 CET] <kerio> not sure why
[18:50:01 CET] <pgorley> hi all, is there an easy way to find out when a certain function was deprecated? trying to ifdef some things in my application so we use the new apis but still support the old ones
[18:51:26 CET] <c_14> git blame
[18:51:55 CET] <pgorley> thanks
[18:59:48 CET] <BtbN> you should check for when the replacement function was added
[18:59:52 CET] <BtbN> and then version check for that
[18:59:58 CET] <BtbN> And not for when it was deprecated
[20:48:12 CET] <Phi> yo folks
[20:48:49 CET] <Phi> Can someone explain the difference between QSV hardware accel, and the h264_qsv encoder?
[20:49:49 CET] <BtbN> what?
[20:50:42 CET] <Phi> --enable-hwaccel=h264_qsv vs --enable-encoder=h264_qsv
[20:51:26 CET] <Phi> there's two methods of specifying QSV, and two different ways to access in code, I'm guessing
[20:54:59 CET] <jkqxz> No.  The QSV hwaccel isn't a real hwaccel.  It's just used as a strange way of signalling to the QSV decoder that you want output in GPU surfaces rather than in normal memory.  (If you try to use it without the QSV decoder, it is silently ignored.)
[20:56:56 CET] <Phi> alrighty
[21:04:43 CET] <Phi> So if I'm switching from H264 native encoder to H264_QSV, should I need to change anything?
[21:05:03 CET] <Phi> atm I'm getting error -15, incompatible video params
[21:05:23 CET] <Phi> *invalid video params, even
[21:06:46 CET] <Phi> I am getting it to work with the static EXE build, though
[21:06:47 CET] <jkqxz> There should be an explanatory error message of some kind.
[21:07:14 CET] <Phi> h264_qsv : Error initializing the encoder, MFXVideoENCODE_Init returned -15
[21:07:14 CET] <Phi>  Couldn't open video file codec context: avcodec_open2 failed with error -22: Invalid argument.
[21:07:27 CET] <Phi> that's all I have in the log
[21:09:54 CET] <jkqxz> Hmm, yeah.  That stuff is all a bit horrible because it requires specific sets of parameters (which vary by platform/version) and if you get anything wrong it just says "nope!" and refuses to do anything.
[21:12:11 CET] <Phi> not as bad as some issues I've had, though
[21:12:25 CET] <Phi> Visual Studio NuGet *shiver*
[21:16:08 CET] <jkqxz> I would suggest getting the encode setup you want working in ffmpeg, then copy the parameters from a working AVCodecContext there.
[21:17:11 CET] <Phi> well, atm I'm copying from a received RTSP H264 stream
[21:17:43 CET] <Phi> one difference I've noted is that the received stream is YUV420P, and apparently Intel QSV requires NV12
[21:18:18 CET] <Phi> so I may have to swap the U/V planes
[21:19:03 CET] <jkqxz> Yep, all hardware uses NV12.  Converting manually isn't too bad, or swscale can do it for you.
[21:24:51 CET] <Phi> if YUV420P -> NV12 is literally just swapping U/V (which from some docs I read a while back it is) then swscale will be somewhat slower
[21:25:01 CET] <Phi> than just manually doing it
[21:25:20 CET] <JEEB> NV12 is not just swapping chroma planes
[21:25:39 CET] <JEEB> NV12 is a single plane for luma (same as standard planar 4:2:0), but then chroma planes are *packed*
[21:25:54 CET] <JEEB> so it's like YYYYYY.... CbCrCbCrCbCr...
[21:26:09 CET] <JEEB> instead of YYYYYYY... CbCbCbCbCb.... CrCrCrCrCr....
[21:28:16 CET] <Phi> ah, okay
[21:28:20 CET] <Phi> that seems painful
[21:32:40 CET] <JEEB> you probably were thinking of VFW I420 and YV12
[21:32:52 CET] <JEEB> which are YCbCr and YCrCb all planar
[21:33:11 CET] <JEEB> NV12 seems to be efficient if you put effort into it
[21:33:18 CET] <furq> there aren't enough names for pixel formats
[21:33:18 CET] <JEEB> x264 converte itself into it
[21:36:35 CET] <BtbN> Yeah, all the hardware encoders want it, there must be something to it.
[21:36:46 CET] <Phi> kinda dumb this RTSP stream is in YUV420P then
[21:37:01 CET] <JEEB> it's not that if it's not raw
[21:37:17 CET] <JEEB> it's just that software decoders in lavc generally output 4:2:0 in full planar
[21:37:19 CET] <JEEB> :P
[21:38:16 CET] <JEEB> also a lot in libswscale and friends only supports planar 4:2:0
[21:38:32 CET] <JEEB> in any case it should not be too much of a difference :P
[21:38:54 CET] <llogan> atomnuker: is #2686 considered fixed?
[21:40:27 CET] <llogan> bah, wrong channel
[21:41:47 CET] <Phi> so wait, lavc decodes from NV12 to YUV420P then has to be re-encoded?
[21:41:56 CET] <Phi> T_T ima sue
[21:49:58 CET] <JEEB> encoded data is completely separate anyways
[21:50:18 CET] <JEEB> and planar/packed is the same data in a different order
[21:50:32 CET] <JEEB> (and packed together, as the name implies)
[21:51:21 CET] <JEEB> the format you're decoding just decodes into 4:2:0 YCbCr, which format your decoder outputs it as is a completely different discussion
[22:01:26 CET] <BtbN> most encoders, even hardware ones like nvenc, support planar input now
[22:01:38 CET] <BtbN> But convert to nv12 internally, maybe with the help of the hardware
[22:02:06 CET] <Phi> okay
[22:02:10 CET] <Phi> so scurrying on
[22:02:30 CET] <Phi> I have a minimum list of params for MFXVideoENCODE_Init
[22:02:53 CET] <Phi> which is in https://software.intel.com/sites/default/files/managed/47/49/mediasdk-man.pdf on page 224
[22:03:04 CET] <Phi> but I can't see where in ffmpeg to set those
[22:06:09 CET] <Phi> the only ffmpeg QSV example is Linux-only
[22:06:23 CET] <Phi> and is heavily reliant on VA API
[22:39:33 CET] <Phi> sigh
[22:39:39 CET] <Phi> I guess that's a no?
[22:46:35 CET] <jkqxz> The libav example is cleaner <https://git.libav.org/?p=libav.git;a=blob;f=doc/examples/qsvdec.c;hb=HEAD>.
[22:47:39 CET] <jkqxz> Not sure whether it actually works in ffmpeg, though.  It could depend on something hasn't been merged yet, though it looks like it might work.
[22:59:09 CET] <Phi> still, at least it doesn't mention VA API
[22:59:14 CET] <Phi> cheers
[23:03:25 CET] <Phi> hm, it does seem to refer to av_hwframe
[23:06:27 CET] <Phi> ffmpeg's example doesn't use that
[23:40:24 CET] <john_k_h> vp9
[00:00:00 CET] --- Fri Dec  9 2016


More information about the Ffmpeg-devel-irc mailing list