[Ffmpeg-devel-irc] ffmpeg.log.20160504

[burek]

[00:31:17 CEST] <xxDCDxx> sorry, I got disconnected
[00:32:26 CEST] <xxDCDxx> should I try to use LIPO to build universal binaries for the CLI? running the command './build-ffmpeg.sh lipo' after removing '--disable-programs' built universal libs, but not the universal cli
[00:35:14 CEST] <c_14> Looking at the script it only does that for the libraries. Are there no binaries at all or do they just not work?
[00:37:57 CEST] <xxDCDxx> the libraries seem fine, the thing is I also wanted universal CLI files (that can run on arm7, arm64 and x86_64
[00:38:00 CEST] <xxDCDxx> I tried this:
[00:38:48 CEST] <xxDCDxx> http://pastebin.com/m4Ru1fpr
[00:39:40 CEST] <xxDCDxx> but I get the error shown there. I suspect the 'lipo' command is only for libraries, not for executables
[00:43:54 CEST] <c_14> I don't think there's an equivalent for binaries
[01:10:15 CEST] <Plorkyeran> lipo is how you create fat executables
[01:12:54 CEST] <Plorkyeran> thin/arm64/share/ffmpeg does not sound like an executable, though
[01:13:15 CEST] <Plorkyeran> the executables are in the bin directories
[01:13:24 CEST] <neuro_sys> I probably asked this before, but I couldn't find it again. What's the opposite of colorkey? Make anything that is not a particular color transparent?
[01:15:30 CEST] <andrey_utkin> neuro_sys, i think you could go just hacking the source code of colorkey filter, should be easy for your case :)
[01:20:12 CEST] <neuro_sys> sounds like it
[01:20:43 CEST] <neuro_sys> but I might go another way that doesn't require stencil masking
[01:20:51 CEST] <furq> 16:07:02 ( neuro_sys) what is a filter that does the inverse of colorkey?
[01:20:52 CEST] <furq> 16:08:14 ( c_14) presumably overlay over a color?
[01:21:03 CEST] <neuro_sys> furq: ha!
[01:25:53 CEST] <llogan> neuro_sys: maybe something like this " "[0:v]colorkey,alphaextract,negate[alf];[0:v][alf]alphamerge[fg];[1:v][fg]overlay"
[01:32:25 CEST] <neuro_sys> llogan: looks useful, I'm trying, thanks
[01:38:23 CEST] <neuro_sys> I seem to have it wrong: "[1:v]colorkey=color=black[cout];[0:v]colorkey=color=white,alphaextract,negate[alf];[0:v][alf]alphamerge[fg];[fg][cout]overlay[out]"
[01:39:53 CEST] <llogan> not much i can say without more constext such as your full command and complete console output pasted into a pastebin site and the link provided here
[01:44:04 CEST] <neuro_sys> okay, hold on
[01:45:04 CEST] <neuro_sys> llogan: http://imgur.com/a/3XaZG
[01:45:15 CEST] <neuro_sys> these are thw two images I'm working on
[01:45:58 CEST] <neuro_sys> I was hoping to use colorkeying both ways in order to avoid compression artifacts from being visible
[01:46:24 CEST] <neuro_sys> as well as when the base image has the target surface ocluded by various objects
[01:47:22 CEST] <llogan> i have to go now, but i think you can get it to work overlaying the top image over the bottom one. no need to colorkey the bottom one
[01:49:02 CEST] <neuro_sys> the top image has some artifacts at the top
[01:49:08 CEST] <neuro_sys> but yeah I should get rid of them first
[02:29:19 CEST] <xxDCDxx> it turns out that I cannot fork new processes from iOS, so compiling the ffmpeg CLI is of no use. Now I'm trying to rename main() to mainFFMPEG() in ffmpeg.c/.h and I'll call that with the proper arguments from my app. Is there any change I need to make to the makefile so it's built as part of the library instead of a standalone executable?
[02:32:06 CEST] <pzich> is there a reason you aren't using a static library like https://github.com/chrisballinger/FFmpeg-iOS ?
[02:33:20 CEST] <xxDCDxx> yeah, I tried that. The thing is that I want to replicate the command 'ffmpeg -i recording.mp4 -vsync vfr -q:v 9 recording.mjpeg', but I don't know how to do it with the raw lib API
[02:33:48 CEST] <xxDCDxx> I'm finding it a little bit complicated using the standalone libs
[02:43:52 CEST] <xxDCDxx> so, is it wise to go with the ffmpeg.c route, or should I learn the libraries and replicate the behaviour myself?
[02:44:39 CEST] <xxDCDxx> getting some pointers would be much appreciated ;)
[03:13:03 CEST] <neuro_sys> would I get a significant performance gain from using libav instead of CLI in my case where I do something like ffmpeg -i %04d.png /* complex filters */ output.mp4
[03:13:20 CEST] <neuro_sys> where I generate png's in memory and write to disk for passing to ffmpeg as such
[05:27:35 CEST] <Abbott> is it possible to disassemble a variable bitrate video into it's frames, edit them, and recompile the variable frames correctly back into the original clip?
[05:28:32 CEST] <Abbott> or do you have to resample the frames at a constant bitrate, edit, then assemble again? Would there be a loss of data? (is this considered encoding when resampling the frames?)
[05:32:13 CEST] <furq> that depends on the codec
[05:32:20 CEST] <furq> if it's not an intra-only codec then the answer is probably no
[05:32:25 CEST] <furq> although i'm no expert
[05:32:48 CEST] <furq> that would apply regardless of bitrate as well
[05:39:32 CEST] <kepstin> neuro_sys: you could get a significant speed gain by skipping the whole png thing and just writing the raw, unencoded video into ffmpeg. Using it as a library could help a bit there by reducing the number of data copies done.
[11:22:19 CEST] <neuro_sys> kepstin: thanks
[11:26:20 CEST] <neuro_sys> btw, what do you think might this masking's purpose? int rowStride = (width * 3 + 3) & ~0x3;
[11:31:19 CEST] <xxDCDxx> hi, I'm trying to replicate what the command 'ffmpeg -i recording.mp4 -vsync vfr -q:v 9 recording.mjpeg' does, but using the libraries, any pointers? I have no experience using video libraries
[11:35:11 CEST] <neuro_sys> seems like it rounds width to multiples of four
[14:16:52 CEST] <Admin7762Frd> good morning everyone
[14:22:28 CEST] <Admin7762Frd> anyone around to help me figure out why i am getting these errors " http://pastebin.com/raw/urJhU2AU " Enter command: <target>|all <time>|-1 <command>[ <argument>] ".. it plays the files for a few seconds and then moves on to the next .. these are 1 gig files. not sure why its ending lik ethat
[14:24:00 CEST] <BtbN> most of these nvenc options look invalid.
[14:25:33 CEST] <Admin7762Frd> really?
[14:25:48 CEST] <Admin7762Frd> :( what is wrong with it ?
[14:27:52 CEST] <Admin7762Frd> NVENC_OPTS="-preset lowLatencyHQ  -qp 21 -numB 1 -goplength 180 -rcmode 32 -fps 30 -ptsdelay 0.0" you mean ?
[14:28:07 CEST] <Admin7762Frd> i been using those options forever for the transcoder
[14:37:32 CEST] <Fyr> how to show the list of codecs that are available for the specified container?
[14:38:06 CEST] <Admin7762Frd> ffmpeg -codecs ?
[14:39:06 CEST] <Fyr> Admin7762Frd, I see the output, but I don't see what accounts for mp4.
[14:40:10 CEST] <Fyr> Admin7762Frd, what codecs can ffmpeg mux into mp4?
[14:44:52 CEST] <Admin7762Frd> Fyr , i wish i knew enough about it .. but you can do mkv ,, and other stuff
[14:45:23 CEST] <Admin7762Frd> BtbN > what is wrong with my nvenc options ?
[14:45:35 CEST] <Admin7762Frd> its not using the nvenc built into ffmpeg
[14:47:14 CEST] <Admin7762Frd> reason i didn't use that is because  the nvenc does not decode and using vdpau gives me a bunch of errors too that i have not figured out yet like " Cannot open the X11 display .
[14:47:14 CEST] <Admin7762Frd> VDPAU init failed for stream #0:0.
[14:47:14 CEST] <Admin7762Frd> vdpau hwaccel requested for input stream #0:0, but cannot be initialized. "
[14:52:49 CEST] <Admin7762Frd> ok moved to vdpau & nvenc built into ffmpeg.. but still getting that output error
[15:11:39 CEST] <relaxed> Admin7762Frd: did you ever report the differences between NvTranscoder and nvenc on the bug tracker?
[15:12:22 CEST] <Admin7762Frd> nope.. i think i got it close enough now that its the same so now i switched.. the problem was always vdpau having problems decoding
[15:12:31 CEST] <Admin7762Frd> where is the transcoder does the decoding too which is nice
[15:12:55 CEST] <Admin7762Frd> that said i got the vdpau going right now along with nvenc but my error still is coming up
[15:12:59 CEST] <Admin7762Frd> let me post a new pastbin
[15:13:46 CEST] <Admin7762Frd> http://pastebin.com/raw/MbvbCJe5
[15:14:02 CEST] <Admin7762Frd> see that one.. the error still pops up with " Enter command: <target>|all <time>|-1 <command>[ <argument>]
[15:14:19 CEST] <Admin7762Frd> oh and also " Parse error, at least 3 arguments were expected, only 1 given in string 'anada.Movie0020.mp4' "
[15:15:27 CEST] <relaxed> quote $lines
[15:20:55 CEST] <DEATH> https://trac.ffmpeg.org/wiki/CompilationGuide/WinRT <= needs amending for latest msys64
[15:21:31 CEST] <DEATH> We expect msys64 to inherit PATH from the calling process, that does not work with latest msys64, unless you edit msys2_shell.bat
[15:21:32 CEST] <DEATH> rem To export full current PATH from environment into MSYS2 uncomment next line
[15:21:32 CEST] <DEATH> rem set MSYS2_PATH_TYPE=inherit
[15:29:01 CEST] <Admin7762Frd> death . i did quote the lines but still.. :(
[15:29:29 CEST] <Admin7762Frd> > /home/ffmpeg-3.0.1/ffmpeg -y -re -i "$lines"
[15:29:37 CEST] <Admin7762Frd> same result
[15:34:53 CEST] <Admin7762Frd> must have something to do with the fifo
[15:34:56 CEST] <Admin7762Frd> just not sure what
[15:56:29 CEST] <Admin7762Frd> any other clues ?
[16:34:23 CEST] <Workflow> Guys, is there a way to make snapshotting quicker? For example, i'm exporting every fps=1/120, is it down to CPU? I have 4 cores assigned to a dedicated VM, but only 2 are being used roughly.
[16:38:19 CEST] <Fyr> =(
[16:40:21 CEST] <Fyr> guys, I receive the following error:
[16:40:21 CEST] <Fyr> http://pastebin.com/Zdufa8AY
[16:40:21 CEST] <Fyr> what could that be?
[16:40:33 CEST] <Fyr> the output file is 790 bytes-sized.
[16:52:32 CEST] <xxDCDxx> using the MJPEG codec, what is the equivalent of: "-q:v 9" in code?
[16:53:01 CEST] <xxDCDxx> I'm doing transcoding, as in this file: https://ffmpeg.org/doxygen/2.8/transcoding_8c-example.html
[17:03:15 CEST] <jkqxz> xxDCDxx:  AVCodecContext.global_quality  (The scale may be different, though.)
[17:06:39 CEST] <xxDCDxx> thanks!
[17:07:20 CEST] <Admin7762Frd> getting closer.. now i get " Hardware accelerated decoding with frame threading is known to be unstable and its use is discouraged. "
[17:07:31 CEST] <Admin7762Frd> can you tell me.. why would it do that for vdpau ?
[17:07:51 CEST] <Admin7762Frd> i tried using threads 1 but  then decoding doesn't work
[17:18:18 CEST] <xxDCDxx> jkqxz: it doesn't seem to do the trick, quality is the same (quite low). I tried values ranging from 1 to 9*FF_QP2LAMBDA, with no apparent change
[17:19:31 CEST] <jkqxz> For MJPEG?  The scale should be is 1-100, so 9 will be very low.  (Though I admit I don't actually know whether that encoder even respects the quality value at all...)
[17:22:35 CEST] <xxDCDxx> I'll try 100
[17:22:47 CEST] <xxDCDxx> there should be a way, given that command line does it, no?
[17:23:33 CEST] <jkqxz> Um, yeah.  That is quite strong evidence, I agree.
[17:24:00 CEST] <xxDCDxx> (e.g., -q:v 30 looks like crap, -q:v 9 looks nice)
[17:25:00 CEST] <jkqxz> That seems the wrong way around.  Are you sure you are encoding with MJPEG?
[17:25:37 CEST] <xxDCDxx> I'm pretty sure. According to this post, that's correct: http://stackoverflow.com/a/32151594/269753
[17:25:57 CEST] <jkqxz> Or maybe it's inverted somehow.  JPEG usually has a quality scale of 1-100, but it could be remapped somehow in lavc.
[17:26:38 CEST] <xxDCDxx> "You can use -q:v instead. For MJPEG the effective range is a linear scale of 2-31, and a lower value will likely end up with a higher quality output."
[17:26:46 CEST] <xxDCDxx> that's what that online dude claims
[17:27:18 CEST] <xxDCDxx> "enc_ctx->global_quality = 100" and "enc_ctx->global_quality = 1" seem to produce the exact same results
[17:29:25 CEST] <jkqxz> FF_QP2LAMBDA is 118.  Try something bigger: 3000?
[17:31:12 CEST] <xxDCDxx> tried 3000 and 10000, same results
[17:31:17 CEST] <xxDCDxx> the value seems completely ignore
[17:31:33 CEST] <xxDCDxx> *ignored
[17:31:47 CEST] <xxDCDxx> because "-q:v 30" makes it look way worse than when I tried enc_ctx->global_quality = 1
[17:40:44 CEST] <jkqxz> AVCodecContext.flags |= AV_CODEC_FLAG_QSCALE?
[17:45:51 CEST] <xxDCDxx> doing "enc_ctx->flags |= AV_CODEC_FLAG_QSCALE" seems to improve the quality a bit, but global_quality seems to still be ignored
[18:06:02 CEST] <xxDCDxx> AVCodecContext.compression_level is also ignored
[18:07:43 CEST] <zarathushtra> i have a video and an audio stream that i got with "mkvextract tracks $1.mkv 0:$1.avi 2:$1.ogg", i need to put it on an avi file and i can do it with "ffmpeg -i $1.ogg -i $1.avi -c copy out.avi", but i also need to put a .srt on it... i tried before the ffmpeg using a video filter, is that the correct approach?
[18:24:59 CEST] <furq> zarathushtra: -vf subtitles=subs.srt
[18:25:08 CEST] <furq> you'd need to reencode for that because it burns the subs into the video
[18:25:51 CEST] <zarathushtra> furq: can i reencode using the tracks or the out.avi
[18:28:53 CEST] <PlanC> I have a video with varying audio bitrate
[18:29:06 CEST] <PlanC> I just want to extract the audio
[18:29:26 CEST] <__jack__> PlanC: ffmpeg -i source -map 0:a -c copy output.mkv
[18:29:36 CEST] <PlanC> how should I know what to set the audio bitrate of the MP3 to?
[18:33:28 CEST] <PlanC> ___jack___: is there a way for me to just the audio bitrate of the video file?
[18:34:50 CEST] <PlanC> the reason I'm asking is because I have a shit ton of videos from my lectures but I'm running out of room so I'm trying to remove the videos and just save the audio
[18:35:01 CEST] <furq> you don't set the audio bitrate if you're extracting the stream
[18:35:19 CEST] <PlanC> furq: some of my files don't have a bit_rate in the stream
[18:35:25 CEST] <PlanC> they just have "formats"
[18:36:38 CEST] <furq> if the command __jack__ posted didn't work then paste the full output
[18:36:51 CEST] <furq> pastebin, rather
[18:37:56 CEST] <PlanC> furq: I haev another command that I use though
[18:38:19 CEST] <PlanC> ffmpeg -i video.mp4 -b:a 128k -y output.mp3
[18:38:26 CEST] <furq> that's reencoding
[18:38:37 CEST] <PlanC> that's what I need to do though
[18:38:44 CEST] <PlanC> because MP3 works the best for me
[18:38:52 CEST] <PlanC> it's small and the audio quality is good
[18:39:09 CEST] <PlanC> (small being really important because I have limited space)
[18:40:57 CEST] <zarathushtra> trying to use "-sub_charenc CP1252" i receive: Codec AVOption sub_charenc (set input text subtitles character encoding) specified for output file #0 (teste.avi) is not an encoding option.
[18:53:39 CEST] <__jack__> PlanC: best things you can do is something like the command is wrote : is will extract the same audio that you have (meaning: no quality loss, really fast processing)
[18:54:53 CEST] <__jack__> PlanC: if you want to reencode the audio, to get more space, please note that it will be lossless; you should use aac or other codec, as the perform really better than mp3 (best quality per size, lower size for the same quality)
[18:56:21 CEST] <__jack__> (but, throwing away the video will give you already almost all space back, so you probably don't need the reencode stuff)
[19:37:52 CEST] <zarathushtra> what is the difference between H.264 MP and H.264 BP
[19:38:31 CEST] <klaxa> sounds like main profile and baseline profile
[19:38:33 CEST] <kepstin> that's "main profile" and "baseline profile", see https://en.wikipedia.org/wiki/H.264/MPEG-4_AVC#Profiles
[19:39:15 CEST] <zarathushtra> Phmm
[19:39:18 CEST] <zarathushtra> cool
[19:39:35 CEST] <zarathushtra> im trying to convert a simple .mkv to a file that my tv can read
[19:39:35 CEST] <zarathushtra> http://www.samsung.com/hk_en/consumer/learningresources/TV/contents_lib/images/wiselink_Europe.pdf
[19:41:06 CEST] <kepstin> if you're encoding with libx264, you can use the '-profile main' option, and it'll limit itself to using only features in main profile.
[19:42:17 CEST] <kepstin> as long as you also stay within the supported video size, and pick an audio codec it likes, that should be good enough.
[19:43:24 CEST] <zarathushtra> mkvmerge shows that the video stream is Track ID 0: video (MPEG-4p10/AVC/h.264)
[19:43:46 CEST] <zarathushtra> ffmpeg: Stream #0:0: Video: h264 (High), yuv420p, 1280x720, SAR 1:1 DAR 16:9, 23.98 fps, 23.98 tbr, 1k tbn, 47.95 tbc (default)
[19:44:02 CEST] <zarathushtra> (High) means high profile? thus i need to reencode?
[19:46:25 CEST] <zarathushtra> oh nvm i was looking at the wrong data... my tv wont support more than 720x480... thats why -.-
[21:24:12 CEST] <operator> getting error with ffmpeg: http://pastebin.com/q5eaFak6
[21:24:33 CEST] <operator> is the problem my video file or missing codecs?
[21:27:55 CEST] <llogan> operator: did you "Consider increasing the value for the 'analyzeduration' and 'probesize' options"?
[21:28:18 CEST] <llogan> can you provide a short input sample? what made the input file? do you know what format the audio is supposed to be?
[21:37:54 CEST] <rsully> Is there a particular linux distro or script that makes compiling ffmpeg with options/dependencies far easier? I want to setup a VM for encoding that has up-to-date everything as easy as possible
[21:40:05 CEST] <llogan> rsully: as for repository offerings Arch keeps up to date with latest release branch. Debian unstable is a little slower but it is often relatively up to date.
[21:40:58 CEST] <llogan> as for compiling, i guess that just comes down to personal preferance: manually, build scripts, user repos, etc
[21:44:26 CEST] <operator> llogan: file created by blackmagic intensity pro 4k card
[21:46:04 CEST] <llogan> i imagine that should be some sort of PCM audio
[21:47:30 CEST] <furq> rsully: it's not quite what you asked for but freebsd makes it very easy to compile a customised ffmpeg
[21:47:34 CEST] <operator> llogan: i can give you 1sec example, is it enough?
[21:47:39 CEST] <furq> as long as the stuff you want is in make config
[21:48:02 CEST] <llogan> operator: if it allows the issue to be duplicated
[21:53:00 CEST] <operator> llogan: I'm getting the same error with this file: http://pc.cd/S2DctalK
[22:05:17 CEST] <llogan> operator: do you have some from the same hardware where the audio is recognized properly?
[22:16:33 CEST] <durandal_1707> llogan: its invalid guid, s24le pcm is stored
[22:18:37 CEST] <rsully> furq I essentially want to take my Mac homebrew config options and stick it on my server
[22:19:02 CEST] <rsully> llogan I would need to go build script route I think, as I need options that can't be distributed as binaries
[22:19:37 CEST] <Fyr> is here anyone who used SubtitleEdit?
[22:20:52 CEST] <rsully> furq where would i look for the freebsd options to know if that would meet my needs?
[22:25:16 CEST] <c_14> rsully: the options are listed here https://svnweb.freebsd.org/ports/head/multimedia/ffmpeg/Makefile?revision=414501&view=markup . FreeBSD only has a relatively old version of ffmpeg though
[22:25:24 CEST] <rsully> That's no good then
[22:27:02 CEST] <furq> oh have they not upgraded to 3.x yet
[22:27:13 CEST] <furq> i figured mine was on an old version because i'm still on 9.3
[22:27:30 CEST] <furq> that's disappointing, they were generally only a few days behind with 2.x releases
[22:29:16 CEST] <c_14> probably because of the major api changes in 3.0
[22:29:18 CEST] <c_14> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=207547
[22:37:14 CEST] <durandal_1707> operator: it looks like audio is pcm_s24le
[22:45:27 CEST] <operator> so, audio makes trouble
[22:51:42 CEST] <Fyr> audio always brings trouble.
[22:51:59 CEST] <Fyr> whithout it's much better.
[23:04:20 CEST] <llogan> rsully: you can make your own build script, or use ABS or AUR in Arch Linux
[23:04:54 CEST] <rsully> ABS / AUR ?
[23:05:17 CEST] <llogan> https://wiki.archlinux.org/index.php/Arch_User_Repository
[23:05:23 CEST] <llogan> https://wiki.archlinux.org/index.php/Arch_Build_System
[23:06:39 CEST] <llogan> or to be more specific, make your own PKGBUILD
[00:00:01 CEST] --- Thu May  5 2016