[Ffmpeg-devel-irc] ffmpeg.log.20160821

[burek]

[00:48:38 CEST] <VamoMenem> Unknown option "--enable-nvresize". :|
[04:02:51 CEST] <VamoMenem> someone know who i can compile ffmpeg whit scale_cpp filter?
[04:14:51 CEST] <furq> the what filter
[04:28:13 CEST] <VamoMenem> furq scale_npp filter
[04:28:42 CEST] <VamoMenem> cpp was a memory leak
[04:29:09 CEST] <furq> --enable-cuda --enable-libnpp
[04:29:26 CEST] <VamoMenem> awesome!
[04:29:32 CEST] <VamoMenem> now go to compile again!
[06:38:07 CEST] <VamoMenem> im gonna be crazy, try to complile ffmpeg with --enable-cuda and --enable-libnpp return ERROR: CUVID not found and installing NVIDIA NVENC SDK or  bumblebee-nvidia primus or libnvcuvid1 dont resolve the error
[06:41:22 CEST] <VamoMenem> ERROR: CUVID not found  in a nice red (:
[06:45:49 CEST] <c_14> Can you upload your config.log to a pastebin service? (At least the first and last 50 lines)
[06:46:09 CEST] <c_14> Or just look at it yourself, it should show what it tried doing when it failed somewhere near the end
[06:49:00 CEST] <VamoMenem> c_14 conflig.log says  fatal error: cuviddec.h: No such file or directory
[06:49:30 CEST] <c_14> Do you have that file somewhere on your system? If so, add the directory to the include path
[06:49:54 CEST] <VamoMenem> i now locating
[06:50:24 CEST] <VamoMenem> i have in /usr/local/cuda-7.5/include/cuviddec.h
[06:50:39 CEST] <VamoMenem> is where i have to put the headers of cuda utils i gess
[06:51:24 CEST] <VamoMenem> how i can add this directory?
[06:53:47 CEST] <c_14> --extra-cflags="-I/usr/local/cuda-7.5/include"
[06:57:06 CEST] <VamoMenem> now working but stops in ERROR: libnpp not found. In config.log says /usr/bin/ld: cannot find -lnppi and /usr/bin/ld: cannot find -lnppc
[06:58:52 CEST] <VamoMenem> god is playing whit me
[06:59:06 CEST] <c_14> Let me guess, you have a /usr/local/cuda-7.5/lib or something like it
[06:59:15 CEST] <c_14> --extra-ldflags="-L/usr/local/cuda-7.5/lib"
[06:59:32 CEST] <c_14> Well, in this case it'd be libnpp
[06:59:41 CEST] <c_14> Basically the library path to wherever the libnpp libraries are
[07:00:11 CEST] <c_14> If they're dynamically linked you might also want to set an rpath (otherwise you'll have to set LD_LIBRARY_PATH on runtime in order to avoid linking errors)
[07:02:52 CEST] <VamoMenem> how i do that? i just install the nvidia sdk
[07:03:07 CEST] <c_14> Do you have a libnpp.so or libnpp.a ?
[07:03:25 CEST] <c_14> Eh, libnppc.so or libnppc.a
[07:03:33 CEST] <VamoMenem> i located /usr/local/cuda-7.5/lib64/libnppc.so so i think i need to change lib to lib64
[07:03:51 CEST] <c_14> Ye, just change the lib to lib64 in the --extra-ldflags I gave you
[07:03:52 CEST] <VamoMenem> a lot of .so
[07:03:56 CEST] <c_14> And add --enable-rpath
[07:04:08 CEST] <VamoMenem> ha ok
[07:04:22 CEST] <VamoMenem> to carry the libs whit the bins later
[07:05:14 CEST] <VamoMenem> worked!!
[07:06:30 CEST] <VamoMenem> i go to cry, so beutifull the make working
[07:06:36 CEST] <VamoMenem> thanks!
[07:06:53 CEST] <c_14> np
[07:19:49 CEST] <VamoMenem> happened the --enble-rpath issue
[07:20:01 CEST] <VamoMenem> when i try to run ffmpeg: ./ffmpeg: error while loading shared libraries: libnppi.so.7.5: cannot open shared object file: No such file or directory
[07:22:09 CEST] <c_14> Either rebuild with enable-rpath, or do LD_LIBRARY_PATH=/usr/local/cuda-7.5/lib64 ./ffmpeg
[07:23:55 CEST] <VamoMenem> to check was working  LD_LIBRARY_PATH=/usr/local/cuda-7.5/lib64 ./ffmpeg -filters
[12:21:35 CEST] <maziar> why i get this error : 1 frames left in the queue on closing
[12:26:00 CEST] Last message repeated 1 time(s).
[12:41:47 CEST] <tripkip> I'm having issues with multiple files  on my 2009 MBP in latest Kodi release, which uses VTB hwdec. Kodi team points to ffmpeg. I'm trying to figure out if I have the same issues with ffplay. Is VTB used by default when running ffplay or is there a spefic parameter to enable it (can't find it in the doc)?
[13:04:50 CEST] <Threads> tripkip how about posting a errorlog of some sort to pastebin and paste the link here if you have a error log
[13:06:01 CEST] <tripkip> Yes that's what I'm aiming for, but I need to know the proper command so that ffplay uses VTB like Kodi Krypton does
[13:07:09 CEST] <tripkip> I came this far (don't know if it's correct): "./ffplay -codec:v h264_videotoolbox file.mkv" - I'm only getting audio output in the player though..
[13:10:19 CEST] <tripkip> Seems that the decoder h264_videotoolbox is not available in ffplay
[13:11:16 CEST] <tripkip> "./ffplay -codecs | grep 264" returns " DEV.LS h264                 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (decoders: h264 h264_vda ) (encoders: libx264 libx264rgb h264_videotoolbox )"
[15:58:35 CEST] <vash1> hey there
[15:59:04 CEST] <vash1> how can i find spectrum difference between two almost identical mono wavs?
[15:59:38 CEST] <fritsch> go to audacity and look at the spectrum?
[16:00:18 CEST] <vash1> already did
[16:00:25 CEST] <flux> I guess spectrum doesn't have comparison tools for spectrum. octave perhaps might be more appropriate if you have sufficient memory? ..and some mathematical aptitude to work with octave and its docs
[16:00:26 CEST] <vash1> but its not changing somehow...
[16:00:40 CEST] <vash1> i tried this trick http://manual.audacityteam.org/man/invert.html
[16:01:26 CEST] <vash1> i see..
[16:01:28 CEST] <flux> to what end? I don't think it affects spectrum at all.
[16:02:08 CEST] <bencoh> flux: fourier is a linear operation, so ...
[16:02:25 CEST] <flux> bencoh, so ... ?-)
[16:02:27 CEST] <fritsch> bencoh: think what you see in spectrum and what this invert does
[16:02:40 CEST] <fritsch> and the result for the spectrum is? :-)
[16:02:57 CEST] <flux> doesn't invert simply change the phase by 180 degrees, and spectrum doesn't visualize that?
[16:03:05 CEST] <fritsch> yeah, exactly
[16:03:53 CEST] <bencoh> fritsch: isn't the idea there to do the sum of stream A plus invert(B)?
[16:04:07 CEST] <vash1> let's say if we have wav file encoded in mp3 @ 128kbps and mp3 @ 320kbps - will not the frequency difference of both will be 16+kHz ?
[16:04:45 CEST] <vash1> from 16 kHz and above (since mp3 @ 128 has frequency cut off on 16kHz)
[16:06:30 CEST] <bencoh> fritsch: looks like that's what they want to achieve on that page (A-B)
[16:06:55 CEST] <bencoh> (and then spectrum(A-B) ~= spectrum(A) - spectrum(B))
[16:07:39 CEST] <vash1> to be clear, basically I'm trying to find the difference between central channel of DTS-HD MA versus to DTS core.
[16:08:53 CEST] <vash1> maybe my question wasn't formulated correctly.
[16:09:18 CEST] <fritsch> bencoh: the problem starts with: what is B concerning A and what means this inversion of A for the spectrum
[16:11:19 CEST] <vash1> so how can I compare with ffmpeg dts-hd-ma.wav and dts-core.wav and output the difference to output.wav?
[16:11:20 CEST] <bencoh> fritsch: he wants a spectrum comparison ... I guess visualizing the "difference" of the two spectrums is a good start
[16:11:31 CEST] <fritsch> yes, for sure
[16:11:41 CEST] <bencoh> well then 17:06 < bencoh> (and then spectrum(A-B) ~= spectrum(A) - spectrum(B))
[16:11:48 CEST] <bencoh> ?
[16:12:00 CEST] <fritsch> he generates B from A
[16:12:08 CEST] <vash1> having output.wav i can analyze it with any spek software
[16:17:32 CEST] <fritsch> bencoh: perhaps more concrete - did you check the "invert" he is doing and how it is done in _spatial_ domain?
[16:17:54 CEST] <bencoh> I didn't
[16:18:08 CEST] <fritsch> yeah - and that's my point, he is not computing on the spectrum
[16:18:21 CEST] <fritsch> he is "Invert flips the audio samples upside-down, reversing their polarity"
[16:18:30 CEST] <bencoh> I just assumed he'd do it in time domain
[16:18:38 CEST] <fritsch> hehe
[16:18:47 CEST] <fritsch> that would make sense
[16:18:52 CEST] <fritsch> but not what he is doing
[16:29:41 CEST] <vash1> already figured that out. here is the screenshot. https://s14.postimg.io/696bai0hb/Screen_Shot_2016_08_21_at_17_25_33.png thanks for your help fritsch bencoh
[18:42:53 CEST] <codespells> Can ffmpeg used with avisynth scripts? If so what would a cmd look like?
[19:30:36 CEST] <furq> codespells: if it was built with --enable-avisynth then you can just do -i foo.avs
[20:02:50 CEST] <codespells> furq whoa cool. So I just just do something like ffmpeg -i myscript.avs  -c:a copy -c:v copy output.mkv and the script will be executed and saved in a mkv container?
[20:05:06 CEST] <furq> sure
[20:05:26 CEST] <furq> as long as you have avisynth installed and it's the same arch as your ffmpeg binary
[20:05:53 CEST] <codespells> that part I can sort out.
[20:11:00 CEST] <codespells> neat! I take it what I did in my cmd is totally uncompressed right? Also would be nice to drop this into Huffyuv or Lagarith. Not sure what is better if I am going to further work on my file.
[20:26:28 CEST] <tripkip> Any OSx user here that can tell me how to run ffplay with VTB enabled?
[20:35:05 CEST] <parsec> tripkip, I'm an OSX user, but I only use ffmpeg and ffprobe
[20:35:39 CEST] <parsec> And I only use it for video conversion to avi with specific settings
[20:54:50 CEST] <codespells> is there someway to output to ffmpeg from within a avisynth script?
[20:59:00 CEST] <furq> you stop writing the script
[21:07:10 CEST] <codespells> furq, lol yeah normally. Thing is that I am using AvsPmod as a editor so I can see what is going on. idk maybe there is a way to do this with ffplay or something instead of AvsPmod ?
[22:00:29 CEST] <sbraz> does it harm quality in any way if i use vf=rgb24 so that i can crop to an odd amount of pixels when encoding to x264?
[22:01:12 CEST] <durandal_1707> sbraz: I just fixed crop filter
[22:01:34 CEST] <durandal_1707> you need recent ffmpeg version
[22:02:03 CEST] <durandal_1707> and add exact option set to 1
[22:02:10 CEST] <sbraz> durandal_1707: so it'll automatically convert the colorspace?
[22:02:21 CEST] <BtbN> x264 only encodes YUV
[22:02:37 CEST] <klaxa> there is libx264rgb though
[22:02:49 CEST] <BtbN> But that won't really play anywhere.
[22:02:57 CEST] <sbraz> BtbN: yeah but if i crop from pixel 1, it has to recompute yuv values, right?
[22:03:08 CEST] <durandal_1707> sbraz: I fixed crop so it crops odd lines
[22:03:57 CEST] <sbraz> durandal_1707: my understanding of how this works might be a little bit limited, but if it can crop on odd pixels, it will require some approximation, won't it?
[22:04:07 CEST] <sbraz> interpolation even
[22:04:18 CEST] <durandal_1707> nope
[22:05:26 CEST] <sbraz> durandal_1707: even if my source is yuv too?
[22:08:50 CEST] <durandal_1707> no need to upconvert to yuv444p and downconvert to yuv420p
[22:09:07 CEST] <sbraz> if my source video is yuv420, the cb and cr values apply to 2 pixels in the same column, right?
[22:09:20 CEST] <durandal_1707> yes
[22:09:53 CEST] <sbraz> so if i were to crop from line 1 to line 3, how would that work?
[22:11:31 CEST] <sbraz> durandal_1707: oh by recent you really meant "recent" haha
[22:12:21 CEST] <sbraz> i'll compile the git version i guess
[22:13:31 CEST] <durandal_1707> afaik interpolation is not needed
[22:14:09 CEST] <sbraz> durandal_1707: the patch looks very simple, think i could apply it to v2.8.7?
[22:14:24 CEST] <durandal_1707> Yes
[23:33:26 CEST] <ito__> hi, I have been trying to make a .webm file the video comes out blurry, I pasted the command and output here   http://pastebin.com/yjAJ70jK
[23:34:55 CEST] <furq> maybe try setting the bitrate to something higher than 200kbps
[23:35:50 CEST] <ito__> furq, you mean the default setting, I tried many times but didn't want to put all the times up, last I tried was -vb 1M
[23:39:02 CEST] <ito__> furq, this is the last one http://pastebin.com/DR4FJXAG
[23:56:47 CEST] <azonenberg> Hi guys
[23:56:59 CEST] <azonenberg> I'm trying to do some post-processing on archival security camera footage
[23:57:36 CEST] <azonenberg> basically, take in a directory full of 1080p25 h.264 files (say, one every 30 minutes over a whole day)
[23:57:50 CEST] <azonenberg> a) concatenate them into a single video file
[23:57:53 CEST] <azonenberg> b) reduce the bit rate
[23:58:03 CEST] <azonenberg> and c) reduce the framerate by dropping frames, while keeping 1:1 playback speeds
[23:58:36 CEST] <azonenberg> The idea is to allow me to keep a fairly long "tail" of footage with much reduced disk space compared to the high quality 1080p25 videos that I only keep for a couple of days unless something bad happens
[23:58:48 CEST] <ito__> thanks, I guess I will try the forum
[23:59:06 CEST] <azonenberg> I have a filter graph set up using filter_complex concat to append the original input files (as well as transcoding to MP4 since I had trouble outputting h.264 at low bit rate)
[23:59:54 CEST] <azonenberg> But when i try -vf fps= i have problems
[00:00:00 CEST] --- Mon Aug 22 2016