[Ffmpeg-devel-irc] ffmpeg.log.20170224

burek burek021 at gmail.com
Sat Feb 25 03:05:01 EET 2017


[00:14:29 CET] <funyun> hi. can anyone help me with this error? http://pastebin.com/ke9pFz5E
[00:14:56 CET] <funyun> i only get it when i add "-profile:v high -level 5.2"
[00:19:32 CET] <BtbN> [h264_nvenc @ 000001ad98ab9f40] The variable bitrate rate-control requires the 'qmin' and/or 'qmax' option set.
[00:19:58 CET] <BtbN> Also highly possible that your card simply does not support level 5.2.
[00:20:18 CET] <BtbN> You usually use the level setting to limit the level, not set it to a high one.
[00:21:02 CET] <funyun> BtbN: is there a specific qmin qmax setting? i've tried multiple variations and i get the same error only without that qmin line
[00:21:13 CET] <funyun> BtbN: i have a gtx 1080
[00:21:20 CET] <funyun> will that support 5.2?
[00:21:33 CET] <BtbN> no idea, doesn't look like it if it works without that level parameter.
[00:22:21 CET] <funyun> BtbN: ah i see. thanks for the help
[00:24:02 CET] <BtbN> There also is no level 5.2 restriction. The highest you can set is 5.1.
[00:24:09 CET] <BtbN> https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/nvenc_h264.c#L70
[00:27:13 CET] <funyun> BtbN: yep that's what it was. i just tried 5.1 and it worked fine. sucks but thanks for the help
[00:27:23 CET] <BtbN> what?
[00:27:35 CET] <BtbN> You do realize it's a restriction-setting?
[00:27:48 CET] <funyun> what does that mean?
[00:27:49 CET] <BtbN> Not a "use this level"
[00:27:58 CET] <BtbN> It will most likely still use the lowest level possible.
[00:28:16 CET] <BtbN> so restricting it to the highest level makes no sense to begin with
[00:29:23 CET] <funyun> BtbN: no, the output was indeed 5.1. but i need 5.2 because anything else stutters on my player
[00:29:37 CET] <BtbN> so don't specify a restriction then
[00:29:55 CET] <BtbN> And I highly doubt that.
[00:30:06 CET] <BtbN> The higher the level, the more complex it gets to decode it
[00:30:20 CET] <funyun> BtbN: that's what i thought
[00:30:22 CET] <BtbN> That's the entire point of the levels
[08:32:18 CET] <k_sze[work]> ugh, ffmpeg is using over 1.5GB resident memory, WAT?!
[08:32:48 CET] <k_sze[work]> I specified -c:v libx264 -preset veryslow, but still, 1.5GB, seriously?
[08:42:40 CET] <JEEB> depending on the resolution, input format etc that's probably not too much
[08:43:43 CET] <JEEB> at veryslow I think the lookahead in x264 is like 60 raw video frames?
[08:44:08 CET] <JEEB> and then the rest of ffmpeg et al also has to have different buffers
[09:06:32 CET] <ChuckLePlant_> good day, if I have compiled ffmpeg with cuvid support, how can I make sure I initialize AVCodec using cuvid instead of cpu based codec? Currently I call avcodec_find_decoder(AV_CODEC_ID_H264);
[09:25:25 CET] <JEEB> ChuckLePlant_: if there's no separate AVCodecID, there's avcodec_find_encoder_by_name
[09:25:31 CET] <JEEB> to which you can stick a string in
[09:26:28 CET] <ChuckLePlant_> JEEB where can I find the valid strings to be used?
[09:29:05 CET] <JEEB> in the actual decoder definitions
[09:29:12 CET] <JEEB> it's the same name as you'd use with the ffmpeg cli tool
[09:29:23 CET] <JEEB> ffmpeg -decoders or -encoders is one way :P
[09:31:33 CET] <JEEB> there might be a "give me all strings for encoders for this one codec id" so you could go through them, but didn't look far enough
[09:36:52 CET] <ChuckLePlant_> thanks!
[11:35:55 CET] <acamargo> morning! people, I have a GPU 0: GeForce GTX 745 (UUID: GPU-ba41d615-c63f-1824-634d-15564856f881) and I want to transcode mpegts for streaming. But I'm receiving [h264_nvenc @ 0x423cde0] OpenEncodeSessionEx failed: out of memory (10) when I am going to use more than 268MiB (the board has 4GiB). how can I use full memory?
[11:37:19 CET] <jkqxz> Is that when opening a third encode session?
[11:40:24 CET] <acamargo> yep :-)
[11:40:28 CET] <acamargo> http://pastebin.com/jwVLUkmj
[11:41:16 CET] <jkqxz> I'm afraid you've been deemed unworthy because you didn't give enough money to nvidia.  You need to have an "enterprise" card (i.e. one costing 10x as much) to encode more than two streams at once.have more than two streams
[11:41:48 CET] <acamargo> really?
[11:41:51 CET] <jkqxz> Yes.
[11:43:56 CET] <acamargo> so, is there a limit of processes to use gpu at same time?
[11:46:33 CET] <jkqxz> No, it's a limit on encode sessions specifically, independent of processes.  (Presumably enforced by the driver.)
[11:50:18 CET] <acamargo> :-( any official doc about that?
[11:51:49 CET] <DHE> https://devtalk.nvidia.com/default/topic/800942/session-count-limitation-for-nvenc-no-maxwell-gpus-with-2-nevenc-sessions-/
[11:52:27 CET] <DHE> It's a 2 session limit, but each session has no speed limit. If you're not doing realtime transcoding (ie. not over-the-air, but from files on disk) it probably won't matter too much
[11:58:57 CET] <acamargo> thank you so much guys, very helpful information ;-)
[12:48:03 CET] <ChuckLePlant_> Hi
[12:48:18 CET] <ChuckLePlant_> I'm building ffmpeg with cuda support, and get an issue with name mangling
[12:48:33 CET] <ChuckLePlant_> http://pastebin.com/RAzDgYry
[12:49:56 CET] <ChuckLePlant_> What that says is that ffmpeg tries to find _cuInit (unmangled).. and cuda.lib defines _cuInit at 4 (mangled with stdcall convention)
[12:50:37 CET] <ChuckLePlant_> In cuda.h CUDAAPI is defined as stdcall if _WIN32 is defined, So I define it like this:
[12:50:44 CET] <ChuckLePlant_> --extra-cflags="-D_WIN32 -IC:\\cuda_8\\include"
[12:50:59 CET] <ChuckLePlant_> (Full configure line here: http://pastebin.com/dMDRJNnA)
[13:17:42 CET] <ChuckLePlant_> What is the cuda version compatible with ffmpeg?
[14:03:54 CET] <Darby_Crash> hi guys. I'm working on Debian Jessie and i have wrote this script to compile a static build of ffmpeg using only repository for the libs https://clbin.com/UTclp
[14:04:52 CET] <Darby_Crash> i have a issue, libx265 and librtmp not found using pkg-config
[14:05:33 CET] <Darby_Crash> Can someone help me?
[14:09:57 CET] <Darby_Crash> pkg-config --list-all | grep rtmp show me right result and also x265
[14:10:28 CET] <Darby_Crash> where is my bad?
[14:11:03 CET] <jkqxz> Look in config.log for the actual error.
[14:11:57 CET] <Darby_Crash> collect2: error: ld returned 1 exit status
[14:11:57 CET] <Darby_Crash> ERROR: x265 not found using pkg-config
[14:12:27 CET] <furq> look higher
[14:13:39 CET] <Darby_Crash> https://clbin.com/fIp8B this is my config.log
[14:14:05 CET] <furq> /usr/bin/ld: cannot find -lnuma
[14:15:18 CET] <Darby_Crash> thanks furq, how can i fix this?
[14:15:42 CET] <furq> install libnuma?
[14:18:27 CET] <Darby_Crash> you're right, i haven't it
[14:18:37 CET] <Darby_Crash> thanks mate ;)
[14:25:52 CET] <Darby_Crash> furq i have fixed for libx265 but librtmp don't work
[14:26:23 CET] <furq> librtmp sucks anyway
[14:26:32 CET] <furq> ffmpeg's native rtmp stuff works fine
[14:28:29 CET] <Darby_Crash> is native library equal?
[14:30:05 CET] <Darby_Crash> https://clbin.com/MTsHk config.log
[16:03:16 CET] <TerNer> Guys, can I write encoded data to memory not a file?
[16:03:20 CET] <TerNer> in C
[16:06:44 CET] <JEEB> yes
[16:07:02 CET] <JEEB> if you want custom IO you can write an AVIO thing
[16:07:19 CET] <JEEB> you basically can implement your own read,write,seek etc
[16:07:58 CET] <JEEB> https://github.com/jeeb/matroska_thumbnails/blob/master/src/istream_wrapper.h
[16:08:48 CET] <JEEB> and then https://github.com/jeeb/matroska_thumbnails/blob/master/src/matroska_thumbnailer.cpp#L124..L137
[16:08:56 CET] <JEEB> this is my old code, but could be useful
[16:09:55 CET] <JEEB> this was for reading, but you can do the same for writing
[16:10:05 CET] <JEEB> (do note that APIs could have changed since, I have not updated that thing at all)
[16:19:51 CET] <DHE> http://ffmpeg.org/doxygen/3.1/avio_reading_8c-example.html This documents how to do it for reading. Gives you an idea what to look at in the docs for writing.
[16:20:55 CET] <TerNer> oh I got disconnected
[16:21:21 CET] <TerNer> Someone can let me know how can I write encoded data to memory not a file?
[16:22:13 CET] <JEEB> 17:08 < JEEB> https://github.com/jeeb/matroska_thumbnails/blob/master/src/istream_wrapper.h
[16:22:16 CET] <JEEB> 17:09 < JEEB> and then https://github.com/jeeb/matroska_thumbnails/blob/master/src/matroska_thumbnailer.cpp#L124..L137
[16:22:19 CET] <JEEB> 17:09 < JEEB> this is my old code, but could be useful
[16:22:25 CET] <JEEB> and DHE linked an example in the code
[16:22:26 CET] <JEEB> 17:20 < DHE> http://ffmpeg.org/doxygen/3.1/avio_reading_8c-example.html
[16:22:38 CET] <JEEB> both of these examples are for reading, but it's very similar for writing
[16:22:50 CET] <JEEB> you just define the write function instead of read
[16:22:58 CET] <TerNer> JEEB, thank you for your help
[16:24:11 CET] <TerNer> but JEEB, overriding writing function of avio is not providing "correct" encoded data
[16:28:28 CET] <JEEB> it shouldn't differ in any way, so your AVIO stuff is most likely incorrect then
[16:29:13 CET] <TerNer> JEEB, I just replaced some code of ffmpeg example
[16:29:39 CET] <DHE> keep in mind ffmpeg itself uses the same APIs for its own network and file access
[16:32:02 CET] <TerNer> DHE: There's no way to write in memory?
[16:32:54 CET] <DHE> not built-in. you'll have to provide your own avio functions
[16:33:00 CET] <DHE> similarly to the example for reading
[16:33:52 CET] <TerNer> oh
[19:21:46 CET] <zamba> hi! i'm in the process of digitalizing a bunch of old video cassettes.. for this i'm using ffmpeg.. and the following options: -f video4linux2 -channe 1 -i /dev/video0 -f alsa -i default -vf crop704:556:4:6,yadif,hqdn3d,format=yuv420p ..
[19:22:01 CET] <zamba> on the status output i'm seeing dup=2 and drop=4.. what does this mean?
[19:22:07 CET] <zamba> could that case to A/V desync?
[19:44:42 CET] <Darby> hi guys i have a working script to compile ffmpeg in static mode but resulted binary required shared library
[19:45:35 CET] <furq> which shared library
[19:47:52 CET] <Darby> https://clbin.com/aR7ae hi furq
[19:49:17 CET] <Darby> script is this https://clbin.com/d3g0s
[19:50:15 CET] <furq> looks ok to me
[19:50:17 CET] <furq> did you run make clean
[19:51:04 CET] <Darby> but i want run it on another machine
[19:51:28 CET] <furq> oh
[19:51:35 CET] <furq> --extra-ldflags="-static"
[19:52:24 CET] <Darby> only this furq?
[19:52:53 CET] <furq> should be
[19:52:58 CET] <furq> you can check what these are using
[19:53:04 CET] <furq> https://www.johnvansickle.com/ffmpeg/
[19:57:11 CET] <Darby> i have compiled it from myself because i need something of more
[19:57:27 CET] <furq> i mean you can check what config flags those builds are using
[19:57:51 CET] <Darby> ahhhh thanks :)
[19:57:53 CET] <furq> i forget if you need -static-libgcc -static-libstdc++ on linux
[19:58:36 CET] <Darby> what must i add?
[19:58:53 CET] <Darby> both -static-libgcc -static-libstdc++
[19:58:58 CET] <furq> i don't know
[19:59:15 CET] <furq> you need those for mingw, i don't remember if you need them for linux
[19:59:23 CET] <furq> those go in --extra-ldflags
[20:08:56 CET] <thebombzen> Darby: furq: it's generally not recommended to compile -lc or -lstdc++ statically
[20:09:22 CET] <furq> it is on mingw
[20:09:22 CET] <thebombzen> because every Linux system is going to have libc and libstdc++, and compiling those statically makes it less portable
[20:09:38 CET] <thebombzen> same with -lz and -lm etc.
[20:09:46 CET] <thebombzen> well yea on mingw
[20:09:48 CET] <furq> and i've never bothered building static for linux because distro packages are fine
[20:09:59 CET] <furq> hence my advice to check the builds of someone who knows what they're doing
[20:10:05 CET] <thebombzen> building static for linux seems silly to me tbh
[20:10:21 CET] <furq> it makes sense if your distro isn't as good as mine
[20:10:33 CET] <thebombzen> "distro isn't as good as mine"
[20:27:42 CET] <relaxed> I started the linux static builds because people used to come in here all the time with an ancient version trying to get support.
[20:53:25 CET] <dl2s4> hi, best command to downsample a flac 24bit to 16bit. best compression?
[20:58:29 CET] <relaxed> dl2s4: try, ffmpeg -i input.flac -sample_fmt s16 -c:a flac -lpc_type 3 -lpc_passes 8 -lpc_coeff_precision 0 -prediction_order_method 5 -min_partition_order 1 -max_partition_order 8 output.flac
[21:00:22 CET] <relaxed> that gave me the best compression from a particular sample, your results may vary
[21:01:11 CET] <furq> is that better than -compression_level 12
[21:02:09 CET] <relaxed> hmm, "ffmpeg -h encoder=flac" doesn't list -compression_level so I never tried it
[21:03:24 CET] <furq> apparently it is
[21:04:04 CET] <furq> with that said the difference between no options other than -c:a flac and flac -8 is a couple of kbps
[21:04:41 CET] <relaxed> Yeah, I just checked the source
[21:06:11 CET] <dl2s4> thanks
[21:06:43 CET] <relaxed> from what I remember ffmpeg's flac gave me better compression than `flac`
[21:07:02 CET] <furq> yeah it does
[21:07:07 CET] <furq> it's not enough to write home about though
[21:08:44 CET] <Darby> thanks thebombzen, have you some help for me?
[21:08:46 CET] <furq> i wonder if those settings are generating non-subset files
[21:08:48 CET] <dl2s4> oh why ffmpeg is better?
[21:10:07 CET] <relaxed> -compression_level 12 compressed better than my command :)
[21:24:16 CET] <blue_misfit> hey guys, any advice on using -shortest? I've got a video that has 60ms more audio than video and I want to clip it down to make the audio match the video
[21:24:38 CET] <blue_misfit> when I add -shortest prior to the output options I still get an output with an audio track that's 60ms longer than the video
[22:30:00 CET] <Darby> furq now don't compile more
[22:30:36 CET] <durandal_1707> ?
[22:31:06 CET] <DHE> libavcodec: so there's an AVCodecContext field 'channels' that you should be able to set and force the number of output channels you want in your output, right? like if the input is surround but I request stereo, I get only stereo out. right?
[22:31:13 CET] <Darby> i have added only --extra-ldflags="-static" and now i have this message: /usr/bin/ld: cannot find -lchromaprint
[22:31:36 CET] <DHE> Darby: you need the ".a" version of libraries for static linking, not the .so files. a lot of distributions don't ship a full set
[22:32:16 CET] <Darby> DHE i use Debian Jessie
[22:32:21 CET] <durandal_1707> DHE: no, downmixing is made with swresample
[22:32:56 CET] <DHE> durandal_1707: this has worked in the past. but I've made a lot of code changes and maybe it's my fault.
[22:33:12 CET] <durandal_1707> it never worked
[22:33:25 CET] <c3r1c3-Win> Does the 'abuffer' filter still exist by that name?
[22:33:26 CET] <Darby> how can i compile a really static binary?
[22:33:48 CET] <Darby> without shared
[22:33:49 CET] <DHE> Darby: you need the .a files for all the features you're using. or you have to strip out features you can't find static components for.
[22:34:36 CET] <Darby> how can i use .a files?
[22:34:37 CET] <durandal_1707> c3r1c3-Win: why you need it?
[22:35:12 CET] <Darby> where can i found those files?
[22:35:49 CET] <c3r1c3-Win> durandal_1707: Yes. Trying to call it via avfilter_get_by_name("abuffer"), I get a NULL on return
[22:35:54 CET] <DHE> Darby: that's up to your distribution. for fedora and most other RPM based distro look for package names with "-static" in the suffix. eg: zlib-static
[22:36:23 CET] <Darby> many thanks DHE
[22:36:45 CET] <Darby> now i do a search
[22:37:12 CET] <durandal_1707> c3r1c3-Win: its still available
[22:38:04 CET] <durandal_1707> what ffmpeg version you use?
[22:38:40 CET] <c3r1c3-Win> durandal_1707: Hmm. . not sure. Letme find out (It's a precompiled build by the project head for Windows releases.
[22:39:21 CET] <c3r1c3-Win> avfilter.dll is version 6.73.100.0
[22:39:38 CET] <DHE> durandal_1707: just figured it out. I'm also feeding it AAC audio instead of AC3, and AC3 does support in-codec downmixing but AAC does not...
[22:40:11 CET] <furq> Darby: -dev packages on debian should have static libs
[22:40:22 CET] <furq> if they don't then you need to build the libs yourself or just build ffmpeg without them
[22:41:01 CET] <DHE> or ask someone who does have them to build you a static version. I have a gentoo system just for static building (gentoo has static libs for almost everything)
[22:41:28 CET] <Darby> furq i have installed libchromaprint-dev
[22:41:32 CET] <durandal_1707> c3r1c3-Win: what compile flags? perhaps it got disabled
[22:41:52 CET] <c3r1c3-Win> That I don't know. I'll see if I can find out.
[22:42:50 CET] <Darby> without --extra-ldflags="-static" i can compile ffmpeg, with it i have /usr/bin/ld: cannot find -lchromaprint
[22:43:13 CET] <durandal_1707> c3r1c3-Win: abuffer should be build unconditionally
[22:43:40 CET] <durandal_1707> meaning you dont need to register it
[22:44:13 CET] <c3r1c3-Win> That's what I thought, but when calling it I get NULL back.
[22:45:08 CET] <c3r1c3-Win> Everything in the code links and I get no compile-time errors (this is on Windows 10 running vs2015).
[22:50:20 CET] <durandal_1707> c3r1c3-Win: do you call avregister_filter_all before?
[22:51:23 CET] <c3r1c3-Win> No. The first call is to avfilter_graph_alloc (which clears), and then I call avfilter_get_by_name("abuffer").
[22:51:56 CET] <durandal_1707> call it first before anything else
[22:54:07 CET] <c3r1c3-Win> The build flags for ffmpeg are as follows:
[22:54:08 CET] <c3r1c3-Win> ./configure --enable-memalign-hack --enable-gpl --disable-doc --arch=x86_64 --enable-shared --enable-nvenc --enable-libx264 --enable-libopus --enable-libvorbis --enable-libvpx --disable-debug --cross-prefix=x86_64-w64-mingw32- --target-os=mingw32 --pkg-config=pkg-config --prefix="/home/me/packages/win64" --disable-postproc
[22:55:04 CET] <c3r1c3-Win> durandal_1707: I'll insert said call first. Thanks!
[22:57:51 CET] <c3r1c3-Win> durandal_1707: Is "avregister_filter_all" an actual function?
[22:58:43 CET] <durandal_1707> see doc/examples in source code
[22:59:30 CET] <durandal_1707> yes it is function
[22:59:40 CET] <durandal_1707> no arguments
[23:01:20 CET] <DHE> Darby: are you okay with a version of ffmpeg without chromaprint ?
[23:02:19 CET] <Darby> i have tryed to delete chromaprint but after i have issue with libass
[23:02:48 CET] <Darby> --extra-ldflags="-static" don't work for me
[23:02:56 CET] <furq> https://packages.debian.org/stretch/amd64/libass-dev/filelist
[23:03:00 CET] <furq> that should be fine
[23:03:22 CET] <furq> libchromaprint-dev doesn't have a static lib, so you should report that as a bug i guess
[23:04:12 CET] <furq> to debian, in case that wasn't clear
[23:04:22 CET] <Darby> i have it https://packages.debian.org/jessie/amd64/libass-dev/filelist
[23:05:51 CET] <Darby> collect2: error: ld returned 1 exit status
[23:05:51 CET] <Darby> ERROR: libass not found using pkg-config
[23:06:01 CET] <furq> what does config.log say
[23:06:38 CET] <Darby> ok i have erased chromaprint and tryed to compile again
[23:06:52 CET] <Darby> this is my config.log
[23:07:36 CET] <Darby> https://clbin.com/SOVuI
[23:08:25 CET] <furq> /usr/bin/ld: cannot find -lenca
[23:08:26 CET] <furq> /usr/bin/ld: cannot find -lfribidi
[23:08:31 CET] <relaxed> you're going to need to compile a static libass along with static libs it depends on
[23:08:48 CET] <Darby> https://packages.debian.org/jessie/amd64/libenca-dev/filelist
[23:09:09 CET] <Darby> here i can't find static library
[23:09:12 CET] <furq> there isn't one
[23:09:57 CET] <c3r1c3-Win> durandal_1707: Hmm... I just cloned the repo and did a search through all of the code and can't find anything that mentions avregister_filter_all
[23:09:59 CET] <Darby> ok chromaprint and libass i must compile from myself, right?
[23:11:01 CET] <durandal_1707> c3r1c3-Win: in doc/examples
[23:12:07 CET] <durandal_1707> c3r1c3-Win: avfilter_register_all
[23:12:23 CET] <c3r1c3-Win> durandal_1707: Searched there, nothing.
[23:12:46 CET] <c3r1c3-Win> Ah! Thank you!
[23:15:13 CET] <livingbeef> I'm converting image sequence into webm (vp9) and when I use -lossless 1 about 11 (from 36 total) frames at the end is dropped. Is that normal?
[23:16:11 CET] <livingbeef> Just ommiting `-lossless 1' fixes it, but I'd like the losslessness
[23:18:51 CET] <livingbeef> (ffmpeg 3.2.2, libvpx 1.5.0)
[23:24:52 CET] <c3r1c3-Win> durandal_1707: Thanks again!
[23:26:05 CET] <livingbeef> It seems like -quality best might work - after it finishes rendering at whopping 0.1fps
[23:30:33 CET] <livingbeef> though quality-wise, it's awful... if you know any fix, I'll appreciate it
[00:00:00 CET] --- Sat Feb 25 2017


More information about the Ffmpeg-devel-irc mailing list