[Ffmpeg-devel-irc] ffmpeg.log.20160912

[burek]

[00:25:45 CEST] <someone212> Hey, i am used to do HEVC encodes with libx265 and the -crf argument for quality. Now i wanted to give hevc_nvenc a try, but don't know how to achieve a similar behavior. How can i set such a constant quality factor for the nvidia encoder?
[00:27:44 CEST] <DHE> there doesn't seem to be a crf option. you can use -qp for roughly constant quality, or bitrate-targetting mode
[00:31:12 CEST] <TwinTailed> Hi. Why do I get an av_send_frame() error only when the packet is of H264 codec? If it was another codec (aac for example, or AVI), packet is written successfully. Here is my code: http://pastebin.com/W2C19xZK
[00:33:48 CEST] <DHE> no logs to the console?
[00:34:10 CEST] <someone212> Thanks for the answer DHE, sadly it doesn't recognize -qp either
[00:36:19 CEST] Action: DHE checks to see if he's got something mixed up.
[00:37:37 CEST] <DHE> huh... must be an x264-only option as well... I swear there's a way to do this but I can't find it..
[00:38:17 CEST] <someone212> Yeah i couldn't find any documentation on it, so that's exactly where i am at
[01:00:37 CEST] <Lokie> JEEB furq DHE thank you very much for the help today, I immediatelly went on re encoding everything again and forgot myself
[01:01:28 CEST] <TwinTailed> Why does avcodec_send_frame function fail on H264?
[01:03:28 CEST] <evilolive> Need to export a 1min test clip of my video since it is HUGE. What would I add to say this?
[01:03:31 CEST] <evilolive> ffmpeg -i source.avi -c:a copy -c:v copy out.avi
[01:03:54 CEST] <evilolive> uncompressed ofc!
[01:04:50 CEST] <someone212> -ss 5:00 -t 1:00 in front of "-i" would get you a 1min clip starting roughly at minute 5
[01:05:02 CEST] <evilolive> thank you someone212
[01:05:31 CEST] <someone212> i guess with uncompressed you mean codec copy (like you wrote it)
[01:05:52 CEST] <TwinTailed> FFMPEG community sucks hardly... Mail list, IRC, and even bug trackers didn't benefit me even 1 bit... what a disgrace
[01:05:59 CEST] <evilolive> yeah =) Silly me. But yes it is utvideo
[01:07:13 CEST] <someone212> Twintailed sometimes the problem is the question. You ask why a function fails on a codec. That's not much to go on at all.
[01:07:26 CEST] <TwinTailed> someone212: why?
[01:07:48 CEST] <TwinTailed> the function goes well on other codecs
[01:07:56 CEST] <TwinTailed> even video codecs
[01:08:07 CEST] <someone212> Also after you told everyone they "suck hardly" the chances are pretty low to get help
[01:09:39 CEST] <TwinTailed> doesn't matter, the chances were really low since the beginning
[01:10:33 CEST] <furq> i chose to read it as "ffmpeg community sucks? hardly!"
[01:12:15 CEST] <TwinTailed> its a pretty simple problem... decoding and encoding H264 frames, should be pretty basic as its probably the most used codec
[01:12:41 CEST] <TwinTailed> but since the documentation is not well descripted too, it makes it hard for newcomers
[01:12:52 CEST] <Chloe> TwinTailed: asking ''why does a funciton fail" with no more info isnt really very helpful
[01:13:19 CEST] <TwinTailed> Chloe: it returns error 22 (Invalid arguments)
[01:13:27 CEST] <TwinTailed> I sent the code in pastebin
[01:13:33 CEST] <TwinTailed> what is required too?
[01:13:33 CEST] <Chloe> well, then it probably means what it says
[01:13:35 CEST] <JEEB> TwinTailed: did you look at the decoding examples under docs btw?
[01:13:43 CEST] <TwinTailed> JEEB: Yes
[01:13:45 CEST] <JEEB> I think the decode video example should handle AVC
[01:13:56 CEST] <JEEB> that said it's been really a while since I did any lavf/lavc coding
[01:13:57 CEST] <TwinTailed> I tried transcoding.c
[01:14:01 CEST] <TwinTailed> it fails with avc
[01:14:10 CEST] <JEEB> transcoding can be funky, you just noted decoding
[01:14:21 CEST] <JEEB> or do you mean decoding with transcoding.c example fails?
[01:15:13 CEST] <TwinTailed> they use deprecated functions
[01:15:42 CEST] <TwinTailed> im trying to use av_send_packet/receive_frame combo
[01:17:05 CEST] <JEEB> not sure if any help at this point since it's from like 2013-14, but this is how my demuxing and decoding looked back then https://github.com/jeeb/matroska_thumbnails/blob/master/src/matroska_thumbnailer.cpp#L98
[01:17:34 CEST] <JEEB> (it uses a custom avio context for the IStream handling that Windows uses, which you cna ignore)
[01:18:04 CEST] <JEEB> should probably poke at it again to see how different the API is nowadays to actually use
[01:22:06 CEST] <TwinTailed> I'll try using the deprecated functions and see if they go well, and if they do, I'll submit a ticket, otherwise I'll just wait for someone to help
[01:34:08 CEST] <AnonBaiter> so I have these parameters:
[01:34:28 CEST] <AnonBaiter> -acodec adpcm_4xm -ac 2 -vn -f data -i (inputfile) -t 3 (outputfile)
[01:34:38 CEST] <AnonBaiter> but when I run these parameters I get this:
[01:34:47 CEST] <AnonBaiter> "Output file #0 does not contain any stream"
[01:34:48 CEST] <AnonBaiter> What should I do about this?
[01:34:59 CEST] <evilolive> Cant ffmpeg output my interlaced video as it was doing this? " ffmpeg -ss 7:30 -t 1:00 -i india_2001.avi -c:a copy -c:v copy out.avi " It became progressive.
[01:35:33 CEST] <evilolive> Just a copy of both the audio and video.
[01:35:46 CEST] <evilolive> the org file is in fact interlaced
[01:37:04 CEST] <evilolive> Going to later deal with the interlace using VapourSynth ... prolly QTGMC
[01:38:35 CEST] <AnonBaiter> I mean, how can I scan the codecs?
[01:40:58 CEST] <Chloe> AnonBaiter: ffprobe?
[01:43:03 CEST] <AnonBaiter> usage example?
[01:44:15 CEST] <Chloe> ffprobe filename
[01:47:06 CEST] <AnonBaiter> "Invalid data found when processing input"
[01:47:35 CEST] <Chloe> AnonBaiter: are you trying to ffprobe the samples which arent supported, because that wont work
[01:47:44 CEST] <AnonBaiter> wait
[01:47:57 CEST] <AnonBaiter> I'll upload the "unsupported" file
[01:51:18 CEST] <AnonBaiter> https://www.sendspace.com/file/26vf63
[01:53:26 CEST] <Chloe> AnonBaiter: I thought you were trying to ffprobe the at3/msf files
[01:53:48 CEST] <AnonBaiter> I have other priorities
[01:57:32 CEST] <Chloe> no idea what happned but that file crashed my hexeditor, my browser, and my irc client
[01:59:34 CEST] <AnonBaiter> huh?
[01:59:53 CEST] <AnonBaiter> oops
[02:00:44 CEST] <TwinTailed> Guys I am doing anything wrong here? It is failing to decode audio or video frames: http://pastebin.com/b7NMk7dC
[02:06:52 CEST] <TwinTailed> OK, I give up :) I have tried everything to decode/encode H264, its not working in anyway, I haven't been able to produce one successful H264 frame.
[02:20:15 CEST] <evilolive> can ffmpeg deal with interlaced video? I just did a -c:v copy on a video that was interlaced and it became progressive
[02:21:02 CEST] <DHE> usually the fact that it's interlaced is in the video stream itself, so copying it couldn't possibly change that
[02:22:42 CEST] <evilolive> DHE yeah would think so... I mean it is just a copy. Wonder if there is a flag that needs to be passed for TFF or BFF ?
[02:22:59 CEST] <DHE> what's the full commandline, including file types?
[02:23:34 CEST] <evilolive> ffmpeg -ss 7:30 -t 1:00 -i india_2001.avi -c:a copy -c:v copy out.avi
[02:25:05 CEST] <evilolive> As you can see I am just making a test clip that I will be using for VapourSynth.
[02:28:24 CEST] <DHE> avi? what codec is that?
[02:28:48 CEST] <DHE> actually, does a 1 minute segment starting from the beginning (omit -ss option) work?
[02:29:03 CEST] <evilolive> avi is a container.
[02:29:27 CEST] <DHE> that's why I asked what codec
[02:29:44 CEST] <evilolive>  Stream #0:0: Video: dvvideo, yuv420p, 720x576 [SAR 16:15 DAR 4:3], 25000 kb/s, 25 fps, 25 tbr, 25 tbn
[02:30:06 CEST] <evilolive> so dvsd it looks like
[02:32:09 CEST] <evilolive> Think the commercial name is DVCPRO. Stuff they use to use in old Video8 camcorders
[02:33:56 CEST] <evilolive> DHE, good point. Will try from the start. Maybe the TFF BFF info is lost if I just make a random clip...one sec
[02:37:00 CEST] <evilolive> DHE, nope no dice. Mediainfo saying this is now progressive...
[02:38:50 CEST] <c_14> Can you upload the output of an ffprobe on the input and the output to a pastebin service?
[02:39:17 CEST] <evilolive> sure
[02:45:00 CEST] <evilolive> c_14, http://paste.debian.net/817540/
[02:46:09 CEST] <evilolive> If I test both files in Mediainfo I can see more info.
[02:47:00 CEST] <evilolive> hmmm seems I am not the only one who has this issues... http://forum.doom9.org/showthread.php?t=165464
[02:50:57 CEST] <c_14> Hmm, no idea. Probably file a bug
[02:55:53 CEST] <evilolive> Looks interlaced still. Just passed it though QTGMC and that made it looks nice again. Seems the BFF/TFF flag was lost and mediainfo took for granted that it was progressive..idk
[04:02:16 CEST] <AnonBaiter> so, any ideas about this?
[04:02:40 CEST] <AnonBaiter> ffmpeg -acodec adpcm_ima_oki -ac 2 -f data -i [inputfile] -t 3 [outputfile]
[04:02:49 CEST] <AnonBaiter> I mean, how can I treat input file as raw dat?
[04:02:52 CEST] <AnonBaiter> *data?
[04:03:29 CEST] <AnonBaiter> especially if it's headerless?
[04:40:24 CEST] <relaxed> AnonBaiter: what do you know about the audio?
[04:47:00 CEST] <AnonBaiter> I can only know about the audio if I can figure out the right codec
[04:47:32 CEST] <AnonBaiter> sometimes you need to experiment with different things to see the right result
[04:49:03 CEST] <AnonBaiter> but first you need a headless audio file
[04:49:17 CEST] <relaxed> AnonBaiter: try something like, ffmpeg -acodec pcm_u16le -f u16le -ar 48000 -i audio ...
[04:50:58 CEST] <relaxed> add -ac 2
[04:53:51 CEST] <AnonBaiter> Well I got "Unknown audio codec" error when adding all that before -i
[04:53:57 CEST] <AnonBaiter> *"Unknown input format"
[04:54:11 CEST] <AnonBaiter> or maybe I'm just that incompetent
[04:59:37 CEST] <relaxed> you don't what created the audio?
[04:59:52 CEST] <AnonBaiter> that's what I'm trying to find out
[05:00:15 CEST] <AnonBaiter> the only hint so far are the audio information at the end of the file
[05:56:26 CEST] <s0126h> how do i find  bitrate of audio on  .mkv file
[06:23:18 CEST] <doc|work> hi, not sure this is on-topic but if if it is, let me know. Hopefully someone can help as #gstreamer is dead at the moment. I've written a python+gstreamer script which is supposed to play an mp4 file. When I run it I get http://paste.ubuntu.com/23167371/
[06:23:47 CEST] <doc|work> anyone got any ideas what I might be missing?
[06:24:24 CEST] <doc|work> I've done a lot of searching and everything I can find to help debug the problem seems to be fine
[09:52:10 CEST] <sim590> I'm transcoding some videos (http://ix.io/1maf). The initial quality is really good, but the result is bad. I lost quality. My command is: `ffmpeg -i %in -y -vcodec libvpx -acodec libvorbis -cpu-used 16 -threads 8 -f webm %out`
[10:10:20 CEST] <Spring> sim590, you'll have to specify some more quality options to get good results from vpx
[10:10:54 CEST] <mdoe> Greetings. I've discovered a small problem in the source code for one of the encoders for ffmpeg, who would I talk to about that?
[10:10:59 CEST] <Spring> or indeed any quality options, see https://trac.ffmpeg.org/wiki/Encode/VP9 for examples
[10:11:53 CEST] <annoymouse4210> how can I split video clip with specific interval?
[10:12:15 CEST] <Spring> mdoe, some of the devs hang out here off/on, or you could file a bug ticket
[10:13:13 CEST] <annoymouse4210> solved.
[10:13:42 CEST] <mdoe> I suppose i could change it on git hub, it's one tiny bit that causes GOP drift when encoding with OMX :D
[10:13:52 CEST] <mdoe> probably a typo
[10:30:27 CEST] <sim590> Spring: Thanks. I'll check this out!
[10:35:03 CEST] <mdoe> I signed up to github and made a pull request.... first time for everything ^_^
[11:48:44 CEST] <brontosaurusrex> is vp9 concating basically working?
[11:52:04 CEST] <brontosaurusrex> Idea would be to split encoding to multiple parts and then concat the result into one file.
[11:54:41 CEST] <BtbN> That's not up to the codec
[11:55:16 CEST] <brontosaurusrex> BtbN, thats on container level? (webm)
[12:23:52 CEST] <damian_> Hi. When I try to add a watermark to a video using command described here: https://ffmpeg.org/ffmpeg-filters.html#Examples-67 it works but... the watermark is PNG with transparency and the transparency is visibly lower (less transparent watermark) when the source video is h265 or prores. Is that some bug in ffmpeg?
[12:24:13 CEST] <damian_> I'm using ffmpeg v 3.1.2
[12:25:12 CEST] <damian_> So: h264 source - watermark OK. h265/prores - watermark has too low transparency.
[13:15:43 CEST] <Locoloco> Hi! I found a package providing FFmpeg as a shared library for a closed source program. That library is provided without any license file. I am sure this is a mistake (the license files are completely empty but some have custom names). Should I bother? Should I try to contact the guy or am I not the "right" person to do so (not a FFmpeg contributor)?
[13:16:40 CEST] <Locoloco> I tried to reach the guy through his bug tracker but it seems to be disabled...
[13:21:05 CEST] <Spring> there should be an ffmpeg hall of shame :p
[13:23:34 CEST] <Locoloco> There used to be one, wasn't it?
[13:50:59 CEST] <nutopia> [mpeg2video @ 0536aa60] concealing 160 DC, 160 AC, 160 MV errors in I frame
[13:51:21 CEST] <nutopia> DC = discontinuities?
[13:51:28 CEST] <nutopia> what are AC and MV ?
[13:57:34 CEST] <jkqxz> DC and AC refer to the transform coefficients (many codecs encode the DC and AC parts differently in the bitstream, so they are reported separately).  MV refers to the motion vectors used for inter prediction.
[14:00:16 CEST] <jkqxz> So that messsage is really just telling you that there were 160 macroblocks in that frame which were totally broken.
[14:19:28 CEST] <nutopia> thanks jkqxz
[16:58:46 CEST] <mcjack> Hi guys, I am using ffmpeg now successfully on mac, but I'm having issues on windows
[16:58:59 CEST] <mcjack> configure says compiler test failed
[16:59:07 CEST] <mcjack> using msys shell
[16:59:27 CEST] <mcjack> the compiler is called correctly, but link is missing libraries
[16:59:39 CEST] <mcjack> libcmt.lib and kernel32.lib
[17:00:00 CEST] <mcjack> they are in different paths, but I don't get them both stuffed into the LIB variable
[17:00:26 CEST] <mcjack> tried with quoted and semicolon between, which invalidates the whole path
[17:01:03 CEST] <mcjack> or with colon, that's how the lionk.exe expects it, but that doesn't work either
[17:01:30 CEST] <mcjack> can anybody tell me, how is the correct way to add several lib paths for link.exe?
[17:09:47 CEST] <mcjack> or maybe another way around the compiler test
[17:18:34 CEST] <JonG> I have been running an ffmpeg binary in my Android app for a while now without issue. My Nexus 5X has recently been updated to Android Nougat and I have started experiencing a problem. FFmpeg exits immediately when running even a simple command. It gives an error code: 134. What is error code 134 (or where can I see all the error codes)? Even with
[17:18:34 CEST] <JonG> -loglevel debug I am given no further information, not even the ffmpeg build details that normally appear.
[17:21:04 CEST] <jkqxz> 134 is > 128, so it's a signal.  Assuming Android has the same signal numbers as normal Linux (?), that's SIGABRT.
[17:21:34 CEST] <jkqxz> Try running it in a debugger and see where that happens - it seems likely that it's going to be somewhere before anything in ffmpeg even runs.
[17:24:42 CEST] <iive> jkqxz: android runs linux kernel, so the signal must be the same.
[17:27:49 CEST] <JonG> OK, thanks for that info. I use a compiled ffmpeg binary, so I can't really use a debugger on ffmpeg itself, can I?
[17:32:05 CEST] <jkqxz> You can.  Even if it doesn't have symbols, you would get some sort of stack trace.  It's entirely possible that it dies before ffmpeg main() runs at all, anyway, and that would be visible.
[17:34:42 CEST] <jkqxz> iive:  Signal numbers in Linux are not consistent across platforms, so I don't see any reason for them not to be different in Android.
[17:35:16 CEST] <iive> hum... really?
[17:43:36 CEST] <iive> doesn't look to be true for 4.7 kernel. all signals are unconfitionally defined to generic values. also posix standard.
[17:45:57 CEST] <iive> hum... no i missed some files. there are other defines for other archs
[17:57:01 CEST] <pgorley> hi all, i have an rtp stream on which i call avformat_find_stream_info, it worked well on 2.6, but i'm updating to 3.1 and now it takes all 30 seconds (my AVFormatContext->max_analyze_duration)
[17:57:07 CEST] <pgorley> any clues as to why this is?
[17:57:13 CEST] <pgorley> using udp, if it helps
[18:16:01 CEST] <mcjack> ok, got the windows build working now, instead of calling all the vsvars.bat etc I just opened the msys shell from within a msvc build shell, so all is set up correctly for configure and make&
[18:16:24 CEST] <mcjack> thanks anyway
[18:16:45 CEST] <mcjack> &for listening
[19:12:10 CEST] <rngkll> hi, Im looking for information on how to generate a dash  .mpd file with ffmpeg, is this possible?
[19:12:50 CEST] <JEEB> yes, it has a  basic MPEG-DASH muxer that also writes a simple manifest
[19:12:59 CEST] <JEEB> just set output file name to something dot mpd
[19:13:22 CEST] <JEEB> it will then output the segments as well as the manifest
[19:14:10 CEST] <rngkll> in mi command I have -f webm_dash_manifest -live 1 -adaptation_sets "id=0,streams=0,1" master.mpd
[19:14:23 CEST] <JEEB> oh, webm dash
[19:14:24 CEST] <JEEB> no idea about that
[19:15:07 CEST] <JEEB> google provided some WebM DASH patches but I think those only covered the ISOBMFF DASH muxer
[19:15:16 CEST] <JEEB> the webm one probably just outputs segments
[19:15:30 CEST] <JEEB> but feel free to try
[19:15:46 CEST] <JEEB> documentation is your friend https://www.ffmpeg.org/ffmpeg-all.html
[19:16:16 CEST] <rngkll> thanks, I prefer dashif compatible but for the moment I only found info ron webm
[19:16:26 CEST] <rngkll> *for webm
[19:19:14 CEST] <rngkll> thanks
[20:08:01 CEST] <fred1807> if I use this https://pypi.python.org/pypi/ffmpeg-normalize  to update the movie file, we are talking about lossy rencode ?
[20:10:51 CEST] <kepstin> fred1807: of the audio, yes. you can stream-copy the video, of course.
[20:11:10 CEST] <fred1807> ok
[20:11:45 CEST] <kepstin> (i haven't looked at the code of that python thing, so I don't know if it actuall *does* stream copy the video or not)
[20:22:16 CEST] <ferdna> how do i copy the stream directly?
[20:22:28 CEST] <ferdna> i need to add this to my ffserver configuration
[20:22:40 CEST] <ferdna> i can do it with ffmpeg no problem... but i need it to be done in my ffserver
[20:25:54 CEST] <ferdna> can i do,., VideoCodec copy?
[20:28:53 CEST] <kepstin> no idea. Current versions of ffmpeg don't even have ffserver any more :)
[20:30:49 CEST] <ferdna> kepstin, yeah i know about that... =\
[20:30:50 CEST] <ferdna> https://lists.ffmpeg.org/pipermail/ffmpeg-user/2013-March/014172.html
[20:30:55 CEST] <ferdna> but i found my own answer
[20:36:30 CEST] <furq> kepstin: oh nice
[20:36:33 CEST] <furq> did it finally get the chop
[20:36:44 CEST] <kepstin> oh, I thought I heard it had?
[20:36:49 CEST] Action: kepstin hasn't actually looked :)
[20:37:41 CEST] <kepstin> hmm, I guess it isn't actually gone yet.
[20:38:02 CEST] <furq> looks like it'll be gone in 3.4
[20:38:02 CEST] <furq> er
[20:38:03 CEST] <furq> 3.2
[22:34:25 CEST] <xjrn> where is the legend for     ..FVA...
[22:36:12 CEST] <kepstin> xjrn: if you scroll up, it should be at the top of the list :)
[22:37:41 CEST] <xjrn> that i dont find
[22:37:55 CEST] <kepstin> hmm, what is that output from? it doesn't look like the codecs or filters list
[22:38:05 CEST] <xjrn> abench AVOptions:   action            <int>        ..F.A... set action (from 0 to 1) (default start)
[22:38:26 CEST] <xjrn> -filters i do have legend for.  for -h encoder= no
[22:38:30 CEST] <xjrn> err filter=
[22:39:02 CEST] <xjrn> looks like Encoder Decoder File Video Audio ? ? A=? ?
[22:39:15 CEST] <xjrn> looks like Encoder Decoder Filter Video Audio ? ? A=? ?
[22:41:43 CEST] <kepstin> on the parameters, it's "Encoding, Decoding, Filtering, Video, Audio, Subtitle, Export, Readonly"
[22:41:47 CEST] <kepstin> not sure what they all mean
[22:42:41 CEST] <kepstin> the same code is used to print avoptions from filters and encoders/decoders
[22:44:14 CEST] <xjrn> in filters, Export shows up as A
[22:45:51 CEST] <kepstin> hmm, I haven't seen any filters with anything in the second-last spot. which one are you looking at?
[23:18:48 CEST] <kode54> JonG had a known issue with FFmpeg on Android
[23:19:40 CEST] <kode54> Nougat doesn't like text relocations
[23:20:05 CEST] <kode54> attempting to dlopen a .so which contains any will result in an error
[00:00:00 CEST] --- Tue Sep 13 2016