[Ffmpeg-devel-irc] ffmpeg.log.20160506

[burek]

[00:19:10 CEST] <vade> ah, my source buffer was fucking things up, not the destination. working
[00:53:48 CEST] <bostonmacosx> hello there
[00:56:50 CEST] <durandal_1707> hello
[00:57:12 CEST] <rsully> furq, relaxed: if either of you are around, I'm back, trying to figure out these next errors from building statically. this happens during `make` step: http://pastie.org/private/2ywrurvsk70pmmhs9mfmxq
[00:58:47 CEST] <furq> rsully: fwiw i'm just googling the missing symbol names
[00:58:50 CEST] <furq> apparently this one is from glib
[00:59:29 CEST] <furq> i've static linked harfbuzz without static linking glib, so i guess your harfbuzz is compiled with weird options
[00:59:39 CEST] <rsully> default on ubuntu 16.04
[01:00:01 CEST] <furq> well yeah mine was cross-compiled so i had to do it myself, so presumably it avoids the dependency
[01:00:19 CEST] <rsully> so I need to compile harfbuzz then?
[01:01:17 CEST] <furq> do you have libglib2.0-dev installed
[01:01:27 CEST] <furq> that should hopefully fix it
[01:02:01 CEST] <rsully> Nope, installing now
[01:05:59 CEST] <rsully> No difference
[01:07:42 CEST] <bostonmacosx> So I'm trying to get my head wrapped around HLS
[01:08:20 CEST] <bostonmacosx> To do a multi bit rate delivery I would need to enocode multiple streames at different bit rates and then server then up with 1 m3u8 file
[01:11:21 CEST] <bostonmacosx> to rule them all...
[01:11:47 CEST] <bostonmacosx> I'm thinking of using CLAPPR as my web player since it seems to understand HLS playback..I'm not sure what else does
[01:12:08 CEST] <rsully> furq going to try to build harfbuzz with --enable-static
[01:24:15 CEST] <rsully> furq how important is graphite2 support for harfbuzz / ffmpeg?
[01:28:17 CEST] <furq> no idea
[01:29:01 CEST] <rsully> lots of warnings, but it didn't fail build this time (with harfbuzz built without graphite2 support, statically)
[01:29:07 CEST] <rsully> (and rebuilding libass against it)
[01:29:26 CEST] <rsully> aaand ffmpeg seg faults
[01:30:27 CEST] <rsully> how do I even debug this
[01:30:39 CEST] <rsully> I have a core file in cwd
[01:33:07 CEST] <rsully> meh no debug symbols so this is useless
[01:50:47 CEST] <rsully> umm compilign --with-debug but no debugging symbols found by gdb???
[01:52:30 CEST] <klaxa> did you use ffmpeg_g ?
[01:52:47 CEST] <rsully> no, I just found that mentioned in mailing list. I will try now.
[01:53:31 CEST] <c_14> You can also --disable-stripping, but honestly just use ffmpeg_g
[01:53:41 CEST] <rsully> yep understood.
[01:53:51 CEST] <rsully> gotta recompile because i cleaned too early
[08:12:56 CEST] <twerd> Hello, is there a command to encode a video with a gop of IBBP where each B and P frames DO NOT include I macroblocks?
[08:14:14 CEST] <twerd> We would like to have motion vectors available for every macroblock. Alternatively, what code file is responsible for deciding whether a macroblock should use MV or not?
[10:19:22 CEST] <Fyr> guys, what is the structure of chapter file for FFMPEG?
[10:19:31 CEST] <Fyr> I can't google out an example.
[10:49:45 CEST] <CoJaBo> So, if I upgrade to the latest ffmpeg, I get the error "[concat @ 0x4676120] Unsafe file name "; it seems to be suggested to use -safe, but that just results in a different error, "/dev/fd/63: could not find codec parameters". Ideas?
[11:10:34 CEST] <JoshX> Hi all!
[11:11:03 CEST] <JoshX> with the matroska container (amongst others) it is possible to store the system time in each the videostream as well..
[11:11:21 CEST] <JoshX> so you can  get the absolute frame time per frame, but also the system time of the recording for each frame
[11:11:31 CEST] <JoshX> is it possible to enable this with ffmpeg?
[11:11:38 CEST] <Fyr> maybe, as an attachment
[11:12:09 CEST] <JoshX> how would i enable this on the command line? I know some axis camera's can produce a mkv file with this setting enabled
[11:12:18 CEST] <JoshX> and i'm using the vi3o python package..
[11:12:34 CEST] <JoshX> and i'm very close to mimiccing this behaviour using a generic cam and ffmpeg
[11:12:51 CEST] <JoshX> (and the recording systems timesource as the time)
[11:13:09 CEST] <JoshX> it should return an epoch unix timestamp per frame
[11:13:29 CEST] <JoshX> and it does, only the start time is 0, so i'm getting 'system timestamps' from 1970-01-01
[11:13:59 CEST] <JoshX> so ffmpeg should insert the system time every x frames, ususally once per second, in the mkv stream as a unix expoch
[11:14:06 CEST] <JoshX> would this be possible?
[11:22:31 CEST] <JoshX> Fyr: do you have and ideas?
[11:28:10 CEST] <Fyr> JoshX, do you need to burn time stamp every frame?
[11:28:37 CEST] <Fyr> somehow, at the bottom-right corner?
[11:28:40 CEST] <JoshX> the axis cam does it once per second
[11:29:08 CEST] <Fyr> or are you trying to set the timeline at 1970-1-1?
[11:40:47 CEST] <JoshX> Fyr: i want the system unix timestamp in the videostream once per second or so
[11:41:01 CEST] <JoshX> *would like ;)
[11:41:16 CEST] <Fyr> oh, I don't know, sorry.
[11:50:05 CEST] <JamJams> Since ->codec->framerate.num is gone, what should we use as a replacement. ->codecpar-> doesn't have the framerate element?
[16:52:05 CEST] <neuro_sys> any pointers on using ffmpeg api (libav) to feed in-memory images as inputs and as video output?
[16:52:35 CEST] <Prelude2004c> --> /home/ffmpeg-3.0.1/ffmpeg -i - -codec copy -f mpegts - 2>/var/log/channels/${CHN}/input.txt < "$fifo" & encode+=($!)|/root/sandro/FFNVTranscoder/NvTranscoder -size 1280 720 -i /dev/stdin -o /dev/stdout -bitrate 1800000 -vbvMaxBitrate 2500000  -vbvSize 1000000 -deviceID $2 $NVENC_OPTS |/home/ffmpeg-3.0.1/ffmpeg -fflags +genpts -i - -c:v copy -c:a aac -b:a 128k -ac 2 -timestamp now -f hls -hls_time 6 -hls_flags delete_segments -hls
[16:52:35 CEST] <Prelude2004c> _list_size 5 ${HLS_PATH}/${CHN}/${CHN}2M.m3u8 2>/var/log/channels/${CHN}/720p.txt    ( Anyone know why i get all the garbage data on the screen ?? ) .. am i not pushing something to /dev/null ?
[17:01:51 CEST] <DHE> neuro_sys: there's a doc/examples directory with code examples, including some that generate their own images algorithmically
[17:02:53 CEST] <neuro_sys> DHE: thanks, I'm checking (I was considering the option of using ffmpeg-cli to pipe-in image streams but whatevr)
[17:13:41 CEST] <Prelude2004c> hey.. anyone know what i am doing wrong ?
[17:13:54 CEST] <Prelude2004c> soon as i run that thing... i get nothing but the garbage from the actual stream on my linux shell screen
[17:14:11 CEST] <Prelude2004c> tried so many different ways of doing /dev/null to hide that.. but nothing works.. not sure wh
[17:14:13 CEST] <Prelude2004c> why*
[17:14:23 CEST] <Prelude2004c> is it because i am running with /bin/bash?
[17:16:01 CEST] <DHE> the very first ffmpeg outputs to stdout but you never piped that anywher
[17:16:13 CEST] <c_14> You mean stderr
[17:16:49 CEST] <Prelude2004c> i was trying to push to fifo file
[17:17:19 CEST] <Prelude2004c> wait.. that is only first part.. ok. let me see if i understand
[17:17:41 CEST] <Prelude2004c> --> /home/ffmpeg-3.0.1/ffmpeg -i - -codec copy -f mpegts - 2>/var/log/channels/${CHN}/input.txt < "$fifo" & encode+=($!)|/root/sandro/FFNVTranscoder/NvTranscoder . doesn't that mean i pipped it to the transcoder ?
[17:18:32 CEST] <c_14> no
[17:18:42 CEST] <Prelude2004c> ok well then that is where i am confused :)
[17:18:42 CEST] <DHE> no, because '&' is a command separator which runs the command on the left side in the background
[17:18:42 CEST] <c_14> The '&' makes it fork
[17:18:43 CEST] <Prelude2004c> ok...
[17:19:31 CEST] <DHE> so.. you're getting ffmpeg to convert the output to mpegts, which gets fed into an 'nvtransocder' (nvidia I assume) which then goes back into ffmpeg
[17:19:35 CEST] <Prelude2004c> ahhhhh ok.... so   do i put that fifo thing at the end of the line.. or do i just put the & encode ..
[17:19:35 CEST] <DHE> there is nvenc support in ffmpeg these days
[17:20:04 CEST] <Prelude2004c> ya i know... i am having problems with the decoding.. the nvtranscoder works well.. doesn't require vdpau
[17:20:15 CEST] <Prelude2004c> i tried to use vdpau... locks up the nvidia card.. for some reason randomly
[17:20:23 CEST] <Prelude2004c> this nvidia built library works very well
[17:20:31 CEST] <Prelude2004c> and does the decode/encode together on the ard
[17:20:33 CEST] <Prelude2004c> card*
[17:22:08 CEST] <Prelude2004c> so i get you 100% .. on the < "$fifo " & .. so. the question is... i can't move that fifo < fifo & encode to the nd because then it wont pipe to nvtranscoder only to the last session which is ffmpeg.. so.. that means.. do i just move the & encode+=($!) to the end ?
[17:22:58 CEST] <Prelude2004c> i mean how would the fifo know which param to use from fifo .. since the files were passed to varialbe encode ( if i am saying that right )
[17:24:17 CEST] <neuro_sys> I think this is the one of the few good tutorials on ffmpeg library? http://dranger.com/ffmpeg/
[17:25:46 CEST] <neuro_sys> which happens to be outdated...
[17:26:04 CEST] <Prelude2004c> ya.. ok i got it.. i removed the encode+=($!)
[17:26:15 CEST] <Prelude2004c> and the & was removed too
[17:26:29 CEST] <Prelude2004c> not sure if i did that right though
[17:26:32 CEST] <Prelude2004c> but it is working
[17:28:57 CEST] <neuro_sys> what is YUV exactly?
[17:31:30 CEST] <furq> https://en.wikipedia.org/wiki/YUV
[17:31:53 CEST] <neuro_sys> wow, this website looks good
[17:32:02 CEST] <DHE> those component cables used in old analog HD video? They're not actually red, green blue data
[17:35:20 CEST] <neuro_sys> http://www.amazon.com/Video-Demystified-Handbook-Digital-Engineer/dp/0750683953
[17:35:24 CEST] <neuro_sys> lLooks neat
[17:38:41 CEST] <neuro_sys> writing PNGs to disk takes so much time, I should definetely stream them in-memory to ffmpeg
[17:38:59 CEST] <neuro_sys> and in raw rgba
[17:43:47 CEST] <Prelude2004c> for files in '${HLS_SOURCE}/${CHN}/*|sort -R'; do echo $files;done .. what is wrong with this /
[17:43:55 CEST] <Prelude2004c> trying '' or ""
[17:44:26 CEST] <Prelude2004c> and tried for files in ${HLS_SOURCE}/${CHN}/*; do echo $files |sort -R;done
[17:45:01 CEST] <Prelude2004c> silly me.. nevermind
[17:47:26 CEST] <Prelude2004c> well i found the source of one of my errors..
[17:47:28 CEST] <Prelude2004c> for files in ${HLS_SOURCE}/${CHN}/*;do echo $files;done|sort -R|while read f;do /home/ffmpeg-3.0.1/ffmpeg -re -i $f
[17:47:47 CEST] <Prelude2004c> Parse error, at least 3 arguments were expected, only 1 given in string 'METV/advertise4.mov'
[17:48:09 CEST] <Prelude2004c> getting errors like that.. .. what is wrong with that and can it be imrpoved ?
[17:57:42 CEST] <Prelude2004c> very odd..
[17:57:48 CEST] <Prelude2004c> for files in ${HLS_SOURCE}/${CHN}/*;do echo "$files"|sort -R;done|while read f; do /home/ffmpeg-3.0.1/ffmpeg -re -i "$f" . gives me the parse error
[18:03:45 CEST] <UukGoblin> hello, I'm trying out the NVENC stuff. I've compiled ffmpeg successfully and I'm trying to recode a h264 file. I haven't yet figured out how to insert a scaling filter, but something seems wrong already. I'm running AV_LOG_FORCE_COLOR=yes ~/soft/ffmpeg-3.0.2/bin/ffmpeg -i ./orig_vid.h264 -c:v nvenc -profile:v high444p  -preset hq nvenc.h264 -v trace | aha
[18:03:53 CEST] <UukGoblin> here's the output: http://yat.uukgoblin.net/ffmpeg.html
[18:04:49 CEST] <UukGoblin> I've no idea what might be causing the 'EncodePicture failed' thing, does anyone have a clue?
[18:08:52 CEST] <Prelude2004c> UukGoblin > try running it without the profile .. any difference ?
[18:11:58 CEST] <UukGoblin> Prelude2004c, yes! thanks, I got the profile thing from https://trac.ffmpeg.org/wiki/HWAccelIntro
[18:12:06 CEST] <UukGoblin> works without it :-)
[18:13:01 CEST] <Prelude2004c> yup..
[18:13:07 CEST] <UukGoblin> so I added a scaler with -vf scale=1920x1080, but I believe this is a software scaler, right? Is there any way I could an accelerated one?
[18:13:55 CEST] <Prelude2004c> um... you can scale that way.. accelerated what do you mean ?
[18:14:29 CEST] <UukGoblin> actually I guess I'd need to decode h264 via the video card as well for it to make sense
[18:14:34 CEST] <Prelude2004c> -size 1920x1089
[18:14:41 CEST] <UukGoblin> Prelude2004c, like, have the gfx card do the scaling
[18:14:45 CEST] <Prelude2004c> sure.. you are playing in my pain in life :P haha.
[18:15:01 CEST] <Prelude2004c> ya if your using ffmpeg.. you can use vdpau to do it
[18:15:05 CEST] <durandal_1707> there are hw scaler filters
[18:15:40 CEST] <Prelude2004c> oh.. ya i am not too familiar with that.. haven't gotten that far with respect to scaling :)
[18:15:49 CEST] <Prelude2004c> i do have one question however.
[18:16:22 CEST] <Prelude2004c> ( last question of the day hopefully.. :) -- ( printf '%s\n' "${HLS_SOURCE}/${CHN}"/*|sort -R|while read f; do /home/ffmpeg-3.0.1/ffmpeg -re -i "$f" -codec copy ) what do you think? here is the odd part.. the files being passed to ffmpeg are still giving me this sort of error ( Parse error, at least 3 arguments were expected, only 1 given in string 'METV/advertise8.mov' ).. so two things why an error when i introce the while read.. a
[18:16:22 CEST] <Prelude2004c> nd lastly why cuttong off since the path for hte files is /mnt/extra/CMETV/advertise8.mov
[18:16:28 CEST] <UukGoblin> durandal_1707, could you point me to a name or sth? :-)
[18:17:17 CEST] <durandal_1707> search for hw in filter docs, you need recent ffmpeg
[18:17:42 CEST] <vade> interesting about the hw scaler - thats good to know
[18:17:56 CEST] <vade> durandal_1707: - is your handle a reference to marathon?
[18:18:04 CEST] <UukGoblin> durandal_1707, on https://ffmpeg.org/ffmpeg-filters.html, I can only find hw_frames_ctx and hwupload_cuda
[18:18:19 CEST] <durandal_1707> vade: haha
[18:20:50 CEST] <vade> :P
[18:22:20 CEST] <durandal_1707> UukGoblin: IIRC there are more, perhaps they are not documented, r
[18:23:43 CEST] <UukGoblin> Prelude2004c, -size is actually an unrecognized option ;-)
[18:26:53 CEST] <UukGoblin> I've recompiled with --enable-vdpau, but I'm not sure if it's actually using it for decoding. CPU usage stays pretty high, about half of each of 8 cores (@100 fps transcoding through nvenc)
[18:35:16 CEST] <UukGoblin> interesting, it looks like it's not possible to move the decoded frame back to RAM with VDPAU. So it'd /have/ to do all the filtering in hardware, if it's supposed to work at all. However, https://developer.nvidia.com/nvidia-video-codec-sdk mentions a NVCUVID API, for hardware video decoding. Maybe that'd be more suited for later (i.e. if I wanted to use some software-only filters)
[18:37:59 CEST] <durandal_1707> there are filters for this
[18:39:00 CEST] <UukGoblin> hm :-)
[18:40:37 CEST] <durandal_1707> what gives you ffmpeg -filters
[18:40:39 CEST] <durandal_1707> ?
[18:42:54 CEST] <UukGoblin> durandal_1707, nice, lots of things: http://pastebin.com/NAecmbHw
[18:58:39 CEST] <Prelude2004c> hey guys.. quick question.... is there some way to insert a png image logo on a video without transcoding it ?
[18:58:59 CEST] <Prelude2004c> -vf can't be used with copy. the video is already done.. and i don't want to encode it again.. but since i am creating the segment files.. i figure .. is there some way ? maybe not
[18:59:38 CEST] <vade> by definition youre changing the video content so it requires a decode - composite and re-encode
[19:00:05 CEST] <Prelude2004c> oh :( ... damn.. seesm nvtranscoder doesn't have an option for logo insert :(
[19:01:14 CEST] <vade> transcoding is decoding -> encoding. If you are segmenting many times you can just move raw packets from streams to one container to another witout transcoding, just fudging timestamps of the packets
[19:01:26 CEST] <vade> which are you doing?
[19:02:27 CEST] <neuro_sys> Do you think open source has similarities to communism as an economic model?
[19:03:02 CEST] <neuro_sys> sorry I have happy-friday induced over-moderation alcohol consumption today
[19:03:59 CEST] <furq> there are successful open-source projects, so no
[19:04:09 CEST] <vade> oh snap -
[19:04:10 CEST] <vade> haha
[19:08:14 CEST] <UukGoblin> neuro_sys, no, because copying software is nearly free, whereas copying resources isn't ;-)
[19:11:09 CEST] <c_14> >nearly free
[19:11:15 CEST] <c_14> You've obviously never sen the bit gulags, have you?
[19:11:33 CEST] <UukGoblin> no?
[19:12:23 CEST] <c_14> Forced labor camps hidden inside every computer.
[19:12:34 CEST] <c_14> Few breaks. No vacation. No amenities.
[19:12:58 CEST] <UukGoblin> millions of transistors kept there against their will?
[19:13:18 CEST] <Kadigan|Work> Hi. I just did a major update of my install base, which included ffmpeg 2.8.6. I previously used an ancient version that exposed MPlayer filters (I used mp=eq2 specifically). How can I replicate that functionality, as I need it?
[19:13:27 CEST] <UukGoblin> well, when that stuff realizes what we're forcing it to do, it ain't gonna be pretty
[19:13:55 CEST] <c_14> Kadigan|Work: the filters have been merged
[19:13:59 CEST] <c_14> eq2 is now just eq
[19:14:09 CEST] <c_14> eh, mp=eq2 is now just eq
[19:14:15 CEST] <Kadigan|Work> c_14: there is NOP, HLT and SLP ;) at least...
[19:14:25 CEST] <UukGoblin> ;-D
[19:14:44 CEST] <Kadigan|Work> (there's also a number of HCF instructions :D)
[19:15:23 CEST] <Kadigan|Work> Okay, it seems to work now. I guess it'll be a while before I find out if I still have the nonfree filters in.
[19:15:32 CEST] <Kadigan|Work> Thank You kindly :)
[19:21:34 CEST] <Kadigan|Work> Okay. Have the settings for eq changed? All I get now - using the same settings as I did for mp=eq2 - is all white video.
[19:28:26 CEST] <Kadigan|Work> Okay, seems gamma was moved from 1st to 4th place. (WTH ffmpeg-devs? why can't it be interop?)
[19:28:46 CEST] <Kadigan|Work> Anyway, thanks. Have a good one.
[20:05:28 CEST] <Prelude2004c> hum. whats the timeout param .. if file input times out after x amoutn of time just quite and go on
[20:05:38 CEST] <Prelude2004c> tried -timeout and its giving me incorrect option
[20:05:46 CEST] <Prelude2004c> even tried -stimeout
[21:05:42 CEST] <vade> question about the header docs for swr_convert_frame
[21:06:07 CEST] <vade> it mentions that I can pass in null AVFrame as input to get buffered converted output frames
[21:06:31 CEST] <vade> however, passing null says no matching function
[21:06:39 CEST] <vade> does this mean an empty frame?
[21:07:03 CEST] <vade> like one just frame_alloced but with no backing buffer from av_frame_get_buffer
[21:07:55 CEST] <vade> jesus christ. never mind
[21:08:02 CEST] <vade> <-idiot
[21:08:12 CEST] <vade> of course when you ask you realize your mistake
[21:27:04 CEST] <Fyr> guys, what is the best way to create silent audio?
[21:29:16 CEST] <durandal_1707> anullsrc
[21:29:23 CEST] <Fyr> thanks
[21:44:18 CEST] <OmegaVVeapon> guys, I'm trying to get ffmpeg to timeout when it listens on a Source Specific Multicast and gets no data
[21:44:28 CEST] <OmegaVVeapon> according to the documentation in https://ffmpeg.org/ffmpeg-protocols.html#udp
[21:44:42 CEST] <OmegaVVeapon> timeout= should do the trick
[21:44:58 CEST] <OmegaVVeapon> if I run "time ffmpeg -i udp://232.22.137.169:5698?timeout=5000000"
[21:45:08 CEST] <OmegaVVeapon> I see it exits in 5 seconds
[21:46:16 CEST] <OmegaVVeapon> however, switching to SSM. "ffmpeg -i udp://232.22.137.169:5698?sources=10.10.10.10&timeout=5000000" never exits...
[21:46:26 CEST] <OmegaVVeapon> what gives? :/
[21:52:42 CEST] <CoJaBo> So, if I upgrade to the latest ffmpeg, I get the error "[concat @ 0x4676120] Unsafe file name "; it seems to be suggested to use -safe, but that just results in a different error, "/dev/fd/63: could not find codec parameters". Ideas?
[21:53:43 CEST] <J_Darnley> I think you want "-safe 0", it needs an argument
[22:00:21 CEST] <CoJaBo> J_Darnley: Yeh, that was it; odd it can't give a clearer error message :/
[22:01:08 CEST] <J_Darnley> Well when it eats another option as its argument, ffmpeg doesn't know that.
[22:01:47 CEST] <J_Darnley> Although I might expect an error for something that isn't 0 or 1
[22:01:59 CEST] <J_Darnley> but I wonder whether you just ignored that.
[22:09:54 CEST] <OmegaVVeapon> Anyone with experience on using ffmpeg on multicast feeds?
[23:17:09 CEST] <vade> where does one submit API documentation bugs?
[23:24:02 CEST] <c_14> Same place you submit all other bugs, on trac
[23:24:11 CEST] <c_14> Or just send a patch to the ml
[00:00:00 CEST] --- Sat May  7 2016