[Ffmpeg-devel-irc] ffmpeg.log.20160711

[burek]

[00:01:33 CEST] <DHE> as two files, or in a format that doesn't care about that
[00:14:01 CEST] <jaggz> dhe, I'd like the final to be in .mp4
[00:14:24 CEST] <jaggz> I have a screen capture, and audio recording -- the audio is over an hour, the screencapture failed because my screensaver turned on at 10 minutes
[00:14:55 CEST] <jaggz> the screencap contains audio too, but I'm wondering if I can chop out the video from the 11th minute on to save filespace in the final vide
[00:17:44 CEST] <friki2015> Hi, i've encoded a video at 25fps. ffmpeg says: "25 fps, 25 tbr, 90k tbn, 50 tbc". Using some players (mplayer) it plays at 2x speed (50fps) but ffplay works fine
[00:18:09 CEST] <JEEB> try mpv and latest vlc rather than mplayer
[00:20:49 CEST] <friki2015> the video works fine with both players. May be a mplayer bug or a encoding/packaging problem?
[00:21:14 CEST] <JEEB> most probably mplayer just being mplayer
[00:21:46 CEST] <friki2015> :-)
[00:22:03 CEST] <JEEB> mpv and vlc are probably the two things you should generally test things with. mpv is the least retarded fork of mplayer that uses FFmpeg's libraries, and VLC generally uses something els.
[00:22:15 CEST] <JEEB> so you can thus test two ways of reading the files in many cases.
[00:22:50 CEST] <JEEB> http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/2015-September/073131.html
[00:23:03 CEST] <JEEB> this was a good post on original mplayer's mailing list from last year
[00:23:06 CEST] <friki2015> thanks for the tip. i was thinking that tbc=50 was the problem
[00:24:50 CEST] <friki2015> "MPlayer is still my favorite video player, and I really would hate to see it go down the drain." hehehe love this final words at the post
[00:31:52 CEST] <JEEB> friki2015: also since container timestamps generally win over anything in the actual stream, for matroska you can use mkvinfo and for mp4/mov type things you can use L-SMASH's boxdumper
[00:32:07 CEST] <JEEB> if you want to be ever stare at the timestamps saved in the actual container
[00:32:18 CEST] <JEEB> *if you ever want to stare...
[01:13:33 CEST] <haasn> furq: I remember wanting to show you screenshots of my wow interface. I managed to recover some from chat logs: https://i.imgur.com/PBeX3PJ.jpg https://i.imgur.com/0L24roU.jpg
[04:56:46 CEST] <Blakes5> Hello. I'm trying to setup a ffserver for a http stream. I have what looks like a working conf file and ffmpeg appears to be encoding the stream but I can't seem to access it using VLC. Help?
[05:00:20 CEST] <Blakes5> ffserver.conf is here http://pastebin.com/zJ0P32YK
[05:00:53 CEST] <Blakes5> This is the ffmpeg command ffmpeg -r 30 -s 752x416 -f video4linux2 -i /dev/video0 http://localhost:8090/office.ffm
[05:02:50 CEST] <Blakes5> http://localhost:8090/office.ffm is the address I'm using to try to access it using VLC.
[05:04:01 CEST] <Blakes5> I can see get requests.
[05:04:38 CEST] <Blakes5> [GET] "/office.ffm HTTP/1.1" 404 158 I'm guessing the 404 means ffserver can't find the stream?
[05:05:25 CEST] <Blakes5> I don't understand.
[08:47:19 CEST] <antiPoP> hi, I'm trying to run ffmpeg and I get "ffmpeg: error while loading shared libraries: libx264.so.148: cannot open shared object file: No such file or directory"
[10:39:03 CEST] <asio> I've got 2 movies: movie1.mp4 and movie2.mp4, that I want merged together but one was filmed during sunshine and other when cloudy, is it possible to adjust them by colorbalance/brightness to so one matches the other?
[11:26:56 CEST] <sybariten> oh hai
[11:27:17 CEST] <sybariten> could someone please explain to me the concept of separate input and output frame rates when makeing videos from sequences of still images
[11:31:58 CEST] <spirou> hi
[11:32:33 CEST] <spirou> from still images, there would be no input frame rate would it?
[11:33:26 CEST] <furq> the input framerate is how long you want each image to appear for
[11:33:38 CEST] <furq> the output framerate is what you think it is
[11:34:08 CEST] <spirou> but I assume you can tell ffmpeg about what rate those images come "these are 50 frame/s" and then tell that you want output frame "25 frame/s" and so it will skip all even frames :D
[11:34:20 CEST] <furq> you sure can
[11:35:50 CEST] <furq> it's more likely that you'd want something like -framerate 1 -i %03d.png -r 25
[11:36:10 CEST] <furq> which will duplicate frames
[11:57:31 CEST] <sybariten> furq: Hm. I'm sorry but could you explain this again, butt differenlty
[11:57:37 CEST] <sybariten> eh, BUT not butt
[12:20:53 CEST] <sybariten> it seems like they would control the same thing?
[12:34:07 CEST] <spirou> yeah you could se out framerate so each in-frame will be used multiple times in a video yeah
[12:34:23 CEST] <spirou> *set
[12:36:07 CEST] <spirou> if you want a 25fps video but don't really have that many pictures
[13:48:47 CEST] <Blakes5> I'm trying to setup a ffserver stream. It looks like it's working but I can't access the stream. Can someone help me pick through this? ffserver.conf http://pastebin.com/zJ0P32YK
[13:49:03 CEST] <Blakes5> This is the ffmpeg command ffmpeg -r 30 -s 752x416 -f video4linux2 -i /dev/video0 http://localhost:8090/office.ffm
[13:49:14 CEST] <Blakes5> http://localhost:8090/office.ffm is the address I'm using to try to access it using VLC.
[15:21:08 CEST] <istee> i am trying to convert a h264 file from a dvr cctv camera. it works fine until about five minutes into the video where the console is flooded with messages like
[15:21:11 CEST] <istee> [h264 @ 0x174df60] Cannot use next picture in error concealment
[15:21:14 CEST] <istee> [h264 @ 0x174df60] concealing 396 DC, 396 AC, 396 MV errors in P frame
[15:22:00 CEST] <istee> is there any way to get it to drop frames with these errors? it plays fine with mpv if you skip through it but it ends up freezing and then being just a gray color while converting with ffmpeg
[15:27:10 CEST] <bencoh> the error_concealment flag iirc?
[15:27:13 CEST] <bencoh> istee: ^
[15:29:12 CEST] <istee> i couldnt find any relevant flag @ http://ffmpeg.org/ffmpeg.html
[15:29:40 CEST] <istee> i found references to -ec/-er but they're deprecated?
[15:30:30 CEST] <istee> url was supposed to be http://ffmpeg.org/ffmpeg-all.html , oops
[15:31:44 CEST] <istee> nevermind, i put it after the input and it gave me a vague splitting error. -ec before works
[15:32:22 CEST] <bencoh> input-related options should always sit before input :)
[15:38:01 CEST] <istee> "-ec 0" just kept stuttering between the frame in a loop, not moving forward :(
[16:13:54 CEST] <istee> update: i tried several players and all of them do the same thing when playing the damaged section, requiring stopping said player and reloading. but it works fine when you seek past it
[16:14:35 CEST] <ChocolateArmpits> Did w3fdif get updated? Previously I remember it was slower than yadif, but now on an old two core system it's pushing over 300fps at complex while yadif manages only 220fps at send_field_nospatial
[16:14:39 CEST] <istee> however ffmpeg with both -ss after and before input causes the error
[16:17:15 CEST] <ChocolateArmpits> bwdif pushes around 260fps and looks to have better output than yadif or w3fdif
[16:17:56 CEST] <istee> https://paste.debian.net/hidden/77c0c9d0/ log while trying to play said section with mpv
[16:21:43 CEST] <ChocolateArmpits> istee: Are you looking for any solution ?
[16:23:03 CEST] <istee> i want to be able to transcode whats recoverable from this stream to a non-corrupt DVD
[16:24:08 CEST] <ChocolateArmpits> istee: Ok, pipe the bytes of the file to ffmpeg
[16:25:07 CEST] <istee> https://carl.moe/bad.mp4 i trimmed the section it chokes on 5:20 to be exact w/ -vcodec copy
[16:27:41 CEST] <istee> i can brute force recover by seeking before input and doing -vcodec copy for these sections to get them away but i would like a less hassle solution
[17:34:09 CEST] <aarrgg> Good morning :) I am struggling with FFmpeg, I have a persistent error -->Unknown input format: 'alsa'
[17:34:09 CEST] <aarrgg> ...and my extensive searches are not shedding light on my problem
[17:36:14 CEST] <aarrgg> help anyone?
[17:36:44 CEST] <outcast> hi is there a page with code snippets?
[17:36:59 CEST] <outcast> looking for a sample of how to do a progress bar in c#
[17:38:19 CEST] <DHE> aarrgg: need to compile with ALSA support. you may need to install a package like alsa-dev for your distribution
[17:39:30 CEST] <aarrgg> E: Unable to locate package alsa-dev
[17:41:31 CEST] <aarrgg> DHE , i'm using Jessie on a Raspberry Pi, I have not seen a specific example of compiling ffmpeg with ALSA support
[17:42:46 CEST] <DHE> officially you use ./configure --enable-indev=alsa  (and similar for outdev) but I'm guessing you lack the dev library.
[17:44:21 CEST] <aarrgg> apt-get install alsa-dev     is what I have tried, and thanks for the config line
[17:48:45 CEST] <debianuser> aarrgg: apt-get install libasound2-dev
[17:49:19 CEST] <debianuser> "libasound" is alsa :)
[17:49:54 CEST] <aarrgg> ill try a recompile with your suggestion, thanks.
[17:50:48 CEST] <aarrgg> debianuser: thanks, tried and already the newest version :)
[17:51:03 CEST] <outcast> is the forum in the topic the best place to ask a c# question
[17:55:30 CEST] <bencoh> I dunno if you'll find any C#/friendly dev (I'm not) but I guess you could always ask (provided it's ffmpeg-related ;)
[17:57:32 CEST] <aarrgg> its nearly 2AM in Australia, so i'm compiling and off to bed, thanks for your help, love linux but its documentation leaves a bit to be desired
[17:58:10 CEST] <furq> are you compiling ffmpeg on an rpi
[17:58:29 CEST] <aarrgg> yes on a Rpi
[17:58:33 CEST] <furq> that sounds like a miserable experience
[17:58:43 CEST] <viric> furq: much less noisy than what you are used to
[17:58:48 CEST] <viric> ;)
[17:59:06 CEST] <furq> i've got an rpi with ffmpeg on it
[17:59:10 CEST] <furq> but i compiled it on a real computer
[17:59:14 CEST] <aarrgg> not too bad, 15min? and not often
[17:59:49 CEST] <viric> with a good scheduler, you just queue the task with nice, and at some point it's done
[17:59:51 CEST] <furq> are you disabling everything or something
[18:00:06 CEST] <furq> that sentence is confusing
[18:00:30 CEST] <furq> pretend i didn't say "or something"
[18:00:40 CEST] <outcast> sorry phone rang
[18:01:16 CEST] <outcast> looking for way to use -progress i think it is to make a percentage bar
[18:01:35 CEST] <aarrgg> furq: ffmpeg has not been recognizing alsa....OR...you could recommend a better way to make video with sound :)
[18:01:59 CEST] <furq> there is no better way than to recompile ffmpeg
[18:02:11 CEST] <furq> but you can cross-compile it on a faster cpu
[18:02:34 CEST] <aarrgg> thanks, at least I'm not barking up the wrong tree :)
[18:02:43 CEST] <viric> can ffmpeg do something to record sound/video in sync, if it captures from independent devices? webcam, soundcard, ...
[18:02:52 CEST] <furq> it's slightly complicated if you need external dependencies but i assume debian multiarch handles that ok
[18:03:07 CEST] <furq> i didn't need to deal with it because i just wanted to see how bad omx's h264 encoder actually is
[18:03:10 CEST] <furq> answer: it's bad
[18:03:17 CEST] <viric> I have a sound recorder that deviates ~1s per hour
[18:03:42 CEST] <aarrgg> ....I'm in enough trouble as it is
[18:04:00 CEST] <viric> furq: ah I've never seen it. Worse than x264 ultrafast?
[18:04:28 CEST] <furq> i've not done a side by side comparison
[18:04:36 CEST] <furq> i'd guess yes though
[18:06:08 CEST] <furq> you can't really expect much from a £30 computer though
[18:06:33 CEST] <furq> or £9 if the pi zero has the same gpu
[18:21:49 CEST] <aarrgg> ...having some joy, thanks again, off to bed!
[19:37:59 CEST] <blue_misfit> hey guys - anyone familiar with HLS? I'm struggling to package a stream with 23.98 video while respecting #EXT-X-TARGETDURATION > #EXTINF. If I have 4 second chunks for example #EXTINF is usually 4.004 - presumably due to 23.98 vs 24. How can I fix? #EXT-X-TARGETDURATION is integer only
[19:41:23 CEST] <Bray90820> pts has no value is what does that error mean?
[19:41:28 CEST] <Bray90820> or warring either
[19:41:34 CEST] <Bray90820> *Rather
[19:44:39 CEST] <JEEB> blue_misfit: that's a bug in the muxer then
[19:44:58 CEST] <JEEB> post an issue on the tracker pointing towards https://tools.ietf.org/html/draft-pantos-http-live-streaming-19#section-4.3.3.1
[20:24:55 CEST] <wiak> vp9 at faster speed and higher bitrate or slower speed at lower bitrate?
[20:25:34 CEST] <wiak> -deadline realtime -speed 8 -b:v 6000K
[20:25:56 CEST] <wiak> for live streaming
[20:39:25 CEST] <ChocolateArmpits> wiak: Depends how many concurrent streams you need and how many you can sustain
[20:40:08 CEST] <wiak> nah with vp9 the issue lies with the slow encoder
[20:40:33 CEST] <wiak> encoding to 720p is like 36fps and the source is 30fps
[20:40:48 CEST] <wiak> vp9 has awful multicoresupport
[20:41:00 CEST] <ChocolateArmpits> If it's stable 36fps then I don't see no problem in livestreaming
[20:41:08 CEST] <wiak> true
[20:43:25 CEST] <ChocolateArmpits> Google/Youtube parallelizes their encoding by splitting the file into parts with each encoded separately. So for them multithreading isn't that vital.
[20:43:53 CEST] <ChocolateArmpits> I guess they might do the same for livestreams
[20:44:00 CEST] <ChocolateArmpits> or maybe not
[20:44:16 CEST] <ChocolateArmpits> nah scratch that, doesn't make sense for livestreams
[20:44:25 CEST] <furq> do they actually do that or is that just conjecture
[20:45:44 CEST] <ChocolateArmpits> furq: they do http://youtube-eng.blogspot.com/2016/05/machine-learning-for-video-transcoding.html
[20:45:48 CEST] <ChocolateArmpits> "The only way we can keep up with that rate of ingest and quickly show you your transcoded video in YouTube is to break each file in pieces called chunks, and process these in parallel. Every chunk is processed independently and simultaneously by CPUs in our Google cloud infrastructure"
[20:46:33 CEST] <ChocolateArmpits> The post is all about how they try to improve quality of these chunks
[20:47:55 CEST] <furq> At YouTube we care about the quality of the pixels we deliver to our users.
[20:48:00 CEST] <furq> handcrafted artisanal pixels
[20:49:05 CEST] <furq> i like how it says that's the only way they can keep up
[20:49:17 CEST] <furq> i mean they could make their encoder have less shit multithreading
[20:55:12 CEST] <SuperSQL> Hello. I have a problem I can't figure it out with ffmpeg. Recently I decided to digitize old video camera cassette tapes using PCI video card and ffmpeg to record. Problem that I have is that some tapes are abit damaged and it looks like some frames are being dropped (ffmpeg doesn't report that). Problem now is it looks like when I try to capture using ffmpeg those dropped frames get replaced by old frame, so when playing finished video
[20:55:12 CEST] <SuperSQL> you'd see single frames jumping back in time.
[20:55:50 CEST] <SuperSQL> I used this command ffmpeg  -channel 1 -s 920x576 -input_format yuyv422 -thread_queue_size 10000k -i /dev/video0 -f alsa -ac 1 -thread_queue_size 2000k -i hw:0 -vcodec copy -acodec copy "/home/user/Desktop/012.mkv"
[20:57:15 CEST] <ChocolateArmpits> SuperSQL: Are you sure it's ffmpeg duplicating frames and not the pci card ?
[20:57:17 CEST] <SuperSQL> I tried many things with it, it always ends up the same. There is nothing wrong with the video card since I tried using other video programs such as motv and it worked perfectly, it just dropped those frames
[20:57:32 CEST] <ChocolateArmpits> oh ok
[20:57:33 CEST] <obelisk29> Is there any way to encode h.265 files with the hvc1 atom rather then the default hevc? Premiere encodes using hvc1 and I tried just changing the tag -tag:v hvc1 but premiere won't import it.
[20:59:07 CEST] <SuperSQL> It's really strange, like if the frame buffer overflowed and went to the oldest one in memory
[20:59:33 CEST] <SuperSQL> no warnings or errors
[21:00:00 CEST] <wiak> ChocolateArmpits http://nwgat.ninja:8000/live.webm
[21:00:01 CEST] <wiak> ^^
[21:00:22 CEST] <ChocolateArmpits> wiak: huh ?
[21:01:03 CEST] <wiak> 1080 > 720, 36fps with speed 8, deadline realtime
[21:01:13 CEST] <wiak> 4mbps
[21:01:35 CEST] <ChocolateArmpits> wiak: What's the CPU usage ?
[21:02:16 CEST] <wiak> 20.9%
[21:02:22 CEST] <wiak> windows 10, FX-8350
[21:03:47 CEST] <nick0> Is this the right place to ask about libav issues?
[21:03:55 CEST] <wiak> #libav? hehe
[21:03:59 CEST] <JEEB> if you mean libav* libraries, yes
[21:04:06 CEST] <JEEB> Libav the fork project, no
[21:04:07 CEST] <nick0> err yeah, haha
[21:04:10 CEST] <nick0> the libraries
[21:04:20 CEST] <wiak> wait what  there is a libav fork?
[21:04:29 CEST] <furq> it was really nice of libav to pick such an unambiguous name
[21:04:42 CEST] <wiak> fufork?
[21:04:43 CEST] <wiak> :P
[21:04:56 CEST] <wiak> mey the forking commence
[21:05:59 CEST] <JEEB> wiak: more like very early 2011 and FFmpeg merges most things done there
[21:06:18 CEST] <wiak> https://gist.githubusercontent.com/nwgat/1cb7e8e69d0b1fbc6a57cb7631a807c5/raw/88050aea6f81dd720f1468e83d903d3c527d777d/gistfile1.txt
[21:06:46 CEST] <wiak> vp9 is a pain, its much nicer to use vp8, that one can maintain like 50fps
[21:07:24 CEST] <nick0> Okay, so basically I'm having an issue with the resampler: if I don't flush it I only get 5 seconds of my original input, but if I do, I get my entire input and an extra minute of it played backwards.
[21:07:35 CEST] <nick0> My code is here: http://pastebin.com/u7ThNPXq
[00:00:00 CEST] --- Tue Jul 12 2016