[Ffmpeg-devel-irc] ffmpeg.log.20150804

burek burek021 at gmail.com
Wed Aug 5 02:05:01 CEST 2015

[00:56:31 CEST] <Craytor> When I run this: ffmpeg -f x11grab -vc x264 -s xga -r 30 -b:v -g 300 -i :1.0 session-recording.avi, I get: :1.0: protocol not found
[00:56:42 CEST] <Craytor> Suggestions/help please?
[01:03:47 CEST] <edoceo> What's the best/preferred format container for huffyuv video stuffss? .avi or mpegts or mkv or ??
[01:03:54 CEST] <edoceo> I'm hoping to get one that is "fast"
[01:04:07 CEST] <Nolski> edoceo: I've used .avi which has been pretty fast
[01:04:20 CEST] <Craytor> Can someone help me please?
[01:04:24 CEST] <Nolski> .avi has also given me the least amount of trouble
[01:06:14 CEST] <Craytor> Nolski, http://pastebin.com/wgUreEqs
[01:07:02 CEST] <Nolski> Craytor: What are you intending to do with -i
[01:07:22 CEST] <Craytor> Nolski: Get the display...
[01:09:14 CEST] <Craytor> Nolski: is there something wrong with it?
[01:20:43 CEST] <Nolski> Craytor: Yeah, I don't think ffmpeg -i :1.0 grabs the display
[01:21:07 CEST] <c_14> It does.
[01:21:15 CEST] <c_14> It grabs the X display on :1.0
[01:23:49 CEST] <Nolski> Wouldn't it be :0.0+1?
[01:24:07 CEST] <c_14> The + is just offset stuffs
[01:24:14 CEST] <Nolski> ahh I see
[01:24:37 CEST] <Craytor> So is there any way to get this to work then?
[01:25:09 CEST] <c_14> The first X display is on :0.0 the second on :1.0
[01:26:16 CEST] <c_14> Oh, I see the problem
[01:26:25 CEST] <c_14> >-b:v -g 300
[01:26:35 CEST] <c_14> first of all, those are both output options, second that's broken
[01:27:51 CEST] <Craytor> So what do I do?
[01:27:56 CEST] <Craytor> Remove -b:v?
[01:28:07 CEST] <c_14> What are you trying to do?
[01:28:15 CEST] <Craytor> Record the screen
[01:28:32 CEST] <c_14> ffmpeg -f x11grab -i :1.0 out.avi
[01:29:50 CEST] <Craytor> hmm
[01:30:04 CEST] <Craytor> Okay, I need to open that file to see if it worked, then I'll tell you what's going on ;)
[01:30:13 CEST] <Craytor> But this is the farthest I've gotten yet
[01:36:30 CEST] <Craytor> c_14, the file is being created, but it's content is just black
[01:36:36 CEST] <Craytor> aka, no desktop display
[01:36:53 CEST] <c_14> What resolution is the desktop?
[01:37:50 CEST] <Craytor> I'm not sure, how do I tell?
[01:37:56 CEST] <Craytor> I'm using a VNC connection
[01:38:41 CEST] <Craytor> Got it, 1024x768
[01:39:25 CEST] <c_14> try adding -video_size 1024x768 before the -i
[01:40:38 CEST] <Craytor> Viewing window when I played it was larger
[01:40:40 CEST] <Craytor> But still black
[01:45:49 CEST] <DHE2> anything blocking shared memory? does adding -use_shm 0 help?
[01:45:58 CEST] <DHE2> (this will adversely affect performance, just suggesting it as a debug aid)
[01:48:36 CEST] <Craytor> DHE2, where would I go and put that?
[01:48:43 CEST] <DHE2> before -i like the other option
[01:49:19 CEST] <Craytor> It says it's an unrecognized option.
[01:51:05 CEST] <llogan> ...if you haven't yet
[01:52:31 CEST] <Craytor> http://pastebin.com/Y3qJYfUg
[01:53:51 CEST] <DHE2> must be using the xcbgrab variant which does't have that option... sorry
[01:55:10 CEST] <Craytor> I don't know why, but OBS won't work either
[01:55:32 CEST] <Craytor> Should you be able to use these on a VPS?
[01:59:26 CEST] <well0ne> well, now i have the same problem with concatenating, i have 5 streams with each 200 files which works flawless, withouth asynchronit
[01:59:58 CEST] <well0ne> but some other files, (they are all having the same settings), are getting async
[02:00:09 CEST] <DHE2> I'm using x11grab with Xvnc for over a year now. nigh-flawlessly. it does work
[02:34:37 CEST] <well0ne> i concenate 2 files, same x264/mp3 settings, but when it starts with the seconds file, the audio is async/has gone (depends on the player)
[02:34:44 CEST] <well0ne> any recommendations?
[02:35:14 CEST] <well0ne> i can try to make a sample
[02:40:17 CEST] <well0ne> http://mozz.top/sample.ts
[02:40:38 CEST] <well0ne> Here is the sample file, concenated the last seconds and the first 20 seconds, there you can see that it get async
[02:40:45 CEST] <well0ne> can someone tell me more about it ?
[02:45:24 CEST] <well0ne> http://mozz.top/out1.ts and http://mozz.top/out2.ts
[02:45:28 CEST] <well0ne> are the sources
[02:48:29 CEST] <well0ne> well i think i got it by myself
[02:49:02 CEST] <well0ne> the audio stream is 15 seconds longer than it should be
[02:49:14 CEST] <well0ne> is there a easy way to fix this?
[02:54:42 CEST] <Craytor> DHE2, are you still here?
[02:58:21 CEST] <DHE2> intermittently
[02:58:43 CEST] <Craytor> What version of Ubuntu are you running?
[02:58:56 CEST] <Craytor> What's your VM host as well (if you don't mind)?
[03:05:31 CEST] <DHE2> it's centos for me
[03:06:39 CEST] <Craytor> Hmm
[03:06:41 CEST] <Craytor> Okay
[03:07:05 CEST] <Craytor> I don't know what to do to get this working
[03:07:06 CEST] <DHE2> and it's bare metal, but no GUI installed
[03:07:11 CEST] <Craytor> And you said it works
[03:07:29 CEST] <Craytor> Then how can you use ffmpeg to record your screen?
[03:08:36 CEST] <DHE2> ffmpeg-x11grab -f x11grab -draw_mouse 0 -s 1280x720 -r 30 -i :1.0   ....
[03:08:39 CEST] <DHE2> that's my commandline
[03:09:25 CEST] <Craytor> What's it do?
[03:10:46 CEST] <DHE2> 30fps of the indicated size
[03:11:35 CEST] <Craytor> Yeah, are you recording the terminal or a desktop of something?
[03:12:12 CEST] <prelude2004c> hey guys.. i am having trouble getting somethign working with parallel... i am using ffmpeg to output files and then i am using inotifywait to poush those files via ssh
[03:12:24 CEST] <prelude2004c> the parallel processes have to happen but its not working well for me
[03:13:03 CEST] <prelude2004c> inotifywait -m -e CLOSE_WRITE  /var/www/channels/"${CHN}"| while read -r path action file ;do parallel rsync --timeout=20 -vz -e ssh "$path$file" "":/ramdrive/html/channels/${CHN}/;done
[03:13:17 CEST] <prelude2004c> anyone see what is wrong with it? i keep getting zombie processes when i do &done at the end
[03:14:16 CEST] <DHE2> use -n on ssh (see man page)
[03:17:08 CEST] <Craytor> DHE2 - was that to me?
[03:17:41 CEST] <DHE2> no
[03:18:20 CEST] <Craytor> ok
[03:18:29 CEST] <Craytor> What version of ffmpeg are you using?
[03:18:33 CEST] <Craytor> Did you compile it yourself?
[03:20:05 CEST] <DHE2> yes, I needed to build it custom becaues it runs on a headless system and I really like git versions
[03:20:32 CEST] <Craytor> Okay
[03:20:48 CEST] <Craytor> Could you maybe walk me through an install so I can maybe get tthis working/
[03:20:49 CEST] <prelude2004c> oh -n wasn't for me :)
[03:21:00 CEST] <DHE2> prelude2004c: it was. you're the one with ssh problems
[03:21:11 CEST] <prelude2004c> ssh problems ? no not ssh problems
[03:21:14 CEST] <prelude2004c> :)
[03:21:27 CEST] <prelude2004c> its a parallel problem actually of rsync of the segmented data
[03:22:04 CEST] <prelude2004c> i am sending it off to rsync and ssh but for each inotifywait it has to send a process.. but i dont want to wait for that process to finish before sending another
[03:22:19 CEST] <prelude2004c> i had it working but for some reason never closes the process so i get stuck with thousands of defunct messages
[03:22:24 CEST] <prelude2004c> i dont know how to fix it
[03:22:25 CEST] <DHE2> oh, okay I see...
[03:23:21 CEST] <prelude2004c> yup and i don't know enough to fix it myself
[03:23:49 CEST] <DHE2> the traditional way is called a double-fork. bit of a pain
[03:23:55 CEST] <llogan> doesn't seem like a ffmpeg issue to me
[03:35:57 CEST] <Craytor> DHE2 - should OBS work on the VPS or not?
[03:43:32 CEST] <DHE2> I don't see why not. what does being in a VM matter?
[03:45:58 CEST] <Craytor> There isn't a physical display?
[03:46:19 CEST] <Craytor> DHE2 - also, which one of these builds do you recommend? https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu
[03:57:32 CEST] <Craytor> I tried this again, and VLC is still playing a black stream
[03:57:37 CEST] <Craytor> No content, just black
[03:57:38 CEST] <Craytor> no idea
[04:18:37 CEST] <Craytor> So for some reason, all ffmpeg is doing is create a video that has no content, it's just a video that has black 24/6
[04:18:41 CEST] <Craytor> *24/7
[04:33:17 CEST] <Craytor> DHE2 - you still here?
[05:34:03 CEST] <Craytor> DHE2 - does your vps have a gpu?
[05:34:11 CEST] <Craytor> I can't seem to get mine running - still.
[08:10:33 CEST] <Guest12368> i want to attach png image mp4 video.
[08:10:46 CEST] <Guest12368> using ffmpeg.
[08:10:55 CEST] <Guest12368> can anyone help me on that ?
[08:11:21 CEST] <Guest12368> i want to attach this png logo at end of video.
[08:17:38 CEST] <Boscop> hi, is there a tutorial for processing each frame in a video with a custom filter/effect?
[08:53:54 CEST] <Guest12368> Anyone there ?
[09:00:08 CEST] <Boscop> Guest12368: yes
[09:00:24 CEST] <Boscop> how can i use a custom avfilter with ffmpeg?
[09:03:37 CEST] <Guest73378> we want to overlay png image on mp4 video after specific time. we also want to fade in and fade out that png image.
[09:03:47 CEST] <Guest73378> does anyone help us on this ?
[09:03:57 CEST] <Guest73378>  cas anyone help us on this ?
[09:04:02 CEST] <Guest73378>  can anyone help us on this ?
[09:13:10 CEST] <durandal_1707> Boscop: what custom filter?
[09:13:23 CEST] <Boscop> durandal_1707: that i wrote
[09:13:31 CEST] <Boscop> to process each video frame
[09:14:41 CEST] <Boscop> durandal_1707: using avfilter
[09:14:41 CEST] <durandal_1707> did you compiled it?
[09:15:03 CEST] <Boscop> yes, the question is, how can i "register it" so i can use it with ffmpeg ?
[09:15:51 CEST] <durandal_1707> you need to edit allfilters and Makefile in libavfilter/
[09:16:39 CEST] <Boscop> durandal_1707: do i have to compile ffmpeg manually?
[09:16:51 CEST] <durandal_1707> yes
[09:18:29 CEST] <Guest73378> overlay feature won't help us on this ?
[09:19:46 CEST] <durandal_1707> Guest73378: it is possible, read documentation
[09:20:00 CEST] <Boscop> Guest73378: https://ffmpeg.org/ffmpeg-filters.html#Description
[09:20:09 CEST] <Boscop> the time constraints
[09:20:55 CEST] <Boscop> durandal_1707: so every time i add another filter i have to recompile ffmpeg?
[09:21:15 CEST] <Boscop> isn't it supposed to be a plugin system?
[09:22:41 CEST] <durandal_1707> exactly, open bug report, there is patch somewhere which addresses this
[09:23:24 CEST] <Boscop> ah
[09:25:01 CEST] <JEEB> Boscop: you have to recompile the parts that are required to be recompiled (the makefiles keep an eye on what stuff has changed and in most cases it works)
[09:25:30 CEST] <Boscop> JEEB: on windows should i use mingw or msvc?
[09:25:31 CEST] <JEEB> but yes, it requires linking a new avfilter and ffmpeg :P
[09:25:53 CEST] <JEEB> Boscop: msys2 + mingw-w64 is probably simplest, if you have MSVC 2013 or 2015 you can use those, too
[09:27:20 CEST] <Boscop> JEEB: can i process two video feeds from two usb webcams in real-time using ffmpeg and then stream them?
[09:27:50 CEST] <JEEB> depends on a lot of stuff, but I've seen restreaming webcams with ffmpeg :P
[09:28:11 CEST] <JEEB> although most of that is on lunix, I have no idea of how it'd be accessed on windows
[09:28:23 CEST] <Boscop> JEEB: so i can open multiple usb webcam streams at the same time with ffmpeg?
[09:28:33 CEST] <JEEB> there seems to be a dshow input but I have no idea how good it is
[09:28:51 CEST] <JEEB> Boscop: you are talking as if there's a single way of doing it :P
[09:29:09 CEST] <JEEB> depends on your OS, how stupid your webcams are etc
[09:29:51 CEST] <JEEB> I've seen a person on linux using v4l2 grabbing AVC streams from specific USB web cams and restreaming that
[09:30:48 CEST] <JEEB> since you mention MSVC it implies that you're on windows. thus, you will have to find out what your options are for grabbing the pictures from the web cam
[09:31:06 CEST] <JEEB> as I said, there is a DirectShow input at least, and I remember that some webcams might be DShow inputs
[09:31:15 CEST] <JEEB> but as I also said, I have no idea how good that stuff is
[09:32:18 CEST] <Boscop> JEEB: i can switch to linux for this. do you have a link to a post of the person who did it on linux?
[09:34:12 CEST] <JEEB> it's somewhere in my backlog but I'm lazy to grep my #libav logs :P it should be pretty straightforward if you look at how the v4l2 inputs are used in the ffmpeg documentation
[09:34:21 CEST] <JEEB> ffmpeg-all.html in google and then look there :P
[09:45:05 CEST] <Boscop> JEEB: thanks. btw, do you know what the easiest way is to align two stereoscopic videos, is there a library for it?
[09:45:37 CEST] <Boscop> for timing alignment using audio correlation and vertical alignment etc
[09:47:45 CEST] <JEEB> why the hell would you use audio correllation for timing? one would think you'd use timestamps from the camera (implying they are not stupid enough to not have timestamps on their output picture streams)
[09:48:06 CEST] <JEEB> and even in that case you could theoretically use the wallclock as the main clock
[09:48:52 CEST] <JEEB> for the alignment I have no idea, but if you just want to put two things next to each other libavfilter should have something for that
[09:56:37 CEST] <Boscop> JEEB: i'm traveling so i can't use large equipment, i was thinking of recording with two phones next to each other
[09:57:20 CEST] <Boscop> which wallclock?
[15:47:03 CEST] <mathieu_> Hi, i've got a question about motion_val
[15:47:19 CEST] <mathieu_> It's quite stupid :p
[15:47:41 CEST] <mathieu_> I want to copy the table, and im a bit confused about the syntax
[15:47:57 CEST] <mathieu_> I have this:
[15:48:20 CEST] <mathieu_> copy_motion_val = malloc(2 * mb_width * mb_height * 2 * sizeof(int16_t));
[15:49:18 CEST] <mathieu_> is this correct? and what casting do I need for the malloc?
[15:49:37 CEST] <mathieu_> it's a really strange structure syntax :p
[16:19:49 CEST] <eik> I'm having issues with the trim filter not dropping frames outside the specified range, it rather duplicates them for some reason. Anyone got any ideas on what I'm doing wrong? Command used http://pastebin.com/8pAQZvSY
[16:23:51 CEST] <ubitux> from vfr to cfr maybe
[16:25:23 CEST] <ubitux> eik: you probably want to -ss and -t
[16:27:58 CEST] <eik> ubitux: I was hoping I could get this filter to work as intended. It would be nice to use the start_frame parameter
[16:29:43 CEST] <ubitux> eik: trim filters are automatically inserted with ffmpeg
[16:29:51 CEST] <ubitux> i wonder if you can see them with -v verbose or something
[16:30:01 CEST] <ubitux> (try -v verbose with -ss / -t just in case)
[16:30:44 CEST] <eik> I can't use -ss in this case I think, since it's positional (comes before or after -i / -o changes behaviour)
[16:31:26 CEST] <eik> Due to how I send arguments to ffmpeg from an external app, this is not an option :/
[16:32:36 CEST] <ubitux> no i mean, just to see how ffmpeg does it
[16:32:48 CEST] <ubitux> so you can copy the behaviour
[16:33:02 CEST] <ubitux> it looks weird though
[16:35:26 CEST] <eik> I tried with -v verbose on the trim filter first. It says "*** 900 dup!" then skips to frame 902 at ~30s in. I'll try -ss with verbose now
[16:41:47 CEST] <eik> -ss works as expected, it skips 30s and marks that as the new 00:00:00. While with trim it just duplicates the first 30s and then goes to frame 900
[16:52:05 CEST] <eik> ubitux: Here are the two outputs from trim and -ss. http://pastebin.com/ZPeCgdYq http://pastebin.com/mMeiZQX8 So trim doesn't drop until the start like the docs say, it rather duplicates.
[16:53:28 CEST] <ubitux> you might want to add a setpts after the trim
[16:54:09 CEST] <ubitux> because triming is not going to adjust the time gap between trim chunks (assuming you have multiple ones, but same if you pick a time in the middle like this)
[16:54:39 CEST] <ubitux> so you need to adjust the times of the frame
[16:55:27 CEST] <ubitux> i would guess ffmpeg assumes the frame doesn't change (enhance assume vfr) but is trying to output cfr (so it will dup the last frame N times)
[16:55:46 CEST] <ubitux> i suppose that's the kind of scenario that's happening
[16:56:30 CEST] <ubitux> in your case, maybe a setpts=PTS-STARTPTS would work
[16:56:45 CEST] <ubitux> (since you only have one trim and you want to adjust the gap at start time
[16:57:20 CEST] <ubitux> so trim=30:45,setpts=PTS-STARTPTS ... maybe :p
[17:09:35 CEST] <reza> hi , can i stream live over UDP usnig ffserver?
[18:22:22 CEST] <dv__> slightly OT: why do webradio links often offer .pls and .m3u playlists with the actual webradio HTTP link inside, and not just the HTTP link directly? if it is about possible future stream location changes: these could also be taken care of by using a HTTP 302/303 return code
[18:27:38 CEST] <ChocolateArmpits> dv__: I guess to support adaptive playback
[18:27:46 CEST] <ChocolateArmpits> oh webradio
[18:27:50 CEST] <ChocolateArmpits> beats me
[19:18:29 CEST] <klaxa> dv__: probably so people can store the playlist on their desktop and double-click it
[20:01:25 CEST] <kyleogrg> hi
[20:01:54 CEST] <kyleogrg> What's the most painless way to build FFmpeg with libfdk_aac?
[20:04:08 CEST] <sfan5> use some kind of build script and modify it to enable fdk-aac (e.g. pkgfile from AUR if you're using arch)
[20:13:37 CEST] <kyleogrg> what about windows
[20:14:34 CEST] <sfan5> then you're probably out of luck
[20:15:15 CEST] <llogan> maybe the zeranoe site has a build script you can modify
[20:15:37 CEST] <sfan5> aren't those cross-compiled from linux?
[20:15:41 CEST] <llogan> i don't know
[20:15:43 CEST] <kyleogrg> Oh?
[20:15:45 CEST] <kyleogrg> hmm
[20:16:12 CEST] <ChocolateArmpits> kyleogrg: use this guide: http://taer-naguur.blogspot.com/2013/09/compiling-x64x86-ffmpeg-with-fdk-aac-lib-on-64bit-windows-how-to.html
[20:16:46 CEST] <kyleogrg> ChocolateArmpits: I was about to post something from the same author: http://taer-naguur.blogspot.com/2013/10/ffmpeg-autobuild-tool-x64.html
[20:16:46 CEST] <ChocolateArmpits> In the x264 compile options remove --bit-depth=10 or change it to--bit-depth=8
[20:17:23 CEST] <kyleogrg> I actually successfully used that script to build ffmpeg with libfdk_aac... but
[20:17:40 CEST] <kyleogrg> I want to build libx265 not libx264
[20:18:14 CEST] <kyleogrg> As a relative novice, I did not know how to adjust the script to do this.  Any suggestions?
[20:18:31 CEST] <ChocolateArmpits> I'm not familiar with the script
[20:19:03 CEST] <ChocolateArmpits> In the step-by-step guide however you will probably need to add --enable-libx265 in ffmpeg's compile options
[20:19:13 CEST] <ChocolateArmpits> and precompile x265
[20:19:23 CEST] <ChocolateArmpits> I don't know what options that might have
[20:19:38 CEST] <ChocolateArmpits> But the process should be the same
[20:19:44 CEST] <kyleogrg> hmm ok
[20:19:46 CEST] <sfan5> not really
[20:19:52 CEST] <sfan5> x265 uses CMake
[20:20:01 CEST] <sfan5> x264 uses a "normal" configure script IIRC
[20:20:27 CEST] <kyleogrg> oh
[20:20:28 CEST] <ChocolateArmpits> Wouldn't that just need additional binaries ?
[20:21:06 CEST] <sfan5> probably
[20:21:51 CEST] <kyleogrg> let's say i want to build a "standard" zeranoe ffmpeg.  is that easy?  then can i do a simple mod to add libfdk_aac?
[20:22:16 CEST] <sfan5> it's not that simple
[20:22:37 CEST] <BtbN> The zeranoe builds are everything but standard. There's quite a bunch of stuff enabled.
[20:22:42 CEST] <sfan5> you'd need to add a "section" to the script that downloads/compiled and install fdk-aac
[20:23:28 CEST] <kyleogrg> ok.
[20:23:40 CEST] <kyleogrg> i wish it was simpler
[20:24:40 CEST] <kyleogrg> i can download zeranoe ffmpeg for libx265.  i can use the above-mentioned script for libfdk_aac.  but i can't easily have both...
[20:29:25 CEST] <kyleogrg> maybe i should try vp9 and opus
[20:30:22 CEST] <sfan5> libvpx is slow
[20:30:26 CEST] <sfan5> libvpx-vpp9 even slower
[20:31:28 CEST] <kyleogrg> yeah.. but in terms of an alternative to x265, it's very competetive, i believe
[20:35:29 CEST] <JEEB> lol
[20:35:37 CEST] <JEEB> did you drink some google kool-aid?
[20:35:45 CEST] <JEEB> if so, I'd like a couple of truckloads of it as well
[20:41:16 CEST] <kyleogrg> haha
[20:41:46 CEST] <kyleogrg> is x265 more efficient?
[20:42:33 CEST] <JEEB> x265 isn't too great but so far it's got better rate control, psychovisual opts, more asm and better threading
[20:42:48 CEST] <JEEB> so currently out of the current-gen format encoders it's probably the least retarded
[20:43:22 CEST] <kyleogrg> one thing is "-tune grain" didn't work last time i checked
[20:43:48 CEST] <JEEB> you shouldn't expect libx264 and libx265 be in any kind similar, since they pretty much aren't
[20:44:02 CEST] <JEEB> there are some similarly named features and things that kind of do similar things
[20:44:06 CEST] <kyleogrg> yeah, but i think this is an option in x265
[20:44:43 CEST] <ChocolateArmpits> JEEB: wouldn't it be smarter to add a hack multithreading for vp9 that would render several segments parallellely and then concatenate them ?
[20:44:43 CEST] <JEEB> yeah, it is. I think with libx265 the decision was made that you're supposed to use an option that takes in API options instead of having o9k global options
[20:45:20 CEST] <sfan5> -x265_params iirc
[20:45:23 CEST] <sfan5> or maybe with a -
[20:45:29 CEST] <sfan5> or param without s
[20:45:32 CEST] <JEEB> because it gets at times a bit lulz when people try to use various libx264-specific options with other encoders
[20:45:33 CEST] <sfan5> i can never remember that
[20:45:49 CEST] <JEEB> ChocolateArmpits: you could do the same with encoders too and gain with it
[20:45:57 CEST] <JEEB> if you aren't doing live encoding that is
[20:46:06 CEST] <kyleogrg> do you think libopus is more efficient/better than libfdk_aac?
[20:46:37 CEST] <sfan5> turns out it's -x265-params
[20:46:46 CEST] <ChocolateArmpits> JEEB: well yeah, but if Google isn't ready to add proper multithreading until some time later, this can be a shortterm solution for offline video
[20:46:46 CEST] <JEEB> I am all for opus vs aac
[20:47:01 CEST] <JEEB> if only your clients support opus, go for it
[20:47:15 CEST] <JEEB> you can distribute binaries with libopus so it even makes that thing a non-problem
[20:47:49 CEST] <kyleogrg> well, i wonder if i should try libx265+libopus for maximum efficiency
[20:50:06 CEST] <ChocolateArmpits> kyleogrg: try for lowest bitrate as a challenge
[20:50:14 CEST] <ChocolateArmpits> as in sub-500kbps
[20:51:02 CEST] <kyleogrg> yeah
[20:53:04 CEST] <kyleogrg> i wanted to try an experiment.  i have a 60-minute video which i want to convert to a 60 MB (!) video
[20:53:50 CEST] <ChocolateArmpits> what genre ?
[20:53:53 CEST] <kyleogrg> let's say it's basically a dvd-quality source
[20:53:56 CEST] Action: JEEB is reminded of the floppy sized animation episodes
[20:54:07 CEST] <JEEB> 24min in 1.44MB
[20:54:15 CEST] <kyleogrg> wow
[20:54:27 CEST] <ChocolateArmpits> tell us more
[20:54:39 CEST] <ChocolateArmpits> was that text animation or something ?
[20:54:55 CEST] <ChocolateArmpits> or sprites ??
[20:55:16 CEST] <JEEB> animoo basically
[20:55:34 CEST] <JEEB> not action but still somewhat animated. + x264 is very good at keeping your rate
[20:55:34 CEST] <ChocolateArmpits> so a slideshow ?
[20:55:50 CEST] <kyleogrg> basically i want to convert a video archive into a *proxy* video archive that will fit on a thumb drive.  so i want the maximum quality given the limited file space.
[20:56:33 CEST] <BtbN> Isn't x264 currently still better for low resolutions/bitrates?
[20:56:34 CEST] <sfan5> 2-pass encoding is what you want
[20:56:41 CEST] <kyleogrg> Let
[20:56:43 CEST] <ChocolateArmpits> kyleogrg: 136kbps average overall allowed at 60min/60MB
[20:57:01 CEST] <kyleogrg> Let's say I want to fit 1,000 hours of DVD video on 60GB.
[20:57:11 CEST] <kyleogrg> Which I know is very small.
[20:57:14 CEST] <ChocolateArmpits> BtbN: last I saw x265 was better, if that tears of steel video isn't borked
[20:57:40 CEST] <ChocolateArmpits> from my personal experience at 400kbps only SD conference videos hold up
[20:57:50 CEST] <ChocolateArmpits> with x264
[20:58:03 CEST] <kyleogrg> That's 61.44MB per minute.  So...?
[20:58:47 CEST] <kyleogrg> So x264 is better than x265 at low bitrates?
[20:59:24 CEST] <ChocolateArmpits> well, I wouldn't argue codecs specifically but standards
[20:59:38 CEST] <ChocolateArmpits> so h265 is better than h264 at lower bitrates
[20:59:53 CEST] <BtbN> Of course h265 is in theory better. But there currently is no implementation that's able to make use of it.
[21:00:41 CEST] <kyleogrg> BtbN: I'm not sure what you mean... If I use libx265 on placebo, won't that be better than libx264?
[21:01:31 CEST] <BtbN> Depends, for 4K stuff, yes, very likely
[21:01:39 CEST] <BtbN> for targeting low res/bitrate, i don't think so
[21:01:47 CEST] <ChocolateArmpits> check these out
[21:01:48 CEST] <ChocolateArmpits> https://x265.com/hevc-video-files/
[21:01:54 CEST] <ChocolateArmpits> supposed comparisons
[21:02:25 CEST] <ChocolateArmpits> I wouldn't go by images alone
[21:02:57 CEST] <Daan_> Hi, Could you guys release compiled versions for Unix in the future?
[21:03:20 CEST] <kyleogrg> before i dig into comparisons online, what's your opinion offhand?  that libx264 is better right now for low bitrate?
[21:03:47 CEST] <llogan> Daan_: http://johnvansickle.com/ffmpeg/
[21:03:53 CEST] <BtbN> Daan_, your distribution most likely packages ffmpeg or at least libav for you.
[21:04:00 CEST] <BtbN> For more than that, compile it yourself.
[21:04:22 CEST] <ChocolateArmpits> kyleogrg: I haven't tried x265 but at those bitrates x265 beats x264
[21:04:40 CEST] <kyleogrg> you mean high bitrates?
[21:04:52 CEST] <ChocolateArmpits> low bitrates
[21:04:59 CEST] <ChocolateArmpits> literally sub-500kbps
[21:05:01 CEST] <kyleogrg> ok
[21:05:26 CEST] <ChocolateArmpits> anything higher and I'm getting my doubts if it's worth it
[21:05:45 CEST] <kyleogrg> given my goal (tiny video proxies), I wonder if I should reduce the frame rate to put more quality into each frame
[21:06:43 CEST] <ChocolateArmpits> kyleogrg: well it will speed up encoding
[21:06:56 CEST] <kyleogrg> i could go from 30p to 15p, for instance.  i wonder if that will still be watchable
[21:07:42 CEST] <ChocolateArmpits> From my experience the smaller the viewing window the less noticeable low fps rate is
[21:08:05 CEST] <kyleogrg> watchable enough to say "hey, that's a good clip.  i want to get the original and use it in my project."
[21:08:05 CEST] <ChocolateArmpits> especially for static scenes
[21:08:26 CEST] <kyleogrg> yeah
[21:08:48 CEST] <kyleogrg> this is all sorts of scenes - some static, some random
[21:12:21 CEST] <kyleogrg> what about 30p (30 frames) vs. 30i (15 frames, 30 fields)?
[21:13:07 CEST] <kyleogrg> instead of having 15 progressive frames, what if i had 30 fields?
[21:13:27 CEST] <kyleogrg> wouldn't the motion be smoother?
[21:14:00 CEST] <sfan5> ugh interlacing
[21:15:46 CEST] <kyleogrg> haha
[21:16:02 CEST] <kyleogrg> but i have a unique situation
[21:16:56 CEST] <ChocolateArmpits> I would suggest against non-standard practises
[21:17:13 CEST] <ChocolateArmpits> technically it would better to call it 15i
[21:17:38 CEST] <ChocolateArmpits> besides interlacing flags need additional bitrate
[21:18:01 CEST] <ChocolateArmpits> I don't know if x265 codes the field separately or no
[21:18:21 CEST] <kyleogrg> uh-huh.  would a 15p video be much smaller than a 15i?
[21:19:22 CEST] <ChocolateArmpits> can't say by how much
[21:21:19 CEST] <niknok> does ffmpeg accepts a subtitle file in .pac format?
[21:53:09 CEST] <Rayne> Hello *, I have a problem with the filter "overlay" with a looped image and a video file. Both ffmpeg (Windows) and avconv (Linux) don't stop processing after the "shortest video" is processed.
[21:53:19 CEST] <Rayne> Source: http://nopaste.linux-dev.org/?665597
[21:53:56 CEST] <Rayne> Example 2 returns a "Undefined constant or missing '(' in 'shortest=1'" error
[21:55:23 CEST] <Rayne> Do you have an idea?
[21:57:47 CEST] <pzich> I think you'll have better luck in an avconv-specific channel
[22:04:40 CEST] <llogan> or show the real ffmpeg command and complete console output.
[22:04:55 CEST] <llogan> otherwise you'll have to go to the avconv echo-chamber
[22:06:55 CEST] <kyleogrg> ffmpeg won't let me use "-ar 22050" for libopus
[22:07:03 CEST] <kyleogrg> it complains
[22:08:13 CEST] <JEEB> opus is a wee bit special regarding that kind of stuff. so just don't set an audio rate
[22:08:22 CEST] <kyleogrg> "Error while opening encoder for output stream #0:1 - maybe incorrect parameters such as bit_rate, rate, width or height"
[22:08:25 CEST] <kyleogrg> oh?
[22:09:10 CEST] <llogan> that's not a supported sample rate, but as JEEB said...
[22:09:28 CEST] <kyleogrg> then if i want to reduce the audio file size, what can i do
[22:09:35 CEST] <JEEB> just set the audio bit rate
[22:09:38 CEST] <kyleogrg> just set the bitrate, i guess
[22:09:39 CEST] <JEEB> or quantizer
[22:09:52 CEST] <kyleogrg> is quantizer the quality setting?
[22:10:16 CEST] <JEEB> do not mistake quantizer for quality :P
[22:10:35 CEST] <kyleogrg> does "-vbr on" work?
[22:11:10 CEST] <llogan> there is a -vbr private option. see "ffmpeg -h encoder=libopus"
[22:13:00 CEST] <kyleogrg> hmm.  so is -vbr 0 to -vbr 2 a quality setting?
[22:14:32 CEST] <kyleogrg> oh.  0 = off.  1 = on.  2 = constrained.
[22:14:54 CEST] <Rayne> llogan, PEBKAC. Thank you. No problem with ffmpeg. :-)
[22:42:03 CEST] <kyleogrg> for 2-pass encoding, i'm supposed to do "-f mp4 NUL && \" (windows).  however, i am encoding an mkv file.  should i replace "mp4" with "mkv"?
[22:43:38 CEST] <JEEB> format takes in formats, see a list of them with `ffmpeg -formats`
[22:44:37 CEST] <kyleogrg> okay.  so i guess i will do "matroska"
[22:47:47 CEST] <kyleogrg> ffmpeg automatically optimizes threads right?  that is, it will use all cores by default?
[23:41:50 CEST] <llogan> kyleogrg: depends on the encoder (I'm assuming you're referring to encoding)
[23:42:16 CEST] <kyleogrg> yes.  libx265
[23:45:40 CEST] <llogan> i would guess it does
[23:45:58 CEST] <kyleogrg> ok
[23:53:05 CEST] <kyleogrg> when encoding, how can i scale video res to 33% of original size?
[23:53:56 CEST] <llogan> -vf scale=iw*0.3333333333333333333333333333333333333333
[23:54:20 CEST] <llogan> eh, well, you get the idea
[23:54:32 CEST] <kyleogrg> ok, actually it's a bit more complicated.  how can i first convert to square pixels, and then get 33% of the size....
[23:55:33 CEST] <kyleogrg> the source is a non-square-pixels, 720x480 vid.  so it needs to become 640x480, and then 33% of that.
[23:55:50 CEST] <llogan> if you're lazy use two scales
[23:55:58 CEST] <kyleogrg> i know that this will convert to square pixels: scale=iw*sar:ih
[23:56:21 CEST] <kyleogrg> lazy is fine if it works...
[23:56:26 CEST] <Nolski> Cant you just create a filter graph?
[23:56:35 CEST] <kyleogrg> idk what that is
[23:57:20 CEST] <Nolski> -filter_complex "scale=iw*sar:ih; scale=iw*0.3333333"
[23:57:33 CEST] <kyleogrg> llogan: would this work?  -vf "scale=iw*sar:ih,scale=iw*0.3333...."
[23:58:07 CEST] <llogan> iw*<your desired value>:-1
[23:58:10 CEST] <kyleogrg> Nolski: okay.
[23:58:23 CEST] <llogan> or iw*<whatever>:ih*<whatever>
[23:58:54 CEST] <Nolski> ^ right what he said
[23:59:57 CEST] <kyleogrg> bit confused.  source, which is 720x480, should become 640x480, which should become 211x158 (33%).
[00:00:00 CEST] --- Wed Aug  5 2015

More information about the Ffmpeg-devel-irc mailing list