[Ffmpeg-devel-irc] ffmpeg.log.20120904

[burek]

[00:43] <rud> hmm anyone would care to comment the following (RTMP live-encoding) lines..: http://pastebin.com/dCPy5XT7 looking for ways to optimize the transcoding, and mostly, cpu use / efficiency .. but i'm unsure where / what to do next.
[00:44] <rud> quality is far more than ok, the real deal is cpu optimization (i think)
[00:54] <cbsrobot-> rud: try -preset
[00:54] <cbsrobot-> with one of the following: ultrafast, superfast, veryfast, faster, fast
[00:55] <rud> cbsrobot-: thanks, trying right now
[00:55] <cbsrobot-> f.ex:
[00:55] <cbsrobot-> ffmpeg -loglevel fatal -threads 8  -i rtmp://source.ext/app/stream \
[00:55] <cbsrobot-> -re -acodec libfaac -ar 44100 -b:a 96k -vcodec libx264 -tune ultrafast -s 1280x720 -b:v 896k -r 25 -metadata streamName=demo_720p -f flv rtmp://destination.ext/r/demo_720p
[00:55] <cbsrobot-> ah nope
[00:55] <cbsrobot-> ffmpeg -loglevel fatal -threads 8  -i rtmp://source.ext/app/stream \
[00:55] <cbsrobot-> -re -acodec libfaac -ar 44100 -b:a 96k -vcodec libx264 -preset ultrafast -s 1280x720 -b:v 896k -r 25 -metadata streamName=demo_720p -f flv rtmp://destination.ext/r/demo_720p
[00:56] <rud> gotcha
[00:56] <cbsrobot-> and
[00:56] <cbsrobot-> the threads in front of -i means threads used to decode
[00:56] <cbsrobot-> are you sure about that ?
[00:57] <cbsrobot-> try:
[00:57] <rud> oh
[00:57] <rud> :)
[00:58] <cbsrobot-> ffmpeg -i rtmp://source.ext/app/stream -re -c:a libfaac -ar 44100 -b:a 96k -c:v libx264 -preset ultrafast -s 1280x720 -b:v 896k -r 25 -metadata streamName=demo_720p -f flv rtmp://destination.ext/r/demo_720p
[00:59] <rud> so, no -threads after -re ?
[00:59] <cbsrobot-> nah
[00:59] <rud> k
[00:59] <cbsrobot-> hehe
[01:00] <cbsrobot-> maybe even:
[01:00] <rud> :)
[01:00] <cbsrobot-> ffmpeg -i rtmp://source.ext/app/stream -re -c:a libfaac -ar 44100 -b:a 96k -c:v libx264 -preset ultrafast -crf 23 -s 1280x720 -r 25 -metadata streamName=demo_720p -f flv rtmp://destination.ext/r/demo_720p
[01:00] <rud> sec, i'll paste you my whole env
[01:01] <cbsrobot-> and play with the cry value: lower = better quality but more bitrate, higher = the inverse
[01:01] <cbsrobot-> s/cry/crf/
[01:02] <cbsrobot-> and finally:
[01:03] <cbsrobot-> ffmpeg -i rtmp://source.ext/app/stream -re -c:a libfaac -ar 44100 -b:a 96k -c:v libx264 -preset ultrafast -tune zerolatency -crf 23 -s 1280x720 -r 25 -metadata streamName=demo_720p -f flv rtmp://destination.ext/r/demo_720p
[01:04] <rud> aha good to know, here are my tests pre -crf http://pastebin.com/xDNEwsYE
[01:05] <rud> no sec, sorry
[01:05] <rud> here : http://pastebin.com/xDNEwsYE
[01:05] <rud> arg no, hrm
[01:06] <rud> that's the hls output, sorry, need sleep, pasting proper logs
[01:12] <cbsrobot-> input is 854x480 and you scale it to 720p ...
[01:14] <rud> i know, i have a poor bandwidth here on encoder side, so can't push real 720p, but production site has a full 4meg SDSL ..
[01:14] <rud> here are the full session logs: http://pastebin.com/FBHnjJjj
[01:18] <cbsrobot-> is the cpu load better now ?
[01:18] <rud> hell yes
[01:18] <rud> :)
[01:18] <rud> displayed bitrate of ffmpeg jumped from 700+kbs to 1700+kbs, so yes, great improvement :)
[01:19] <rud> testing now with all transcoders active ..
[01:19] <cbsrobot-> adapt the crf settings for each stream
[01:19] <cbsrobot-> and bandwidth wise you could try https://github.com/mstorsjo/fdk-aac instead of libfaac
[01:20] <rud> nice
[01:20] <cbsrobot-> and then set -b:a 32k
[01:20] <rud> ohh nice !
[01:20] <cbsrobot-> but not sure if cpu likes it
[01:20] <rud> will test, many many thanks for your advices :)
[01:21] <cbsrobot-> or try libaacplus from http://tipok.org.ua/node/17
[01:21] <cbsrobot-> should work out of the box
[01:22] <rud> wuah, load dropped from 14.0 to a humble 1.70 :)
[01:22] <cbsrobot-> nice
[01:22] <rud> (i'm running the commands i pasted + more less another 4 transcoders doing HLS)
[01:24] <cbsrobot-> need to sleep now
[01:24] <rud> indeed :)
[01:24] <cbsrobot-> gn
[01:24] <rud> you too, bye !
[01:27] <rud> managed to drop encoder latency of broadcaster > preview in iOS from 45s to 20s not bad!
[01:29] <physean> hi
[01:29] <physean> i'm trying to make a watermark
[01:30] <physean> i got it to work on windows
[01:31] <physean> but in my VPS, it's throwing errors
[01:32] <physean> i don't have -vf on this version of ffmpeg, it says
[01:32] <physean> ffmpeg: unrecognized option '-vf'
[01:32] <sacarasc> Update.
[01:32] <physean> i tried to apt-get it and it says it's at latest version
[01:33] <sacarasc> Then you'll probably have to compile yourself.
[01:33] <physean> gotcha
[01:33] <physean> i downloaded the sources
[01:33] <sacarasc> Which version is it, out of curiosity?
[01:33] <physean> FFmpeg version SVN-r0.5.9-4:0.5.9-1, Copyright (c) 2000-2009 Fabrice Bellard, et al.
[01:33] <physean> dreamhost VPS, they prolly use an old version for compatibility
[01:33] <physean> i would use vhook but the module is missing
[01:33] <sacarasc> Ubuntu?
[01:34] <physean> i think so
[01:34] <sacarasc> If so, http://goo.gl/9l5Nl
[01:34] <sacarasc> The Ubuntu compile guide!
[01:34] <physean> thanks.
[01:35] <physean> http://wiki.dreamhost.com/FFmpeg
[01:35] <sacarasc> Download the latest version of FFmpeg into a new temporary directory called ffmpeg by typing:
[01:35] <sacarasc>     svn checkout svn://svn.mplayerhq.hu/ffmpeg/trunk ffmpeg
[01:35] <sacarasc> That's like... 20+ months out of date.
[01:36] <cbsrobot-> rud: check rtmp_buffer in the docs
[01:36] <sacarasc> ffmpeg haven't used svn since January 11.
[01:36] <sacarasc> *January 2011.
[01:36] <cbsrobot-> maybe you can lower the vlue to get faster output
[01:36] <rud> cbsrobot-: hmm good call
[01:36] <physean> sacarasc, bear with me for a minute
[01:36] <rud> i have an audio issue right now, chopped audio in generated streams, will investigate, but at least, it's a brilliant start :)
[01:37] <physean> sacarasc, will it mess up my current ffmpeg install? or what?
[01:38] <sacarasc> You could uninstall your current one and follow the guide and all should be good.
[01:38] <physean> in the dreamhost wiki it says the one currently installed is in a "shared directory"
[01:39] <physean>  the absolute path to the shared installation of FFmpeg is:
[01:39] <physean> /usr/bin/ffmpeg
[01:39] <ubitux> you can use a non-installed ffmpeg version
[01:39] <physean> alright.
[01:39] <ubitux> so ./configure, make, and ./ffmpeg ...
[01:39] <physean> i've entered the svn command.
[01:39] <ubitux> we don't use svn since ages
[01:39] <ubitux> don't use it.
[01:39] <physean> gotcha
[01:39] <ubitux> https://ffmpeg.org/download.html
[01:40] <ubitux> you also can use:
[01:40] <physean> i have downloaded and extracted ffmpeg-0.11.1
[01:40] <physean> sounds right?
[01:40] <ubitux> you should use git head
[01:40] <ubitux> 0.11 is kind of old now
[01:43] <physean> http://ffmpeg.gusari.org/static/64bit/ffmpeg.static.64bit.2012-09-03.tar.gz
[01:43] <physean> ?
[01:43] <ubitux> this is a static build ready-to-use
[01:43] <physean> nice
[01:43] <rud> cbsrobot-: really good call, using 1000ms, i get an RTMP > RTMP latency of about 2secs
[01:44] <ubitux> for building it yourself you should read the top of the ffmpeg.org/download.html pasted a few minutes ago
[01:44] <rud> i mean broadcaster > rtmpd > ffmpeg > rtmpd > rtmp-client
[01:44] <physean> ubitux, if it works i'm happy
[01:44] <physean> it seems to launch
[01:44] <cbsrobot-> rud: cool
[01:44] <rud> need to figure why HLS audio is choppy now, but, dog-walk first.
[01:45] <cbsrobot-> try async
[01:45] <rud> asyncts, gotcha
[01:46] <physean> ubitux, can i even create a transparent watermark?
[01:47] <physean> so far i was able to create an opaque one on a windows build
[01:52] <physean> aha
[01:52] <physean> i've successfully ran the command
[01:52] <physean> now to upload the video somewhere
[01:55] <physean> ubitux,
[01:55] <physean> its done.
[01:55] <physean> and transparent.
[01:55] <physean> thanks bro
[02:05] <physean> how does flv keep the codec the same
[02:06] <physean> the input video was h264, the output was mpeg4 something
[02:06] <physean> obviously i've messed something up
[02:06] <physean> i need to put more options
[02:06] <buhman> lavf appears to have trouble demuxing h264 from mpegts (it comes up with weird timestamps), but mplayer's mpegts demuxer seems to do the right thing (kindof). Is it possible to tell ffmpeg to decode using mplayer's "mpegts" the same way mplayer can decode with lavf?
[02:06] <buhman> related: http://ffmpeg.org/trac/ffmpeg/ticket/1715
[02:37] <hakermania> Hey there. As far as I know ffmpeg should support wavpack, but the command that works for most audio files: avconv -i input.mp3 -aq 100 output.wv doesn't seem to work here... Using ffmpeg version 0.8.3-4:0.8.3-0ubuntu0.12.04.1. The manpage isn't referring anything about 'Wavpack'. Am I missing something?
[02:43] <trysten> hey
[02:43] <trysten> how do i put subtitles together
[02:44] <buhman> hakermania: "avconv"? are you serious?
[02:44] <buhman> trysten: concatenate them?
[02:46] <trysten> buhman: do you know what is supposed to happen when i run `ffmpeg -i something.srt`
[02:47] <trysten> because i'm getting Invalid data found when processing input on what i know to be a usable srt
[02:47] <buhman> trysten: are you mapping it properly?
[02:48] <trysten> all i'm doing is running `ffmpeg -i something.srt`
[02:48] <trysten> so whatever the normal mapping is
[02:49] <trysten> my scenario is thus: i have 4 files. two video files and two corresponding subtitle files. they are sequential. i want to combine them into one big file.
[02:50] <buhman> erm, so 1) mux the srts in each, making a mpg container 2) concatenate the mpgs 3) possibly remux to another container, mp4 or something
[02:51] <trysten> ah! mpg can hold srt stream?!
[02:51] <trysten> i didn't know that.
[02:51] <buhman> erm, can it? :S
[02:51] <trysten> i'm hoping so XD
[02:51] Action: buhman could be insane
[02:52] Action: trysten is probably insane
[02:55] <physean> you're both insane
[02:58] <hakermania> trysten: I'm 100% serious
[02:58] <hakermania> ffmpeg is always complaining to use avconv instead
[02:58] <ubitux> it's because you're using ubuntu/debian
[02:58] <ubitux> and they aren't packaging ffmpeg but the fork
[02:58] <trysten> hakermania: try #avconv?
[02:59] <ubitux> and they're using propaganda
[02:59] <rud> hmmm -tune zerolatency is tricky. if using -tune zerolatency -s 426x240  >> black image, if using -tune zerolatency -s 640x360 >> image
[02:59] <hakermania> trysten: Ok, replace avconv with ffmpeg if that's the case.. the result is the same!
[03:00] <trysten> hakermania: what is your error
[03:00] <hakermania> trysten: Unable to find a suitable output format for output.wv
[03:01] <ubitux> again
[03:01] <ubitux> you are *not* using ffmpeg
[03:01] <ubitux> but a fork
[03:01] <ubitux> which we can't support
[03:02] <hakermania> ubitux: I said that the command ffmpeg outputs exactly the same thing
[03:02] <ubitux> the ffmpeg you are using is the ffmpeg of the fork
[03:02] <ubitux> which is a very old and broken version
[03:03] <hakermania> ubitux: why is this?
[03:04] <trysten> hakermania: good question. i've unthinkinly used only the SVN of ffmpeg for a long time. most other versions are usually worthless
[03:04] <ubitux> because it's a fork
[03:04] <ubitux> hakermania: http://blog.pkh.me/p/13-the-ffmpeg-libav-situation.html
[03:04] <ubitux> svn of ffmpeg? please, it's years old :p
[03:05] <ubitux> hakermania: TL;DR: search for "deprecated"
[03:05] <hakermania> ubitux: so, you're mainly against Libav?
[03:05] <trysten> right. i meant git.
[03:05] <ubitux> hakermania: no i'm not, but we can't support it
[03:06] <ubitux> this means until you use ffmpeg we can not really help you
[03:06] <hakermania> ubitux: Can I install 'real' ffmpeg through a package of the main repository ?
[03:07] <ubitux> trysten: to concat subtitles you should use a dedicated tool, since you will likely have some desync with the video; look for gaupol or something like that
[03:07] <ubitux> mmh.
[03:07] <ubitux> !ppa
[03:07] <ubitux> mmh.
[03:08] <ubitux> the ppa looks old
[03:08] <ubitux> maybe debian-multimedia
[03:09] <ubitux> hakermania: you can also use this
[03:09] <ubitux> and 3rd solution, you can also complain to your distribution
[03:09] <physean> ubitux, everythingworks
[03:09] <physean> i love you
[03:10] <physean> for sending me the binary
[03:10] <ubitux> i love you too
[03:10] <hakermania> :( I am reading what happened and now I am sad :( Thanks ubitux
[03:10] <ubitux> no problem :p
[03:11] <trysten> damn
[03:11] <trysten> just finished reading that. pretty sad, but also just a product of cooperation
[03:11] <trysten> strife, that is.
[03:12] <physean> dont be sad, be happy
[03:12] <physean> justicevids will live on
[03:15] <hakermania> ubitux: Why the heck did Debian chose Libav over FFmpeg?
[03:16] <ubitux> because the packager(s) is one of the libav folk
[03:16] <ubitux> so they just trust him
[03:16] <ubitux> him(/them?)
[03:17] <ubitux> complain to your distro if you don't agree with such choice
[03:17] <buhman> ubitux++
[03:17] <buhman> hakermania: arch doesn't package libav
[03:18] <buhman> because our maintainers actually think.
[03:18] <buhman> (I hope)
[03:18] <hakermania> ubitux: where can I do such a complaint? I mean, complain where?
[03:19] <ubitux> well it's good if a distro can package libav, but it's better if ffmpeg is packaged as well (this is the case of gentoo afaik), and if it's not using the ffmpeg name or spreading lies.. :p
[03:19] <ubitux> hakermania: depends on your distro
[03:19] <hakermania> Will I be alone to this complaint? Why don't we (?) organize a big complaint? We shouldn't go as individuals...
[03:19] <ubitux> are you using debian? ubuntu? something else?
[03:19] <buhman> hakermania: a bug report
[03:19] <buhman> hakermania: mark it as "defect"
[03:19] <ubitux> there are already various opened bugs on launchpad afaik
[03:19] <hakermania> I'm at ubuntu, will you guys come and mark as affected to the bug report?
[03:19] <buhman> hakermania: no, because that would be a lie
[03:20] <ubitux> https://bugs.launchpad.net/ubuntu/+source/libav/+bug/939863
[03:20] <ubitux> i'm not sure what's the message right now
[03:21] <ubitux> anyway, i don't know where you are supposed to complain
[03:22] <ubitux> maybe the mailing list, opening a new bug, etc
[03:22] <ubitux> it doesn't matter if it's organized
[03:22] <ubitux> just request what you actually want
[03:23] <hakermania> ubitux: I'm afraid that my voice will not be heard. Anyway, why don't you guys do something about it? Can't you just vindicate the fact that ffmpeg is more complete and stable and feature rich than avconv (irregardless of the leader's past) ?
[03:23] <physean> x-large meat lover pizza, 2 hookers, half ounce of cocaine.
[03:23] <ubitux> hakermania: we did, several times
[03:23] <ubitux> on debian mostly
[03:23] <hakermania> ubitux: how and what was the outcome ?
[03:24] <ubitux> no one is motivated enough to make both packages available
[03:24] <ubitux> and the debian choice depends on the packager
[03:24] <ubitux> so it doesn't matter if it's a sane choice or not, the packager decides
[03:24] <physean> what's with avconv
[03:25] <ubitux> it's technically complicated to provide both ffmpeg and libav btw
[03:25] <ubitux> and the debian packaging insanity doesn't help very much
[03:25] <ubitux> also, libav didn't rename the libraries for instance...
[03:26] <ubitux> (it's a problem for apps built against these libs)
[03:26] <hakermania> ubitux: You should really find a solution. Why don't you submit ffmpeg as a package with slightly different name? For example, if I do a sudo apt-get install ffmpeg here I get avconv, and not the 'real' ffmpeg. So, why don't you rename the package? I don't understand why it is so complicated after all.
[03:26] <ubitux> (even though ffmpeg libraries could just replace libav ones technically)
[03:26] <physean> alright
[03:27] <physean> once i convert a FLV with ffmpeg
[03:27] <physean> it comes out as AVI?
[03:27] <ubitux> physean: what do you mean?
[03:28] <physean> i mean i named my file output.avi
[03:28] <physean> but is the container still FLV?
[03:28] <ubitux> hakermania: it's technically complicated because of the libraries and packages built against them
[03:28] <ubitux> i doubt it
[03:28] <ubitux> what's your cmd line?
[03:28] <hakermania> ubitux: 'built against them' means that you have exactly the same libraries names?
[03:28] <hakermania> so it replaces files inside the system?
[03:28] <ubitux> yes, libav didn't renamed them.
[03:29] <physean> ./ffmpeg -i /145.flv -vcodec libx264 -crf 28 -quality best -preset veryslow -vf "movie=watermark.png [watermark]; [in][watermark] overlay=main_w-overlay_w-10:main_h-overlay_h-10 [out]" output-crf28.avi
[03:29] <ubitux> physean: container changed to avi in that case
[03:29] <ubitux> it is automatically detected
[03:30] <physean> so can i write FLV and it will do FLV?
[03:30] <ubitux> sure, replace the .avi into .flv
[03:30] <physean> ok
[03:30] <ubitux> hakermania: both libraries must be binary compatible
[03:31] <ubitux> this means we should be able to switch the libraries without breaking the app
[03:31] <hakermania> ubitux: there's a special field under debian/control especially designed for this situation: 'Breaks' (http://www.debian.org/doc/debian-policy/ch-relationships.html#s-breaks) Am I wrong?
[03:31] <ubitux> that's more complicated than that
[03:31] <ubitux> it means the apps that need these libs must be built against either libav or ffmpeg
[03:32] <ubitux> so you must have 2 different version of each app depending on ffmpeg libraries
[03:32] <ubitux> and there are quite a bunch of them.
[03:33] <ubitux> also, about the "you should really find a solution"; i'm not using debian or ubuntu, and i hate the way packaging works on them. also, i don't have time to maintain all these packages
[03:33] <ubitux> i'm a developer, i write code
[03:33] <ubitux> at times i even write documentation, and review some others code
[03:34] <ubitux> and i even help users sometimes (fear!)
[03:34] <ubitux> i really don't have the time to also maintain packages in distribution i don't use
[03:34] <hakermania> ubitux: what are you using?
[03:35] <ubitux> is that important? :)
[03:36] <hakermania> ubitux: Not so much :( But if only we had ffmpeg in ubuntu... Anyway, thanks for the help, I guess my only chance would be to go ask at #avconv
[03:37] <hakermania> hey
[03:37] <hakermania> there's no such room :P
[03:37] <ubitux> #libav
[03:37] <ubitux> nothing prevents you from using ffmpeg you know
[03:37] <ubitux> the static build are usable standalone
[03:37] <ubitux> and you can build your own version and use it without installing it
[03:37] <hakermania> ubitux: I am developing an application, so it's not quite for personal use
[03:37] <ubitux> ok
[03:38] <hakermania> It would be nice if the users could take it from the official repositories as a dependency
[03:38] <ubitux> go complain to #ubuntu, #libav or whatever :p
[03:38] <ubitux> we can't do much
[03:39] <ubitux> also you're welcome to try packaging that stuff correctly
[03:39] <trysten> hahahaha
[03:40] <hakermania> ubitux: I don't package myself. Thanks god we have found a guy for this job, because I hate it too...
[03:40] <ubitux> and btw, yes ffmpeg supports wavpack
[03:40] <ubitux> ffmpeg -codecs|grep wavpack
[03:40] <ubitux>  D.A.LS wavpack              WavPack
[03:41] <ubitux> ’ decoding, audio codec, lossy+lossless
[03:41] <ubitux> iirc you were trying to encode, that's not supported.
[03:41] <hakermania> ubitux: avconv is successful on this grepping too. So, it is successful only at decoding?
[03:42] <ubitux> it means there is only a wavpack decoder
[03:45] <hakermania> ubitux: do you know if encoding is going to be implemented soon?
[03:46] <ubitux> i didn't hear anything about that
[03:46] <ubitux> why do you need wavpack btw?
[03:47] <ubitux> if you want some lossless audio, we have a native flac encoder
[03:48] <ubitux> and it's also free, widely used, etc
[03:48] <physean> http://www.justicevids.com/buzz-aldrin-punches-moon-landing-denier/
[03:48] <physean> NEVER PAY AGAIN FOR LIVE SEX!
[03:48] <physean> thank you, ubitux!
[03:55] <hakermania> ubitux: the users want it
[03:55] <hakermania> they prefer it over flac
[03:55] <ubitux> is there any technical reason?
[03:55] <ubitux> but sure feel free to open a trac issue with the feature
[03:56] <ubitux> (https://ffmpeg.org/trac/ffmpeg/)
[03:57] <hakermania> ubitux: not a known technical reason. A site reviewed our application and there was a comment requesting wavpack support and it had taken one thousand liikes or so :P So I guess we have to hear the people's voice
[03:57] <hakermania> ubitux: what can I do with the link you sent me?
[03:58] <ubitux> open a ticket with a feature request for a wavpack encoder
[04:00] <hakermania> not sure how to do this but I guess I will figure it out
[04:04] <hakermania> ubitux: https://ffmpeg.org/trac/ffmpeg/ticket/1716   Did I do anything wrong?
[04:07] <ubitux> looks ok
[04:07] <ubitux> note that it's not a trivial thing to write
[04:07] <ubitux> so it won't be written in two days
[04:08] <DeviceZer0> hello all. I'm playing around with converting a video taken with my cellphone(h264 avc1 1900x1080) to something like flash flv. The example I found when searching around on google is using -qscale. I noticed when running it says: "Please use -q:a or -q:v, -qscale is ambiguous" has qscale been change or removed?
[04:08] <DeviceZer0> Trying -q:v 1 then -q:v 31 seems to have no change on output file
[04:08] <DeviceZer0> nor does -qscale 1 or -qscale 31
[04:09] <DeviceZer0> the full command I am using is: ffmpeg -i 20120830_182839.mp4 -ar 44100 -qscale 1 -vcodec libx264 output-qscale1.flv
[04:10] <ubitux> -q is an alias for -qscale
[04:10] <ubitux> :v is for video, :a for audio
[04:10] <DeviceZer0> hmm.
[04:10] <ubitux> it should affect the output, compare the sizes & md5
[04:10] <DeviceZer0> can you do just -q? or does it need :a or :v?
[04:11] <DeviceZer0> seems it works with just -q
[04:11] <ubitux> well you likely want a different factor for audio & video
[04:12] <DeviceZer0> true.
[04:12] <hakermania> ubitux: thanks for everything! Bye!
[04:12] <DeviceZer0> I am very new to doing audio/video stuff...
[04:14] <DeviceZer0> so checking md5sums they are diff files and the file size is different...but not by much. -q 1 = 28199229 and -q31 = 27866779
[04:16] <ubitux> you're using libx264 to encode so i'd suggest yo use the crf option
[04:16] <ubitux> i thought it was passed along but maybe i'm wrong
[04:16] <DeviceZer0> hmm
[04:16] <ubitux> anyway, try to add a -crf 21 or something
[04:17] <DeviceZer0> and remove the -q/-qscale? or leave it in?
[04:17] <ubitux> i believe the -qscale option will affect the audio encoding in your case
[04:17] <ubitux> try maybe -aq 3 -crf 20 or something
[04:18] <DeviceZer0> ok. will try now
[04:18] <ubitux> you might want to play with the various libx264 options
[04:19] <ubitux> (like preset, tune, profile, etc)
[04:19] <DeviceZer0> yea. I was planning on eventually. Just wanted to do some quick testing
[04:20] <DeviceZer0> there is just *so* much to learn.
[04:31] <trysten> ubitux: thanks for being so helpful!
[04:32] <ubitux> am i?
[04:32] <trysten> a comment on your extroversion, not efficiacy
[04:32] <trysten> efficacy*
[04:33] <ubitux> :)
[04:33] <physean> yo
[04:34] <physean> im doing ffmpeg -vcodec h264
[04:34] <physean> hmmm
[04:35] <physean> i need to upscale a few videos, but i have 200 videos
[04:35] <physean> and only some are low-res
[04:35] <physean> is there a way i can print the resolution of the videos?
[04:36] <trysten> physean: when i run ffmpeg -i something it tells me the resolution on the stream line
[04:38] <ubitux> use ffprobe
[04:39] <physean> hmmm
[04:39] <physean> pretty smart
[04:40] <ubitux> ffprobe -show_streams -show_format -v 0 -of json input.mp4
[04:40] <ubitux> for example.
[04:41] <physean> so let's say i wants the watermark to be a fixed width
[04:43] <physean> coul i write a script to resize the watermark before each encode
[04:43] <physean> or would it be better to upscale all the videos?
[04:45] <ubitux> i don't understand what you want to do
[04:45] <physean> im sorry
[04:45] <physean> im putting watermarks on videos
[04:45] <physean> but the resoultion of the video varies
[04:46] <physean> sometimes it's actually smaller than the watermark width
[04:48] <physean> so i think the correct thing to do is to resize the watermark before each encoe
[04:50] <ubitux> you have access to the size of your watermark
[04:50] <ubitux> as well as the size of the surface you put it on
[04:50] <physean> right
[04:50] <ubitux> so you can do various math with them
[04:50] <physean> so i have to write a script
[04:51] <ubitux> you don't need to
[04:51] <ubitux> you can use expressions
[04:51] Action: physean doesn't follow
[04:51] <ubitux> what's your current command line?
[04:53] <physean> ./ffmpeg -i /$1.flv -vcodec libx264 -crf 28 -quality best -preset veryslow -vf "movie=watermark.png [watermark]; [in][watermark] overlay=main_w-overlay_w-10:main_h-overlay_h-10 [out]" $1.flv
[04:54] <ubitux> do you understand what overlay=main_w-overlay_w-10:main_h-overlay_h-10 does?
[04:54] <physean> very little
[04:54] <physean> i think the 10 is for moving pixels
[04:54] <physean> padding
[04:55] <ubitux> it means x = main_w-overlay_w-10 and y = main_h-overlay_h-10
[04:55] <ubitux> main_w and main_h are the input width and height
[04:56] <ubitux> overlay_w and overlay_h are the overlay width and height
[04:56] <ubitux> and anyway, you can do some crazy math
[04:56] <ubitux> for example (main_w-overlay_w-10)*2+50/overlay_h
[04:56] <ubitux> or anything stupid
[04:56] <ubitux> (there is no meaning in the expression i just gave)
[04:57] <physean> WHOA
[04:57] <ubitux> try to brainstorm something according to what you want
[04:57] <physean> well i don't totally understand
[04:57] <physean> the 10 means what exactly
[04:57] <ubitux> 10 pixels
[04:57] <physean> crop 10 pixels from the overlay?
[04:58] <physean> so, it's centered?
[04:58] <ubitux> it means it will put the picture at x = main_w - overlay_w - 10
[04:58] <physean> - as in minus?
[04:58] <ubitux> so if your input is 800x600, your overlay is 200x100
[04:58] <ubitux> x will be 800 - 200 - 10
[04:58] <physean> ohhhhhhh
[04:58] <physean> im getting it now
[04:59] <ubitux> so the image will be at x = 590
[04:59] <ubitux> similar for the y
[04:59] <physean> gotcha
[04:59] <physean> so i can multiply the overlay too?
[04:59] <physean> wtf
[04:59] <physean> i'll have to ask you about this tomorrow, unless you can refer me to a place
[05:00] <ubitux> now what you might want is play with the ratio between your overlay and the picture
[05:00] <ubitux> so main_w/overlay_w for example
[05:00] <ubitux> but well, i don't know exactly what you want
[05:00] <physean> you got it
[05:00] <ubitux> so adapt to your needs, it's elementary math
[05:00] <physean> yeah, once i understand the basics
[05:01] <physean> it will be easy to adapt
[05:01] <physean> but that's the position?
[05:01] <ubitux> yes
[05:01] <physean> well i want it to be smaller
[05:01] <physean> not just moved
[05:01] <ubitux> oh, ok
[05:01] <ubitux> then with the scale filter
[05:02] <physean> ok
[05:02] <ubitux> not sure about that then
[05:02] <ubitux> since you will likely not be able to access the two sizes
[05:02] <ubitux> meh i'm too tired i need to sleep
[05:02] <ubitux> 'night ppl
[05:03] <physean> not be able to access the two sizes
[05:03] <physean> what sizes
[05:03] <physean> can't i print the size
[05:03] <physean> nvm
[05:03] <physean> goodnight nigga
[07:54] <manticore> i found a huge but in ffplay.exe
[07:54] <manticore> i found a huge bug in ffplay.exe
[07:58] <cbsrobot-> Guest1580: this big : http://msnbcmedia.msn.com/j/MSNBC/Components/Photo/_new/111201-coslog-insect-415p.photoblog900.jpg ?
[07:58] <Guest1580> huh
[07:58] <Guest1580> software bug
[07:59] <cbsrobot-> show me the picture
[08:00] <cbsrobot-> Guest1580: just kidding, but please be more specific &.
[08:00] <Guest1580> it plays with weird noise in the background when playing .mlp files
[08:03] <cbsrobot-> you could report it on http://ffmpeg.org/trac/ffmpeg/report
[08:03] <cbsrobot-> read http://ffmpeg.org/bugreports.html before you post it
[08:04] <Guest1580> Input #0, mlp, from 'a.mlp':
[08:04] <Guest1580>   Duration: N/A, start: 0.000000, bitrate: N/A
[08:04] <Guest1580>     Stream #0:0: Audio: mlp, 88200 Hz, stereo, s32
[08:04] <Guest1580>    5.99 A-V:  0.000 fd=   0 aq=    0KB vq=    0KB sq=    0B f=0/0
[08:04] <cbsrobot-> 88200 Hz - never saw that before
[08:05] <Guest1580> Input #0, mlp, from 'b.mlp':
[08:05] <Guest1580>   Duration: N/A, start: 0.000000, bitrate: N/A
[08:05] <Guest1580>     Stream #0:0: Audio: mlp, 96000 Hz, 5.1, s32
[09:48] <coalado1> I'd like to set my conversion output to a http stream like  -f mp3 http://127.0.0.1:888
[09:48] <coalado1> is this possible?
[10:01] <Tjoppen> "yes", depending on the receiver
[10:05] <coalado1> I just noticed... ffmpeg posts the results, but does not open an own server, right?
[10:05] <Tjoppen> correct
[10:05] <Tjoppen> try ffserver I suppose.  I've no idea how it works though
[13:26] <shroomM> umm... i know the bug report isn't helpful, but maybe any idea how to provide you guys more information ...
[13:26] <shroomM> https://ffmpeg.org/trac/ffmpeg/ticket/1574
[13:30] <Tjoppen> can you fix the encoder?
[18:59] <AlRazi> my configuration: --enable-static --enable-shared --enable-gpl --enable-nonfree --enable-libmp3lame --enable-libaacplus yields Segmentation fault (core dumped)
[19:47] <jabils> hello everyone
[19:47] <jabils> anyone here have experience installing FFMPEG on a Solaris machine?
[19:47] <jabils> i am running into a library issue.
[20:05] <llogan> burek: haven't seen you here in a while
[20:15] <darkstarbyte> Is there software that allow me to view a video and generate cut points for ffmpeg?
[20:17] <darkstarbyte> maybe a script that I can use with ffmpeg that allows me to specify cut points based on what I could write down while using vlc.
[20:34] <relaxed> darkstarbyte: I feel like the select filter should achieve this but it's lacking.
[20:36] <darkstarbyte> Do I really need a full on video editor?
[20:36] <darkstarbyte> I was thinking something like
[20:36] <relaxed> probably
[20:36] <darkstarbyte> I have another thought
[20:37] <darkstarbyte> Could it do something like
[20:37] <darkstarbyte> -cut from_HH:MM:SS:xxx to_HH:MM:SS:xxx
[20:40] <darkstarbyte> time to sync some directories, and start video editing on my fast comptuer
[20:40] <darkstarbyte> computer
[20:50] <burek> llogan, yeah, some private issues emerged, but should be ok now
[20:50] <burek> :beer: :)
[21:04] <saste> darkstarbyte: segment -segment_times
[21:05] <llogan> burek: would it be possible for fflogger to announce new bugs on -devel? of course i don't know if people would like the extra noise
[21:05] <llogan> since we have a ffmpeg-trac mailing list
[21:07] <saste> llogan: not a bad idea, but better to discuss that in #ffmpeg-devel
[21:08] <llogan> i just wanted to see if it was possible first
[23:09] <burek> llogan, sure, it is possible
[23:10] <burek> just see with other people if they agree about it and we can make it happen
[00:00] --- Wed Sep  5 2012