[Ffmpeg-devel-irc] ffmpeg.log.20150125

[burek]

[00:30] <t4nk455> hello
[00:31] <t4nk455> hello to all i have live video source and have movie file have fram inserted on black scream  i need to insert this movi to run overlay the live video and when the movie finish keep the live video output as live to rtmp
[00:32] <t4nk455> hello
[00:34] <t4nk455> hello to all i have live video source and have movie file have fram inserted on black scream  i need to insert this movi to run overlay the live video and when the movie finish keep the live video output as live to rtmp
[00:35] <t4nk455> any one can help me
[01:14] <wodim> is there something like delogo but that accepts some kind of mask? i mean a square is not very good, vlc can do a shape for example
[01:20] <Kuukunen> wodim: http://avisynth.nl/index.php/Rm_logo http://avisynth.nl/index.php/InpaintFunc
[01:21] <wodim> thank you for the idea
[01:24] <zumba_addict> hi folks, I'm a bit confused with the quality of video playing versus the extracted image as a jpeg. When I pause a video and capture a screenshot, it is 50% better than the quality that I got after running ffmpeg -i WP_20150124_14_28_46_Pro.mp4 -r 30000/1001  image-%d.jpeg
[01:25] <c_14> try setting -q:v to something lowish (at least I think low is better)
[01:26] <zumba_addict> ok, I havent' tried that parameter
[01:26] <zumba_addict> can you give an example param please?
[01:27] <zumba_addict> what is the minimum/max value we can assign to -q:v
[01:27] <zumba_addict> just saw this too, http://stackoverflow.com/questions/10225403/how-can-i-extract-a-good-quality-jpeg-image-from-an-h264-video-file-with-ffmpeg
[01:31] <zumba_addict> -q:v 2 improved it
[01:36] <Sparrow99> hi how can I pointing nvEncodeAPI.h with ffmpeg+enable-nvenc ?
[04:17] <zumba_addict> how come this didn't order the numbers correctly? ffmpeg -i image_%04d.jpg -c:v libx264 -r 30000/1001 -crf 22 -preset ultrafast -pix_fmt yuv420p out.mp4
[04:17] <c_14> define "order the numbers correctly"
[04:18] <zumba_addict> i have files from image_0001.jpg to image_1861.jpg
[04:18] <zumba_addict> the files image_0997.jpg to image_0999.jpg got move to another location
[04:18] <c_14> They should be in the right order.
[04:19] <c_14> It's always worked for me.
[04:19] <zumba_addict> it always worked for me too
[04:19] <zumba_addict> but here is what I noticed
[04:19] <zumba_addict> when I do an ls -lrt, the files aren't in order
[04:19] <zumba_addict> looks like ffmpeg is looking at the timestamp rather than sequence
[04:22] <zumba_addict> snippet output of ls -lrt - http://pastebin.com/ff7Y2ys8
[04:23] <zumba_addict> any ideas?
[04:24] <c_14> What's the first image index?
[04:25] <zumba_addict> 0001
[04:25] <zumba_addict> everything looks fine except 0997-0999
[04:26] <c_14> Have you looked at those images in an image viewer to make sure they weren't renamed by accident?
[04:28] <zumba_addict> i figure it ouw
[04:28] <zumba_addict> out
[04:28] <zumba_addict> looks like lightroom doesn't care when it saves it
[04:28] <zumba_addict> here is what i'll do to fix it
[04:28] <zumba_addict> ls -1 | sort which sorts them properly
[04:29] <zumba_addict> then i will recopy them to another folder which will save them on time properly
[04:29] <zumba_addict> so looks like a bug in ffmpeg too
[04:29] <zumba_addict> i looked at the files and the images are correct
[04:29] <zumba_addict> the sequence are correct too
[04:29] <zumba_addict> but the time they are saved is not in order
[04:43] <zumba_addict> i got disconnected
[04:43] <zumba_addict> so it's a bug in ffmpeg right?
[04:44] <c_14> If you can find a way to reproduce it, yes. (or even if you can't if you can provide part of the source files that it happens with, and it always happens with those files)
[04:45] <zumba_addict> what I'm doing now is resaving the files in the correct timestamp
[11:28] <Stu|> quick question, what do people use as a gui for ffmpeg on ubuntu?
[11:29] <zumba_addict> hey guys, what's a good title for my youtube for our snow tubing today? I was thinking highspeed snow tubing but there is already in youtube
[11:45] <pross> snow tubing?
[11:52] <filterAMI> I am getting error while using scale filter: Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height
[11:54] <zumba_addict> here it is, what do you think about my video :) I used ffmpeg for this. WOOHOO on ffmpeg - https://www.youtube.com/watch?v=TRX-VyFO2JU
[11:59] <RobertNagy> I'm getting a lot of "h264 reference count overflow" errors when playing a ffmpeg generated rtmp stream in ffplay. What might be causing this and how can I avoid it?
[12:00] <RobertNagy> I'm running the stream through nginx if that might affect something...
[13:37] <wodim> mpeg4 (native) -> h264 (libx264)
[13:37] <wodim> i'm doing this (480p) and the movie has the same size (around 700MB)
[13:37] <wodim> what gives?
[13:41] <relaxed> what's your command?
[14:24] <wodim> relaxed: ffmpeg -i El\ Dúo\ Sacapuntas\ -\ Yo\ quiero\ ser\ torero.avi -c:v libx264 -c:a libfaac -f yo_quiero_ser_torero.mkv
[14:24] <wodim> relaxed: from 709M to 539M
[14:24] <wodim> i thought it would be less than 300M at least...
[14:32] <__jack___> wodim: you can reduce that further, by reducing quality
[14:33] <wodim> __jack___: i don't really want to, i just thought the resulting size would be much different. my mistake i guess?
[14:34] <__jack___> encoding an already encoded source (something not lossless or not almost-lossless) is not that awesome, because the #2 encoder will try to keep #1 encoder "waste"
[14:34] <__jack___> #2 will keep #1 mistakes
[14:39] <wodim> no fix for that even if it's much slower?
[15:04] <hid> hello
[15:28] <relaxed> wodim: add -preset veryslow
[15:28] <wodim> ok thank you
[15:28] <relaxed> you might want to copy the audio
[15:29] <wodim> maybe because i don't think there will be much of a difference
[15:37] <hid> how do you include neroAacEnc with ffmpeg?
[15:48] <relaxed> hid: it's not possible, use libfdk_aac instead
[15:48] <hid> relaxed: ok thank you :)
[16:06] <kc8hfi> i have this,  http://fpaste.org/174371/42219827/   the video and audio are in sync for a good while, but it slowly gets out of sync.  would pulseaudio be causing that?
[16:07] <kc8hfi> I'm not sure how to handle lines 25-28.
[16:08] <kc8hfi> mediainfo on the resulting file,  http://fpaste.org/174372/19848414/
[16:14] <kc8hfi> the tee'd stream didn't work though.  i used ffplay udp://localhost:1234 to play that stream,  http://fpaste.org/174373/22198815/    tons of errors.
[18:19] <kc8hfi> this is the command,  http://fpaste.org/174393/22062551/   that should create a udp stream,  so when I run this command, http://fpaste.org/174394/22206338/  you can see the error at line 14
[18:23] <c_14> Try f=mpegts instead of f=matroska
[18:26] <kc8hfi> i think i need to tell ffplay what the codec is,  http://fpaste.org/174397/42220673/
[18:32] <jbmcgx> hey there, I was wondering if anyone on here would be able to help me figure out why ffmpeg doesn't seem to be utilizing all the vCPUs available on the server we're running it on
[18:32] <jbmcgx> we're running a c4.8xlarge EC2 instance that has 36 available threads (18 cores with 2 threads on each core). When I run "top" in terminal and check out the CPU usage, ffmpeg seems to be doing the majority of the work (~90%) on one thread and only using 3-4 threads in total.
[18:33] <jbmcgx> our ffmpeg is built from source with "--enable-pthreads". I've tried adding "-threads 0" to the encoding settings and also "-threads 36" but the result is similar
[18:33] <c_14> jbmcgx: what are you doing? complete commandline please (preferably in a pastebin)
[18:40] <kc8hfi> this should work,  http://fpaste.org/174400/20764314/
[18:44] <c_14> Those options don't have any effect in ffplay.
[18:44] <jbmcgx> c_14: hopefully you can make sense of this, its a pretty hefty command... http://pastebin.com/X3A5S1uF
[18:47] <jbmcgx> c_14: essentially what it is, is a video of someone fiddling with a phone and we are sending in motion tracking coordinates and overlaying a frame sequence to alter what is shown on the screen of the phone
[18:49] <jbmcgx> c_14: it also applys an alpha mask as the person's thumbs overlap the phone during the frame sequence we're overlaying
[18:51] <jbmcgx> and were using a modified version of the perspective frei0r plugin (perspective2) to handle the motion tracking coordinates (4 corners, not a flat rectangle)
[18:51] <c_14> oh boy, The deadlock is most likely somewhere in that filtergraph. I'm not sure if filtergraphs as a whole support threading
[18:51] <c_14> The easiest way to get it to thread better would be to split it into several ffmpeg commands, encode using a lossless codec and combine the results.
[18:54] <jbmcgx> so you mean like split the video into chunks to encode, run those at the same time and then concat the result?
[18:58] <c_14> If you can do that easily, ye. That would be your best bet. If you can't easily do that you'd have to look into each of the filters in the filtergraph, how the handle threading and work on improving that...
[19:00] <jbmcgx> c_14: do you know of an easy way to tell when all those chunks have finished? or would we need to make some kind of a system around it to handle that (like to know when the last chunk has finished and continue onto the next command)
[19:05] <c_14> If you fork all the commands as subprocesses, you can wait on them.
[19:10] <jbmcgx> ok gotcha - and then set threads to 0? or is that basically not doing anything
[19:12] <c_14> That's only effecting the x264 encoding, and it's 0 by default
[19:15] <jbmcgx> ok cool - thanks for the help!
[19:46] <zumba_ad_> hey guys, is it possible in ffmpeg to combine video files and overlapping the start and end while fading in and fading out?
[20:56] <cousin_luigi> What does --enable-pic do? I mean, isn't it enough to add -fPIC to CFLAGS?
[20:57] <JEEB> simpler way to do that basically
[21:01] <cousin_luigi> rpmlint complains if I don't add the switch manually
[21:01] <cousin_luigi> I'm not sure why
[21:01] <cousin_luigi> perhaps it only checks the gcc line
[21:02] <JEEB> maybe it only checks CFLAGS ?
[21:02] <JEEB> instead of what is actually used
[21:02] <JEEB> make V=1 will show you all the gritty details, IIRC
[21:12] <cousin_luigi> JEEB: Will -fPIC work on arm?
[21:13] <cousin_luigi> There were all sorts of weird arch options in this build script.
[21:48] <cousin_luigi> gn
[22:04] <zumba_ad_> hey guys, I processed this in ffmpeg :) https://www.youtube.com/watch?v=TRX-VyFO2JU
[23:49] <zumba_ad_> is there a difference between -framerate 30 vs -r 30000/1001
[23:49] <c_14> yes
[23:49] <zumba_ad_> i always use -r
[23:50] <zumba_ad_> just saw framerate from this link, http://video.stackexchange.com/questions/7903/how-to-losslessly-encode-a-jpg-image-sequence-to-a-video-in-ffmpeg
[23:50] <zumba_ad_> how do they differ?
[23:50] <zumba_ad_> i know that when i use 30000/1001, it's actually 29.97
[00:00] --- Mon Jan 26 2015