[Ffmpeg-devel-irc] ffmpeg.log.20141101

[burek]

[00:23] <edman007> alright, I found my problem! and it's ffmpegs fault... I think, my output mpeg only had VOBUs for the first 60s or so... why?
[00:24] <edman007> why is that, and can I do DVD muxing with ffmpeg?
[00:24] <edman007> mplex worked fine
[01:14] <t4nk571> Hello everyone, I'm hoping to find someone to point me in the right direction about ffmpeg pipes?
[01:15] <c_14> |||||||||
[01:15] <c_14> Pipes.
[01:15] <c_14> Pipes everywhere.
[01:15] <c_14> What exactly are you trying to do?
[01:16] <t4nk571> So...in short...pipe all thumbnails generated from a video over to another process
[01:16] <t4nk571> So that the secondary process can crunch the images real time.
[01:16] <c_14> just use pipe:1 as the output file
[01:16] <t4nk571> so...the command would be something like this...?
[01:16] <c_14> Not sure if that works for images.
[01:16] <c_14> hmm
[01:17] <c_14> let me see
[01:17] <t4nk571> ffmpeg -i input.avi image2 -vf fps=fps=1/60 pipe:1 "other command?"
[01:18] <c_14> -f image2pipe
[01:18] <c_14> And it'l bee pipe:1 | other command
[01:18] <c_14> eh -e
[01:19] <t4nk571> the secondary process is actually the facedetect module available here http://www.thregr.org/~wavexx/software/facedetect/
[01:19] <t4nk571> what's the -e? Could you give me a complete example?
[01:19] <c_14> That was just me correcting my mispelled bee
[01:21] <t4nk571> so for every image that ffmpeg dumps...it would send that .png or jpeg over to stdout
[01:21] <t4nk571> ?
[01:21] <c_14> it should
[01:21] <t4nk571> ok
[01:21] <c_14> Not entirely sure what output it defaults to though...
[01:21] <t4nk571> so for that example I would leave off the image%3d.jpg?
[01:21] <c_14> yes
[01:22] <c_14> Or else you'll be writing to the filesystem again.
[01:22] <t4nk571> ok
[01:22] <t4nk571> one more question
[01:22] <t4nk571> would ffmpeg run faster from the pre-compiled package on ubuntu, or compile from source?
[01:23] <c_14> The compiled from source package will presumably be newer and have more fixes/improvements than the pre-compiled one.
[01:23] <c_14> Whether or not that makes it faster depends.
[01:24] <t4nk571> if I put that package in /usr/local/bin can I run it from anywhere?
[01:25] <c_14> If /usr/local/bin is in your PATH, yes.
[01:27] <t4nk571> sweet...lemme give this a whirl, lol. Thanks!
[01:35] <t4nk571> so if the second command would normally go like this...'facedetect -q image.jpg' ...how do I tell it that the info is coming from stdin?
[01:43] <t4nk571> c_14?
[01:44] <c_14> Usually with -
[01:44] <c_14> facedetect -q -
[01:47] <t4nk571> lemme see here
[01:47] <t4nk571> so this is what I have thus far...and it isn't working quit yet
[01:48] <t4nk571> ./ffmpeg -i input.avi -f image2 -vf fps=fps=1/600 pipe:1
[01:48] <t4nk571> oops
[01:48] <t4nk571> hit enter too soon
[01:49] <t4nk571> ./ffmpeg -i input.avi -f image2 -vf fps=fps=1/600 pipe:1 | facedetect -q - | echo $? >> testlog.txt
[01:50] <t4nk571> It writes the exit code one time in the log file
[01:50] <t4nk571> And while it is running it states ignoring invalid SAR: 0/0:00.00 bitrate=N/A
[01:53] <c_14> I think you might need image2pipe
[01:53] <t4nk571> I tried that in place of the image2
[01:53] <t4nk571> -f image2pipe
[01:53] <t4nk571> same result
[01:54] <kevmitch> does facedetect actually support "-" as meaning read from stdin?
[01:56] <t4nk571> I don't know that for sure
[01:56] <t4nk571> it would normally take a path to an image
[01:57] <t4nk571> any idea of another command I could test as the secondary process?
[01:57] <kevmitch> you could try facedetect -q <(./ffmpeg -i input.avi -f image2 -vf fps=fps=1/600 pipe:1)
[01:57] <kevmitch> that might be a bash thing
[01:57] <kevmitch> it might not even work with something other than text
[01:58] <kevmitch> if this is it http://www.thregr.org/~wavexx/software/facedetect/ it doesn't even look like it can handle more than one image at a time
[01:58] <t4nk571> that is it
[01:59] <t4nk571> Could ffmpeg be slowed down to handle it or something?
[01:59] <t4nk571> Drop frames on purpose I guess?
[01:59] <t4nk571> That site shows some bash script to batch process, but that's about it
[02:01] <t4nk571> Or maybe another way to fork a process that crunches the images as ffmpeg dumps them I guess?
[02:01] <t4nk571> I'm not that great with bash yet.
[02:01] <t4nk571> More into python...but weak at that one too, lol.
[02:03] <kevmitch> if you could get ffmpeg -i "input.mov" -an -f image2 "output_%05d.jpg" to output the name of each image after it's written
[02:04] <t4nk571> hmmmm...maybe echo out to a file or something?
[02:04] <kevmitch>  you could pipe that into "| while read file;do
[02:04] <kevmitch> facedetect  $file
[02:04] <kevmitch> done
[02:05] <t4nk571> can you be specific with the syntax? I'm kinda rough on bash
[02:05] <t4nk571> my apologies
[02:05] <t4nk571> I do appreciate the help
[02:05] <t4nk571> irc's are amazing
[02:05] <kevmitch> ./ffmpeg -i input.avi -f image2 -vf fps=fps=1/600 "output_%05d.jpg" | while read file;do
[02:05] <kevmitch> facedetect  $file
[02:05] <kevmitch> done
[02:06] <kevmitch> but that probably won't work because I don't think ffmpeg does output just the image file name to stdout
[02:06] <kevmitch> you need to get it to do that
[02:07] <t4nk571> this seems like something that would be kind of common...that image2pipe call makes me think there has to be some way to write the image and pass it to a process?
[02:07] <t4nk571> I could settle for running another command (facedetect) and have it cleaning up as ffmpeg dumps images
[02:08] <t4nk571> two processes I guess. But I'm not sure that ffmpeg would need to fork or anything...just not sure how to launch two calls at once or what have you..
[02:08] <kevmitch> yes, but that process has to accept data from stdin, but it looks like facedetect only reads single files
[02:09] <t4nk571> so could facedetect crunch on a ls *.jpeg at the same time or something?
[02:09] <t4nk571> It reads a bit slower than ffmpeg dumps
[02:09] <t4nk571> from what I've seen at least
[02:11] <kevmitch> not sure about the performance vis a vis ffmpeg+facedetect, but you might get a cleaner implementation using opencv, which can both decode video to images and perform face detections on individual frames. python bindings make it relatively easy
[02:12] <t4nk571> really? I looked at that library rather quickly and can admit it was huge...and facedetect looked so simple, :-)
[02:12] <t4nk571> taking the easy way out doesn't always pay...
[02:13] <kevmitch> yeah, take a look here
[02:13] <kevmitch> http://docs.opencv.org/trunk/doc/py_tutorials/py_gui/py_video_display/py_video_display.html
[02:14] <kevmitch> you just need to put a face detect on each image in the loo
[02:15] <kevmitch> just give a filename to VideoCapture rather than a number
[02:15] <t4nk571> wow...that doesn't seem bad at all...awesome...as always...thank you very much!
[02:17] <t4nk571> so if I want to make ffmpeg widely available to the opencv library
[02:17] <t4nk571> I have downloaded a precompiled version and have been doing the ./ffmpeg from within that downloaded folder
[02:18] <t4nk571> do I copy the folder or just the contents over into /usr/local/bin so that other processes that depend on it can use it? Im on ubuntu
[02:18] <t4nk571> and Ubuntu's package of ffmpeg doesn't have the necessary filters to do the image dumping at predefined intervals
[02:22] <kevmitch> if you install ffmpeg into an isolated prefix with configure --prefix="$BUILD"/build_libs
[02:23] <kevmitch> then configure opencv with PKG_CONFIG_PATH="$BUILD"/build_libs/lib/pkgconfig ./configure
[02:23] <kevmitch> or something like that
[02:23] <t4nk571> k
[02:23] <t4nk571> thanks again
[02:23] <kevmitch> or you could just make install ;)
[02:23] <kevmitch> and opencv would presumably just find it
[02:23] <t4nk571> i had repeated failures building ffmpeg from source
[02:23] <t4nk571> would make it all the way to the end and it would fail on the part where ffmpeg wraps it altogether
[02:23] <t4nk571> no clue wh
[02:24] <t4nk571> why
[02:24] <kevmitch> oh, so you have a partially completed build ?
[02:25] <kevmitch> if so, opencv will have a hard time using that
[02:25] <kevmitch> in any case, the above example should work fine with both opencv and libav inthe ubuntu repositories
[04:38] <edman007> Hrm, why is when I add -target ntsc-dvd when encoding a video it wont work with -ss or -t set as well, removing -ss/-t or -target makes it work
[09:52] <cjasonb> question - I am looking around the ffmpeg website for list of REQUIRED libraries for compile.  (Something in a recent Suse update broke my ffmpeg adn I'm trying to find out what so I can roll it back)  Is there a place on the site that provides sucha list?
[09:53] <relaxed> cjasonb: broke it how? can you pastebin the error?
[09:58] <cjasonb> no error in the code output.  FFmpeg was being used with the following command  ffmpeg -f alsa -i hw:1 -f v4l2 -i /dev/video0 (etc. etc) to capture from a pinnacle dazzle usb device.  Was working flawless.  Now getting the video as green video, with very chopy sound.  Not sure if issue is v4l2 or some underlying render lib.  Hoping to find a list so I can compare updated libs with required libs
[09:59] <cjasonb> could also have something to do with data strem from the usb port I suppose.
[10:00] <relaxed> can you capture with something else? Have you read https://trac.ffmpeg.org/wiki/Capture/Webcam ?
[10:01] <cjasonb> have not read, will chaeck now.  Unfortunately dazzle device is all I own.  Project is converting 100+ VHS tapes for my local church to DVD.
[10:02] <cjasonb> oh, and I am the only person in the world who does not own a web came, so loop back isnt going to work atm
[10:04] <cjasonb> should point out I had to revert to a 2.3.3 ffmpeg to even get it to work in the first place.  Right now I am looking for more of a "where is the error" so I know what route to take in seeking a fix, than an actual solution (unless someone knows one)
[10:05] <relaxed> what happened when you tired 2.4.2?
[10:07] <relaxed> maybe suse updated your kernel and the newer driver is causing the issue. If you can try booting into an older kernel if it's still around.
[10:07] <cjasonb> hmmmmm
[10:07] <cjasonb> kernel worth a look
[10:07] <cjasonb> hadnt thought of that
[10:09] <cjasonb> the thing that realy bugs me is this.  I installed suse 13.1 (I know 13.2 comming out in days), delayed it's post install update, got ffmpeg doing what I wanted, and then updated, now its broke again.  Logging off to reboot an older kernel.  BBL if still issues
[10:27] <cjasonb> well, kernel does not appear to be the issue.  Previous motv output (motiff based tv front end) was flawless, now it is black screen and realy scratchy audio.  FFMPEG now locks up until I press ctrl-c right after identifying Input #0  (the audio portion).  Video is locking.  Jumping to pastebin to set up a link
[10:31] <cjasonb> pastebin output  http://pastebin.com/BJ81qkaf
[10:32] <cjasonb> the parameter error I am unsure of.  That EXACT line worked previously (copy pasted it to chell)
[10:32] <cjasonb> ^^shell
[10:34] <cjasonb> if i am guessing right, the issue isnt ffmpeg, but a sys lib or a dependancy of v4l2
[10:38] <relaxed> cjasonb: pastebin the output of v4l2-ctl -L
[10:38] <cjasonb> brb
[10:43] <cjasonb> sorry for delay, real life jumped in and bit me   http://pastebin.com/kUwP8QJu
[10:44] <relaxed> did you try viewing the video with anything else, like "mpv tv://" ?
[10:45] <cjasonb> only motv installed and that was zilch.  trying to grab a new viwer...brb
[10:46] <relaxed> sounds like this isn't an ffmpeg issue
[10:47] <cjasonb> im stck behind dsl...these others a re big and will take a while.
[10:47] <cjasonb> I tend to agree with you
[10:47] <cjasonb> was simply asking here first since I figured faster answer here
[10:48] <relaxed> I would google for your capture device + linux and sort by most recent.
[10:48] <cjasonb> been trying that for a few days now  (more)
[10:49] <cjasonb> got rid of ubuntu for suse, and being unfamiliar have done a lot of googling.  The device is trash for linux support unfortuantely.  BUt it is all the church had, so Im stuck with it
[13:12] <emilsedgh> Hi everyone. Im using ffmpeg as my encoder. this is the command I use:
[13:12] <emilsedgh> (well, avconv to be precise)
[13:12] <emilsedgh> avconv -y -f alsa -i hw:2,0 -f video4linux2  -i /dev/video1 -vcodec libx264 -b:v 200k -b:a 50k -strict -2 -acodec aac -f flv rtmp://79.175.176.66:1935/live/myStream
[13:13] <emilsedgh> my problem is that video and audio are out of sync. audio is ~3 seconds delayed
[14:28] <zap0> can ffmpeg do 1 operation on a bunch of files?
[14:29] <Mavrik> not by itself, write a script
[14:58] <kevmitch> better yet use xargs
[14:59] <kevmitch> ls lots*of*files | xargs -i ffmpg -i '{}' <stuff>
[16:10] <ztane> anyone by accident happens to have ffmpeg built for raspbian?
[16:27] <sacarasc> kevmitch: Would that run a separate ffmpeg for each file? Because otherwise it's not going to work.
[20:45] <anshul_mahe> What is BFF and TFF in [Parsed_idet_0 @ 0xb0eb780] Single frame detection: TFF:0 BFF:0 Progressive:49 Undetermined:0
[20:46] <anshul_mahe> can any one link me documentation, I searched
[20:48] <c_14> Bottom Field First vs Top Field First
[20:48] <anshul_mahe> c_14: thanks
[00:00] --- Sun Nov  2 2014