[Ffmpeg-devel-irc] ffmpeg.log.20150530

burek burek021 at gmail.com
Sun May 31 02:05:01 CEST 2015


[00:59:57 CEST] <treeinterest> hi, please help. when using the filter "pad" with color black and saving it as a jpeg, it isn't pure black... but when saving it as a png (and then converting it to a jpeg) it is pure black. how can i get it to use pure black without having to reconvert it?
[01:14:29 CEST] <treeinterest> anybody here?
[01:14:43 CEST] <treeinterest> if i claim im female, will that encourage answers?
[01:44:21 CEST] <treeinterest> ayyy, fixed it. apparently i had to rewrite history and enter into a state of unawareness to find the true meaning of formats
[01:46:22 CEST] <pzich> that's was I was going to suggest
[01:46:26 CEST] <pzich> what*
[01:49:28 CEST] <treeinterest> to be honest, i dont really know -why- it's fixed, but im just gunna be grateful that it works :)
[01:50:01 CEST] <treeinterest> *greatful
[02:29:25 CEST] <chrisjunkie> Weird issue with RTSP/RTP multicast: Scenario is using FFMPEG to stream some Pelco CCTV cameras. We're trying to use multicast with the cameras instead of unicast because 1) The network is overloaded, and out of our control and we're getting lots of dropped packets and 2) Multicast is already being used by the Pelco recording software so we might as well. When we start streaming more than one camera on multicast however, each FFMPEG instance seems to
[02:29:26 CEST] <chrisjunkie> capture the packets of every other ffmpeg instance. It seems like each ffmpeg instance isn't filtering the UDP packets by source address. I've tried UDP options but seeing as this is rtsp then it doesn't work
[02:29:28 CEST] <chrisjunkie> Any ideas?
[02:30:28 CEST] <chrisjunkie> Have tried this as the last command and I still can't get anything to work as expected
[02:30:29 CEST] <chrisjunkie> ffmpeg -rtsp_transport udp_multicast -rtsp_flags +filter_src -i "rtsp://192.168.77.101/stream2m?sources=192.168.77.101" -an -b:v 400k w1.mp4
[02:30:58 CEST] <chrisjunkie> Looking through rtsp.c, rtspdec.c it really doesn't appear that any filtering happens on multicast UDP RTP packets
[03:13:00 CEST] <Hello71> ll
[09:22:50 CEST] <elmargol> Do i Use 23.976fps 23.98fps or 23.97 fps for digital film?
[09:23:05 CEST] <elmargol> digital recordings not actual film
[12:06:18 CEST] <JodaZ> anyone got an efficient way to extract frames for a preview?
[13:28:46 CEST] <animax> hi all. I've discovered the utvideo codec for intermediate format. JEEBsv was so kind to point to. I've encoded a sequence of PNGs now. by encoding them back to PNGs with ffmpeg just to proof that there are no changes to the original footage it came out that the single PNGs had more data. data of the resulting images results in almost 150% of the orignal sum of data in total. furthermore the...
[13:28:48 CEST] <animax> ...colours of the resulting images seem to have flattened slightly.
[13:28:59 CEST] <animax> encoded with http://www.pasteall.org/58733 (line 1) and encoded back with line 2. working on win 7 ult 64bit.
[13:29:12 CEST] <animax> is there a way to achieve a result that's identical to the original data and cover the same amount of data?
[13:29:18 CEST] <JEEBsv> > -pix_fmt yuv420p
[13:29:31 CEST] <JEEBsv> it's pretty obvious you're going through a RGB<->YCbCr conversion
[13:29:46 CEST] <JEEBsv> remove that and it should try to keep the original colorspace as far as I can remember how lavc works :P
[13:30:38 CEST] <JEEBsv> also 420 means 4:2:0 chroma subsampling so red/blue is saved in one sample per 2x2 luminance samples
[13:30:59 CEST] <JEEBsv> + it converts to limited range YCbCr
[13:31:24 CEST] <JEEBsv> so yeah, if you want the same then you stay the fuck away from colorspace conversions
[13:33:13 CEST] <eric___> Good afternoon
[13:33:51 CEST] <eric___> I have a macbook pro, and I use music player daemon with cantata client to play my music
[13:33:56 CEST] <animax> JEEBsv: ok. I'll try out.
[13:34:18 CEST] <eric___> All my music library is AAC (.m4a)
[13:34:37 CEST] <eric___> mpd is compiled in a way so ffmpeg decodes AAC
[13:34:57 CEST] <eric___> But I have a problem, there is no gapless playback
[13:35:18 CEST] <eric___> There is a few milliseconds of blank between songs.
[13:35:42 CEST] <eric___> I'd like to know is ffmpeg can decode .m4a gapless ?
[13:39:50 CEST] <eric___> Can anyone give me an answer please ?
[13:40:54 CEST] <animax> JEEBsv: the data is even bigger now (200%). ?
[13:41:52 CEST] <animax> I left out the pix format.
[13:42:11 CEST] <animax> ffmpeg -i Test_%04d.png -c:v utvideo -pred median out_05.avi
[13:42:40 CEST] <animax> and encoded back with line 2 in pasted code
[13:44:32 CEST] <eric___> I encode my music with dBpoweramp for mac
[13:44:40 CEST] <JEEBsv> animax: in what fucking sense "data is even bigger now"?
[13:45:51 CEST] <JEEBsv> do you just mean that the PNGs created by lavc are less compressed? quite fucking possible
[13:45:51 CEST] <eric___> please any guru tell me...
[13:46:12 CEST] <animax> the same procedure I've discribed above. sum of data of all PNGs is 15,6 MB now. previous size was 7,86 MB.
[13:46:28 CEST] <animax> when encoding back tp png ...
[13:46:31 CEST] <JEEBsv> well yes, because I have no idea how the fuck ffmpeg codes PNG and if it has any options to compress it better
[13:46:56 CEST] <JEEBsv> if you just want to check if the actual coded data within the ut video frames is the same you can use the md5 output
[13:47:29 CEST] <JEEBsv> ffmpeg -i input.png -f md5 - should output a md5 hash onto the standard output (onto your motherfucking terminal)
[13:47:39 CEST] <JEEBsv> or was it framemd5
[13:47:41 CEST] <JEEBsv> probably was :P
[13:47:53 CEST] <JEEBsv> also you might or might not have to set -c:v rawvideo for the data to get decoded
[13:48:09 CEST] <JEEBsv> or not, that might just leave it not decoded
[13:48:43 CEST] <JEEBsv> in any case, there are fucking ways of making sure the coded data is the same
[13:50:25 CEST] <JEEBsv> https://www.ffmpeg.org/ffmpeg-formats.html#framemd5-1
[13:50:29 CEST] <eric___> Is there anybody out there ? :/
[13:50:44 CEST] <JEEBsv> ok, so you shouldn't have to poke it really
[13:51:28 CEST] <JEEBsv> `ffmpeg -i input.png -f framemd5 -` should do it for the PNG file
[13:51:49 CEST] <JEEBsv> and `ffmpeg -i input.avi -f framemd5 -` should do it for a one-frame AVI file encoded out of a single PNG
[13:51:59 CEST] <JEEBsv> this works with multiple ones as well of course :P
[13:53:31 CEST] <JEEBsv> eric___: lavf can work with encoder delay if it was properly signaled by the encoder in the container. it will output negative timestamps on decoded samples that come before it
[13:53:48 CEST] <JEEBsv> how players implement that and *if* they implement it is completely up to the players
[13:54:00 CEST] <JEEBsv> libavformat and libavcodec do let you do it :P
[13:57:25 CEST] <eric___> JEEBsv: Thank you, I'm sorry but as english isn't my native language I don't understand what you wrote :/
[13:58:27 CEST] <animax> JEEBsv: thank you very much. my purpose still is to retain quality untill the final step. so http://www.pasteall.org/58735 should work fine? I was just irritated because the output avi file is growing and the "back coded" PNGs did so also.
[13:58:43 CEST] <JEEBsv> the AVI of course grew because you started feeding more data
[13:59:08 CEST] <JEEBsv> if you wanted to retain fucking quality look at my framemd5 lines :P
[13:59:17 CEST] <JEEBsv> those should hash the decoded pictures
[13:59:34 CEST] <JEEBsv> thus if you get the same hashes you know you have the same data coded in both
[14:00:13 CEST] <JEEBsv> eric___: what I meant is that libavformat and libavcodec support signaling encoder delay in mp4 just fine
[14:00:43 CEST] <JEEBsv> so if the files are done correctly with the correct signaling libavformat and libavcodec should tell you which samples that come from the decoder should be ignored
[14:01:03 CEST] <animax> JEEBsv: can I md5 check the whole png sequence in one go?
[14:01:10 CEST] <JEEBsv> animax: I would say yes
[14:01:35 CEST] <JEEBsv> use the same -i and then "-f framemd5 -" after that to have the output in your terminal
[14:01:45 CEST] <JEEBsv> if you want it in a file, then you just replace the last dash with the file name
[14:01:59 CEST] <JEEBsv> dash means either standard input for input, or standard output for output
[14:02:11 CEST] <animax> 'ffmpeg -i input%04d.png -f framemd5'?
[14:02:27 CEST] <JEEBsv> that lacks the output, which is why I had the dash there :P
[14:02:35 CEST] <animax> oh mom
[14:02:35 CEST] <JEEBsv> (or a file name if you want the information dumped into a file)
[14:03:24 CEST] <animax> what kind of file? .txt?
[14:03:50 CEST] <JEEBsv> it will be text data but the application doesn't care how you name it
[14:04:16 CEST] <JEEBsv> libavformat only cares about the file name when you don't specify the muxer with -f
[14:04:27 CEST] <animax> 'ffmpeg -i input%04d.png -f framemd5 out.txt'?
[14:04:29 CEST] <JEEBsv> for example if your file name ends with dot-avi it picks the AVI muxer :P
[14:04:31 CEST] <JEEBsv> yes
[14:05:00 CEST] <animax> ?
[14:05:19 CEST] <animax> 'ffmpeg -i input.avi -f framemd5 out.txt'?
[14:05:55 CEST] <JEEBsv> yes
[14:06:31 CEST] <JEEBsv> it will output data in this format: https://www.ffmpeg.org/ffmpeg-formats.html#framemd5-1
[14:07:01 CEST] <eric___> JEEBsv: Thanks. How can I compile ffmpeg (what args) so its plays .m4a gaplessly ?
[14:07:42 CEST] <eric___> it*
[14:07:47 CEST] <JEEBsv> eric___: you can't disable the feature without disabling the whole demuxing/decoding part
[14:08:02 CEST] <animax> so doing 'ffmpeg -i input%04d.png -f framemd5 out.txt' with both the original png sequence and the "back encoded" one shloud deliver me the same md5 sum, shouldn't it?
[14:08:35 CEST] <JEEBsv> I don't know why you'd be going back to PNG just for that, just compare the sources and the ut video output
[14:08:44 CEST] <JEEBsv> if those match, you're good
[14:09:04 CEST] <animax> JEEBsv: yes, you are wright. sorry.
[14:09:30 CEST] <JEEBsv> eric___: so it's either incorrectly created files,  your player not using the timestamps correctly or the files having silence at the end
[14:09:37 CEST] <eric___> JEEBsv: I'd like to enable gapless playback for m4a... Sorry if I don't understand
[14:09:58 CEST] <eric___> JEEBsv: Ok, thank you
[14:10:05 CEST] <JEEBsv> (and of course there's the whole thing after you get and handle the decoding correctly in the audio output)
[14:10:39 CEST] <JEEBsv> I mean, all I'm saying that if you have a new enough libavformat and -codec built with mp4/aac support you should have everything you need to get the encoder delay and use it correctly
[14:12:25 CEST] <eric___> JEEBsv: libavformat and lib(?) are to be compiled ?
[14:12:50 CEST] <JEEBsv> ...
[14:13:43 CEST] <animax> JEEBsv: pardon. is there a chance to check the md5 sum in a easier way?
[14:13:52 CEST] <JEEBsv> "in an easier way"?
[14:14:00 CEST] <animax> do I have to do it visually?
[14:14:10 CEST] <JEEBsv> oh, you can script it of course :P
[14:14:16 CEST] <animax> or is there an app?
[14:14:30 CEST] <JEEBsv> scrub from the lines everything but the md5 hash, then compare
[14:14:38 CEST] <animax> ok.
[14:15:02 CEST] <eric___> JEEBsv: thank you for your help. Gonna investigate...
[14:16:13 CEST] <animax> JEEBsv:ok. no chance ...
[14:16:28 CEST] <animax> can't figure out what is what
[14:16:35 CEST] <animax> too much
[14:16:40 CEST] <animax> code ...
[14:16:58 CEST] <JEEBsv> are you by chance opening it in notepad or something that expects windows endlines?
[14:17:11 CEST] <animax> note pad
[14:17:14 CEST] <JEEBsv> yes
[14:17:35 CEST] <JEEBsv> use a proper text editor or output to the terminal which does understand plain \n as an endline
[14:17:59 CEST] <JEEBsv> notepad++ or sublime text 3 are the two I generally keep installed around on windows :P
[14:18:02 CEST] <animax> http://www.pasteall.org/58737
[14:18:18 CEST] <animax> I have notepad++ installed
[14:18:26 CEST] <JEEBsv> yes, that looks simple enough
[14:19:28 CEST] <JEEBsv> anyways, if you just want to make generally sure it's lossless you can just compare the three first manually or so :P if you ever feel like always checking it, you should just script it then
[14:19:50 CEST] <JEEBsv> python or powershell or something should handle it nicely
[14:24:16 CEST] <animax> JEEBsv: you mean hash numbered by 0, 1 und2? line 7, 8 and 9?
[14:25:10 CEST] <JEEBsv> the last part of all of the data lines is the md5 hash, the first number is the stream number which can be different if you have multiple streams in the input
[14:25:49 CEST] <JEEBsv> so basically to make sure you have coded the same data just check the first three or so pictures' md5sum that it matches between your input PNGs and the Ut Video file >P
[14:25:52 CEST] <JEEBsv> :P
[14:26:06 CEST] <JEEBsv> that should mean that you are generally doing the right thing
[14:28:49 CEST] <animax> ok. I start to realize ... can't figure out any difference at all in the whole list of 75 entries.
[14:29:36 CEST] <animax> omg. this is all quite complicated ... :-)
[14:31:36 CEST] <JEEBsv> I guess if you had coded your source PNGs in >8bit per sample you would have gotten different hashes, but then the sizes of your packets would also be different
[14:32:01 CEST] <JEEBsv> (the value before the hash is the size of the data to be hashed)
[14:33:29 CEST] <animax> I coded them in 16 bit afaicr
[14:35:32 CEST] <JEEBsv> then the hashes wouldn't match
[14:35:52 CEST] <JEEBsv> when you run ffmpeg what does it say the input colorspace is for the PNGs?
[14:40:00 CEST] <animax> rgb (pc)
[14:40:14 CEST] <animax> npo
[14:40:22 CEST] <t4nk827> So... i'm having trouble getting the filter ebur128 to add the metering metadata to my audio stream.  My CLI snippet is here http://pastebin.com/embed_iframe.php?i=Qx4MDYDk
[14:40:23 CEST] <animax> rgba (pc)
[14:41:00 CEST] Last message repeated 1 time(s).
[14:43:58 CEST] <JEEBsv> well yeah, both of those are 8bit RGB :P
[14:44:05 CEST] <JEEBsv> (per sample)
[14:44:33 CEST] <animax> JEEBsv: default setting in blender is 16 bit with rgba.
[14:44:55 CEST] <JEEBsv> well it most definitely isn't 16bit according to ffmpeg's decoder if that's the input part :P
[14:44:58 CEST] <animax> ok. maybe I've changed the settings.
[14:46:32 CEST] <JEEBsv> well, how people call those things can differ; people often call 8bit samples with three samples per value as 24bit after all
[14:46:41 CEST] <JEEBsv> and four 8bit samples (RGBA) is often called 32bit
[14:46:42 CEST] <JEEBsv> :P
[14:47:38 CEST] <JEEBsv> anyways, it's not possible to code >8bit samples with Ut Video so not too much use of coding the PNGs with something more than that :P
[14:47:56 CEST] <JEEBsv> or well, there is if you want to keep such renders
[14:48:01 CEST] <t4nk827> Fuller explanation of what I've tried here: http://pastebin.com/sffhn0n3
[14:50:53 CEST] <gr1sha> Can I convert a VideoPicture object to RGB?
[14:50:58 CEST] <gr1sha> using sws_scale?
[14:53:14 CEST] <animax> JEEBsv: ? there is or there isn't?
[14:55:25 CEST] <JEEBsv> there is use if you need to keep around such renders, but you are not going to be able to input that stuff into most editors unless something supports higher bit depth PNG packed in MOV or so :P
[15:05:37 CEST] <animax> JEEBsv: You mean there are formats I can use for storing the footage but they cannot be used with editors?
[15:06:44 CEST] <JEEBsv> of course, ever since you said "this is for editing in video editing software" your alternatives are limited
[15:07:17 CEST] <JEEBsv> blender is possibly a bit less limited because it can use lavf/lavc, but anything payware is going to end up being quite limited in what it can eat and so forth
[15:12:09 CEST] <animax> after all I found out that there is no way to save the imported utvideo after editing in blender with the input utvideo settings. I have to see if all effects and filters will work when dealing right with the PNGs (import and export). so is there a good command to use lossless colour space preserving libx264?
[15:13:06 CEST] <animax> because libx264 is good enough for blender isn't it?
[15:13:35 CEST] <JEEBsv> if the lavc it uses is built with avc support, yes
[15:13:59 CEST] <animax> how can I check?
[15:14:18 CEST] <JEEBsv> "-c:libx264rgb -q:v 0 -g 3" are some settings to test
[15:14:22 CEST] <JEEBsv> uhh
[15:14:25 CEST] <JEEBsv> -c:v libx264rgb
[15:15:11 CEST] <animax> I shall encode and check if blender reads the file?
[15:15:16 CEST] <JEEBsv> yes
[15:15:28 CEST] <JEEBsv> -g 3 should be small enough that it should be quite quick to seek
[15:17:19 CEST] <animax> 'ffmpeg -i img_%04d.png -c:v libx264rgb -profile:v high -preset medium -crf 23 out.mp4'?
[15:17:24 CEST] <JEEBsv> ...
[15:17:38 CEST] <JEEBsv> you truly have no fucking idea what the fuck you are doing
[15:17:45 CEST] <JEEBsv> you wanted lossless, right?
[15:17:52 CEST] <animax> Only 'ffmpeg -i img_%04d.png -c:v libx264rgb out.mp4'
[15:18:13 CEST] <JEEBsv> wlel yes, you need to set the fucking quantizer to zero to get fucking lossless
[15:18:20 CEST] <animax> that might be right ... :-(
[15:18:31 CEST] <JEEBsv> which is what -q:v 0 I think should do, if it doesn't work, use -qp 0
[15:18:48 CEST] <JEEBsv> also you definitely don't want the default fucking GOP length for editing which requires quick seeking
[15:19:07 CEST] <JEEBsv> around -g 3 should bring some efficiency into coding but still be fast
[15:19:27 CEST] <JEEBsv> I fucking spoonfeed you with the parameters and you just go off into a fucking tangent
[15:19:30 CEST] <JEEBsv> fuck you
[15:21:00 CEST] <animax> the whole topic is new for me and english is not my mother language ...
[15:24:08 CEST] <animax> so I really have to struggle with understanding everything you say ...
[15:24:17 CEST] <animax> but I'm trying
[15:25:18 CEST] <animax> what I considered last was shit I see.
[15:31:40 CEST] <animax> ok. you were really quite a big help for me. one last trial. maybe you will answer me. http://www.pasteall.org/58743. is it ok?
[15:33:16 CEST] <animax> if you don't answer anymore: a big thx and bye.
[17:29:24 CEST] <k_sze> So YouTube has these recommendations for videos to be uploaded, right? https://support.google.com/youtube/answer/1722171?hl=en-GB
[17:30:54 CEST] <k_sze> What ffmpeg options do those translate to? The ones for the AAC-LC are easy, but the 2 consecutive B frames, closed GOP, moov atom at the front, and CABAC, I don't know at all.
[17:31:43 CEST] <Mavrik> those are pretty much ffmpeg defaults for x264
[17:31:55 CEST] <Mavrik> -profile:v high
[17:32:03 CEST] <Mavrik> will give you CABAC
[17:33:14 CEST] <k_sze> the 2 consecutive b-frame is not default, if I understand the x264 documentation correctly?
[17:33:18 CEST] <k_sze> the default seems to be 3?
[17:36:22 CEST] <Mavrik> ffmpeg -i <something> -codec:v libx264 -profile:v high -x264opts "b-frames=0" -g 15 -flags +cgop  -pix_fmt yuv420p -crf <quality> -movflags faststart output.mp4
[17:36:24 CEST] <Mavrik> that should be it
[17:36:46 CEST] <Mavrik> note that not all of those are requirements really - youtube will take Main profile video with no problem
[17:36:48 CEST] <k_sze> b-frames=0??
[17:36:49 CEST] <Mavrik> oh, also, b-frames=2 "P
[17:36:52 CEST] <Mavrik> :P
[17:36:53 CEST] <k_sze> heh
[17:37:09 CEST] <Mavrik> also bitrates are mostly guidelines, not hard requirements
[17:37:19 CEST] <JEEBsv> Mavrik: setting profile to high generally does nothing
[17:37:31 CEST] <JEEBsv> as it's where x264 starts, it then checks your settings
[17:37:45 CEST] <JEEBsv> and if you use nothing that requires high profile it will automagically lower the profile
[17:37:48 CEST] <Mavrik> yeah
[17:37:55 CEST] <Mavrik> so it's a bit redundant
[17:38:06 CEST] <Mavrik> as long as it's not baseline that kills quality it's fine
[17:38:36 CEST] <Mavrik> and "-g" is obviously GOP size
[17:38:40 CEST] <k_sze> Why the -g 15 though?
[17:39:00 CEST] <k_sze> oh, GOP of half the frame rate
[17:39:01 CEST] <k_sze> right
[17:39:01 CEST] <JEEBsv> I think it's actually pretty useless to tell people to set profile to high, as it is the highest and most people will just see "ffmpeg" not respecting that parameter
[17:39:19 CEST] <JEEBsv> which leads to a lot of questions like "I set profile to high but I get main"
[17:39:20 CEST] <JEEBsv> :P
[17:39:59 CEST] <JEEBsv> (mostly because by default 8x8 dct is disabled, which IIRC is the only feature that x264 supports from high profile only)
[17:40:20 CEST] <Mavrik> ^^
[17:40:24 CEST] <JEEBsv> only if you use a slower preset that enables 8x8 dct, you will get high profile selected (and you don't even have to set it)
[17:40:47 CEST] <JEEBsv> tl;dr you should only set profile when you want to _limit_ it
[17:41:00 CEST] <JEEBsv> as in, if you specifically need main or baseline
[17:41:27 CEST] <Mavrik> I'll keep that in mind to keep newbies from attacking this channel ;)
[17:41:42 CEST] <JEEBsv> in that case it should actually disable features that are not compatible with that limited profile
[17:44:22 CEST] <k_sze> hmm, -crf <quality> and -b:v 8192k are conflicting, right?
[17:45:04 CEST] <JEEBsv> it's not quality, I got yelled at calling it like that years ago on #x264
[17:45:21 CEST] <JEEBsv> it's a rate factor, which is probably the closest thing we currently have to a constant quality'ish rate control
[17:45:26 CEST] <JEEBsv> but yes, they conflict
[17:53:52 CEST] <k_sze> and if I do a 2-pass encoding, do the libx264 options in pass 1 need to be the same as in pass 2?
[17:57:13 CEST] <k_sze> (I actually don't quite understand why YouTube has recommendations at all.)
[17:57:21 CEST] <k_sze> Aren't they going to re-encode anyway?
[18:01:05 CEST] <JEEBsv> yes
[18:01:29 CEST] <JEEBsv> so in theory if they have fixed their shit you just encode lossless shit that they can decode and push it there
[18:02:04 CEST] <JEEBsv> they used to have issues with lossless H.264 a few years back because their FFmpeg was from like 2007 or so
[18:02:14 CEST] <JEEBsv> but that might have changed since
[18:03:00 CEST] <JEEBsv> the GOP stuff might or might not have to do with how they encode the clips in parts, if they only cut the jobs on IRAPs then the less of them you have the less jobs there will be for their workers
[18:14:15 CEST] <Beam> hey
[18:15:19 CEST] <Mavrik> JEEBsv, yeah, and since they encode for HLS as well, it probably helps them more optimally generate chunks as they reencode in pieces
[18:16:00 CEST] <Beam> how would i go about to make ffmpeg create a live stream from a series of video files chosen live, in a container/format that allows live streaming, as well as upnp/dlna compatible so i can stream it onto the tv?
[18:16:20 CEST] <Beam> aiming to create something like sort of like a video jukebox
[18:16:40 CEST] <Beam> also i would like to add some overlay with live data
[18:16:55 CEST] <JEEBsv> vlc would probably be better for that since it will stay around and can be controlled via telnet etc
[18:17:21 CEST] <JEEBsv> and you IIRC can keep the thing alive while you modify the playlist
[18:17:28 CEST] <JEEBsv> ffmpeg itself probably isn't the best thing for stuff like that
[18:17:49 CEST] <Beam> hrrm i suppose. i hate vlc for streaming though, i always get sync issues
[18:18:55 CEST] <Beam> is there any kind of framework that does what i want, or just a generic framework that uses ffmpeg and has functions to do stuff i mentioned?
[18:19:18 CEST] <JEEBsv> vlc is the closest I know to that :P
[18:19:42 CEST] <JEEBsv> you'd have to write your own thing using the libraries if you want something else
[18:20:34 CEST] <Beam> can vlc add an overlay to a stream, read from something easy to code for?
[18:21:29 CEST] <JEEBsv> it has various overlays as far as I know, but you'd really need to look at the sout documentation for that
[18:48:43 CEST] <JodaZ> Mavrik, any idea how i can encode for HLS in pieces?
[20:58:39 CEST] <rehunted> why trickle does not work with ffmpeg?
[20:58:58 CEST] <rehunted> i'm trying to limit download of a mjpeg stream
[20:59:07 CEST] <rehunted> with http://monkey.org/~marius/pages/?page=trickle
[21:04:58 CEST] <rehunted> I guess I found it
[21:04:59 CEST] <rehunted> A fully statically linked executable (meaning that everything, including libc, is statically compiled into the binary) can't be used with a userspace process like trickle. What eventually happens with a statically linked executable is that it makes system calls directly into the Linux kernel's stable user<->kernel Application Binary Interface (ABI). It does not attempt to load any libraries within /lib, /usr/lib, /usr/local/lib, etc. to resolve symbol
[21:04:59 CEST] <rehunted> s.
[21:05:02 CEST] <rehunted> http://superuser.com/questions/701527/trickle-throttle-all-programs-works-for-statically-linked-executables
[22:38:40 CEST] <ealdeguer> Hi everyone, I have some issues compiling ffmpeg on my mac (yosemite), I did git clone the libfdk-aac and then configure, make make install, but each time I do ./configure my ffmpeg with the --enable-libfdk-aac parameter, it says it can't find it
[22:38:42 CEST] <ealdeguer> any ideas ?
[22:38:45 CEST] <ealdeguer> thx
[22:39:18 CEST] <JEEBsv> `pkg-config --libs fdk-aac` says something?
[22:39:50 CEST] <ealdeguer> yep -L/usr/local/lib -lfdk-aac
[22:40:15 CEST] <JEEBsv> ok, then blame whomever decided that pkg-config shouldn't be used to find fdk-aac
[22:40:38 CEST] <JEEBsv> because you have to manually tell the configure script now where fdk-aac is
[22:40:43 CEST] <JEEBsv> even though pkg-config can tell
[22:41:37 CEST] <JEEBsv> uhh if only I remembered what shell OS X has by default
[22:41:41 CEST] <ealdeguer> like adding some kind of GLIB_LIBS="-lglib-2.0 -lgio-2.0" before ./configure
[22:41:45 CEST] <ealdeguer> sorry for me being a noob haha
[22:42:07 CEST] <JEEBsv> basically you have to put into --extra-cflags="<here>" what pkg-config --cflags fdk-aac gives
[22:42:12 CEST] <Mavrik> ealdeguer, btw, just using "brew install ffmpeg --with-fdk-aac" should get you there way faster
[22:42:22 CEST] <JEEBsv> and --extra-ldflags="<here>" what --libs says
[22:42:41 CEST] <ealdeguer> Hi mavrik, that was my first try with the same result unfortunately :(
[22:42:50 CEST] <ealdeguer> thanks JEEBsv will try
[22:42:53 CEST] <JEEBsv> and if you have time, herp a derp on the ffmpeg-devel list why pkg-config isn't used for fdk-aac
[22:43:05 CEST] <JEEBsv> because it's motherfucking retarded and there's one guy who's opposing it as far as I know
[22:43:16 CEST] <Mavrik> huh
[22:43:22 CEST] <Mavrik> what's the reason for not using pkg-config?
[22:43:29 CEST] <Mavrik> Isn't it like used for everything else?
[22:43:31 CEST] <JEEBsv> yes
[22:43:36 CEST] <JEEBsv> cehoyos is cehoyos
[22:43:57 CEST] <Mavrik> ealdeguer, hrmpf, works just fine for me -_-
[22:44:07 CEST] <Mavrik> but yeah, use --extraldflags/--extra-cflags
[22:44:50 CEST] <JEEBsv> https://patches.libav.org/patch/57514/
[22:45:03 CEST] <JEEBsv> there seems to be a patch on the libav side, but it seems it wasn't merged there, either?
[22:45:29 CEST] <JEEBsv> oh, that's actually pretty new
[22:45:40 CEST] <Mavrik> Let's just start a thread about moving to CMake
[22:45:44 CEST] <ealdeguer> Haha finally used your idea Mavrik, I read too fast, I thought you meant installing just libfdk-aac with brew (what doesn't work) then I did brew install ffmpeg --with libfdkaac
[22:45:57 CEST] <JEEBsv> Mavrik: no never
[22:46:06 CEST] <JEEBsv> cmake has a weird thing about hating pkg-config
[22:46:07 CEST] <Mavrik> Wouldn't that be fun? :)
[22:46:07 CEST] <ealdeguer> and then recompile my ffmpeg with my custom parameters, but thanks to brew the libfdkaac should be now recogniezed
[22:46:17 CEST] <JEEBsv> and instead they reimplement various searching shits
[22:46:39 CEST] <Mavrik> JEEBsv, hmm, I think I have a C++ CMake project here that uses pkg-config to resolve ffmpeg libs :)
[22:46:46 CEST] <Mavrik> but never did invest alot of time in it
[22:46:55 CEST] <Mavrik> but that was more of a bikeshedding post
[22:46:55 CEST] <JEEBsv> I bet that pkg-config thing wasn't done by the cmake guys :P
[22:46:57 CEST] <chrisjunkie> Guys, any way that we can change error concealment on h264 video to only pass the last full-frame to the encoder? I'm viewing CCTV video over a lossy network, and dont care if the picture is jumpy. I.e dont do any error concealment, just wait until we can decode a new full frame by duplicating the old one to match output FPS?
[22:47:17 CEST] <Mavrik> FIND_PACKAGE(PkgConfig) :)
[22:47:56 CEST] <JEEBsv> anyways, I poked #libav-devel about that patch
[22:48:05 CEST] <JEEBsv> if it gets merged, it will hit ffmpeg as well and cehoyos will throw a fit
[22:48:16 CEST] <JEEBsv> and we get to eat pop corn and enjoy the show
[22:50:10 CEST] <ealdeguer> Damn it doesn't work, I'll try now the extra flag idea
[22:53:43 CEST] <ealdeguer> JEEBsv So I did type ./configure --enable-gpl --enable-nonfree [...] --enable-libx264 --enable-libsdk-aac --extra-cflags "-I/usr/local/include"
[22:54:02 CEST] <JEEBsv> ealdeguer: and the ldflags too
[22:54:13 CEST] <ealdeguer> it says Unknown option "-I/usr/local/include"
[22:54:30 CEST] <JEEBsv> --extra-cflags="-I/usr/local/include" ?
[22:55:53 CEST] <ealdeguer> -I/usr/local/include is what returns pkg-config --cflags fdk-aac
[22:57:21 CEST] <JEEBsv> yes, and that's correct
[22:57:32 CEST] <JEEBsv> I mean, just put it that way instead of the space there
[23:00:22 CEST] <ealdeguer> the flags now works but not the ld one
[23:01:11 CEST] <ealdeguer> this is at the end of my configure : --extra-cflags="-I/usr/local/include" --extraldflags="-I/usr/local/include"
[23:01:21 CEST] <JEEBsv> you should have believed my line :P
[23:01:27 CEST] <ealdeguer> is there a dash ?
[23:01:39 CEST] <JEEBsv> yup
[23:01:50 CEST] <ealdeguer> ok sorry for these stupid mistakes
[23:02:36 CEST] <ealdeguer> both the flags are now recognized but still ERROR: libfdk_aac not found
[23:03:04 CEST] <JEEBsv> pastebin config.log
[23:03:05 CEST] <JEEBsv> and link here
[23:05:35 CEST] <ealdeguer> there you go http://pastebin.com/YLxeKGxA
[23:05:38 CEST] <ealdeguer> thx
[23:06:14 CEST] <ealdeguer> I do have in /usr/local/include, a fdk-aac folder, as I did for x264.
[23:06:38 CEST] <ealdeguer> In that fdk-aac folder I did ./configure make make install with no apparent error
[23:07:53 CEST] <JEEBsv> uhh
[23:08:18 CEST] <JEEBsv> so you say you had fdk-aac's source code in /usr/local/include/fdk-aac and you built it there?
[23:08:28 CEST] <ealdeguer> yep
[23:08:31 CEST] <ealdeguer> precisely
[23:08:32 CEST] <JEEBsv> what the flying fuck
[23:09:09 CEST] <JEEBsv> just a protip for the future
[23:09:18 CEST] <JEEBsv> you don't put your initial source code there
[23:09:31 CEST] <ealdeguer> thanks for your advice
[23:09:41 CEST] <ealdeguer> where should I git clone it ?
[23:10:09 CEST] <ealdeguer> in usr/local/include/ffmpeg ?
[23:10:15 CEST] <JEEBsv> somewhere completely unrelated, your home dir or whatever :P the source dir doesn't exactly get used for anything else than making it simple for you to build it
[23:10:23 CEST] <Mavrik> ew
[23:10:28 CEST] <Mavrik> no wonder that also confused homebrew :)
[23:10:29 CEST] <ealdeguer> ok got it
[23:10:34 CEST] <ealdeguer> no more sources in usr local include
[23:10:48 CEST] <JEEBsv> you set a prefix when configuring a la --prefix=/usr/local/ and then make install will install its stuff there
[23:11:08 CEST] <JEEBsv> but still in theory you should have the installed header there, although your directory structure seems to be a mindfuck
[23:11:25 CEST] <ealdeguer> ok I'll do that now with x264 fdk-aac and ffmpeg thanks I try this now
[23:11:40 CEST] <JEEBsv> as you can see at the end of that config.log, it checks for fdk-aac/aacnenc_lib.h , and it does have the -I there
[23:11:46 CEST] <JEEBsv> and it says it can't find it
[23:11:58 CEST] <JEEBsv> but yeah, clean up your /usr/local/ once :P
[23:12:14 CEST] <JEEBsv> although I have no idea what other stuff you have there
[23:14:00 CEST] <ealdeguer> do you think I can remove the include folder with no risks ? as I didn't compile anything apart x264 libfdk-aac and ffmpeg
[23:15:09 CEST] <JEEBsv> I don't remember how homebrew exactly works, but if you check that you have nothing extra in the include and lib and bin directories under /usr/local you can probably clean 'em up :P
[23:16:01 CEST] <ealdeguer> there the content of my usr/local/include folder http://pastebin.com/JdV5QA7i
[23:16:53 CEST] <JEEBsv> I'll just say that I have no idea how you've set your system like so I can't exactly help you there :P
[23:17:40 CEST] <ealdeguer> ok thx anyway
[00:00:00 CEST] --- Sun May 31 2015


More information about the Ffmpeg-devel-irc mailing list