[Ffmpeg-devel-irc] ffmpeg.log.20160311

[burek]

[04:28:42 CET] <jookiyaya> didn't workcan you people tell the difference between x264 video and apple h264 encoded video
[05:03:33 CET] <tacosforall> Hi... I'm trying to render an animation but I'm getting inconsistent keyframing in x264. Each frame should appear four times per 30 frames but I'm seeing inconsistent results. I realize this is a well-documented scenario but I've been coding video and audio software for 20 years (really!) and I'm seeing weirdness no matter what I try with keyframes, time bases, and various filter tricks. Can someone give me some top-level guidel
[05:05:40 CET] <tacosforall> Specific to what I'm doing, I'm using a concat filter (-f concat -i input.txt) with images and duration lines. The file length is exactly what it should be. However the durations jitter.
[05:06:12 CET] <tacosforall> [You really should be able to just specify frames instead of durations, however apparently this is a 10 year open issue that will not get resolved.] So here I am.
[05:13:25 CET] <tacosforall> I'm staring at ffprobe output that's clearly a bug. I'll file it. Thank you for the silence, it was soothing.
[05:42:04 CET] <fling> Which container to use for opus only stream?
[05:47:47 CET] <c_14> ogg?
[05:47:56 CET] <c_14> Or any container that supports opus?
[07:10:38 CET] <limbo_> what's the best way to re-encode an mp3 without losing any quality when mediainfo doesn't show any duration for it, and itunes shows the incorrect duration? I've already done ffmpeg -i <in> -codec copy <out>
[07:19:54 CET] <kepstin> limbo_: you probably want to use the standalone "vbrfix" tool
[07:20:08 CET] <kepstin> it'll analyze the mp3 file and add the missing vbr info header
[07:25:34 CET] <limbo_> ok. where can I find it and how do I use it? Before I ran it through ffmpeg, mediainfo told my it was constant bitrate.
[07:26:11 CET] <kepstin> if the file doesn't have the vbr header, most programs just read the bitrate of the first frame and say the file's cbr with that bitrate
[07:26:25 CET] <kepstin> and then they'll often get wrong or missing length estimates
[07:26:46 CET] <kepstin> cbr mp3 files don't store the length anywhere - you guess it by dividing file size by bitrate
[07:27:08 CET] <kepstin> vbr files need a special header with extra info to determine the length (unless you analyze the entire file)
[07:27:44 CET] <kepstin> but yeah, just google "vbrfix". there's a window binary available, and it's packaged in at least some linux distributions.
[07:30:45 CET] <limbo_> mediainfo tells me that the outputted files are variable bitrate.
[07:37:06 CET] <limbo_> I'm running it, an getting this output (with errors) https://paste.debian.net/414108/ and lines in other runs like "Warning - more than one id3v1 tag found using last found"
[07:50:50 CET] <kepstin> fun. your file is actually multiple mpeg audio files (apparently not even the same mpeg version) which have been concatenated without so much as removing the tags
[07:50:58 CET] <kepstin> no wonder tools aren't working properly with it
[07:51:37 CET] <kepstin> hmm, could be wrong, but either way it's an oddly broken file.
[07:57:17 CET] <limbo_> damn.
[07:57:24 CET] <limbo_> well, that's 2001 for you.
[07:57:49 CET] <limbo_> I have a couple of files, an I guess some of them are fine then.
[07:58:44 CET] <limbo_> any suggestions on how best to re-encode these so I can at least play these on an ipod without them failing to seek properly or jumping aroung the place every twenty or so minutes?
[08:05:30 CET] <limbo_> They're meant to be about 30kbps, so I don't know what to even re-encode them to.
[08:10:16 CET] <kgkg> Hiall! I am trying to forward an IP Camera to youtube but get a "Failed to update header" error. CLI and error: http://pastebin.com/KJX9mZFV
[09:22:40 CET] <wmp> hello
[09:23:11 CET] <wmp> i have capture dark(night) video from security camera
[09:23:22 CET] <wmp> and i want to remove duplicated frames
[09:23:31 CET] <wmp> all works on videos with light
[09:24:19 CET] <wmp> but when is dark ffmpeg not remove any frames, becouse is differences on single pixels
[09:25:22 CET] <wmp> this is my ffmepg options: "-vf "select=gt(scene\,0.00001),setpts=N/(25*TB)" -f segment -segment_time 120 -segment_format mp4 -strftime 1 -reset_timestamps 1"
[09:34:03 CET] <jookiyaya> 1.846153846153846]
[09:39:40 CET] <aRDa_> hi
[09:40:20 CET] <aRDa_> I'm new with ffmpeg I'm recording (from debian jessie) making a video of my screen with the following command: ffmpeg -video_size 1366x768 -f x11grab -r 25 -i :0.0 -qscale 0 video.avi
[09:40:36 CET] <aRDa_> but the sound is not recorded, any ideas why ?
[11:28:25 CET] <cba321> hello is this the dev channel..i got a message from the compilation of ffmpeg to not report a bug but instead contact ffmpeg.org
[11:29:23 CET] <cba321> since the message said "please do not contact Google or someone...i am trying to comply."
[11:53:56 CET] <DHE> cba321: there is a distinct dev channel, but you might be better off making a ticket on the trac page... (assuming one doesn't already exist)
[12:09:29 CET] <cba321> well it looks like there are no developers around at this time...perhaps a few google spies who need to be incarcerated....especially since some of the code is relevant to vp8/vp9 and so on.
[12:10:49 CET] <cba321> nothing is sicker than monstrous 800 pound gorillas trying to steal and capitalize on small people's ideas.
[12:14:07 CET] <cba321> and i say a hearty ...fuck you crooked nigger dick suckin Jew bastards...
[13:23:50 CET] <andrey_utkin> can I ask anybody how to optimize performance of now-stable ffmpeg's aac encoder? http://permalink.gmane.org/gmane.comp.video.ffmpeg.devel/210082
[13:24:09 CET] <andrey_utkin> (link: maillist post by subject "AAC encoder 3x performance drop in 3.0 since Oct 2014")
[13:25:04 CET] <andrey_utkin> oh there was a discussion, gonna read
[13:46:19 CET] <fenris_mid> saluton. i try to compile ffmpeg git-commit 66edd8656b851a0c85ba25ec293cc66192c363ae and get the following error: libavcodec/h264pred_template.c:1276:19: internal compiler error: in vect_transform_stmt, at tree-vect-stmts.c:8176
[13:46:36 CET] <fenris_mid> does anybody have an idea what the problem could be?
[13:46:57 CET] <fenris_mid> is it really just gcc's fault?
[13:46:57 CET] <nevcairiel> which compiler?
[13:47:17 CET] <fenris_mid> g++ (Debian 6-20160109-1) 6.0.0 20160109
[13:47:38 CET] <nevcairiel> so a pre-release compiler then, likely its fault
[13:47:44 CET] <fenris_mid> k
[13:48:07 CET] <nevcairiel> best would be to double check with gcc 5, and report the problem to the gcc developers so its fixed before 6 is released
[13:49:55 CET] <flux> fenris_mid, I bet the developers would be ecstatic if you were able to infer a minimal reproduciton for it ;)
[13:50:23 CET] <flux> probably these vectorization things are quite the bleeding edge stuff anyway in GCC..
[13:51:47 CET] <jkqxz> If you do report it to gcc then you'll probably want to reproduce on something more recent than two months ago first.
[13:56:13 CET] <wmp> I have video from security camera, i want to speed up if isn;t motion on video, how can i do this?
[14:03:41 CET] <andrey_utkin> wmp, cool idea. Maybe a custom video filter could do that. Analyze frame-to-frame difference and increment the timestamps reverse-proportionally to amount of changes
[14:05:56 CET] <BtbN> Or dump it to individual frames, and then compare the difference between the frames, and if it's small enough, discard the later one
[14:06:18 CET] <wmp> but i need to do fast ;)
[14:07:39 CET] <wmp> now i test this: ffmpeg -y -i short.avi -an -vf "select='gt(scene\,0.02)',setpts=0.1*PTS" output_0.02_0.1.avi
[14:11:53 CET] <J_Darnley> That looks like it will create horribly out of order images
[14:12:20 CET] <J_Darnley> oh wait, I misread
[14:12:24 CET] <J_Darnley> ignore that
[14:31:26 CET] <wmp> J_Darnley: unfortunatelly, this dont detect if in room is man, so this dont works...
[14:42:35 CET] <J_Darnley> And if a guy stands still in the middel of the frame there won't be any changes either
[14:43:05 CET] <J_Darnley> It sounds like you want to go to computer vision levels of analysis
[14:43:06 CET] <louiz> You could get a Paranormal Activity effect, automatically.
[14:43:19 CET] <J_Darnley> Somethign that is not available in ffmpeg
[14:44:02 CET] <wmp> :/
[14:48:07 CET] <wmp> J_Darnley: do you know in what software can i do it?
[14:48:33 CET] <J_Darnley> nfi
[14:48:42 CET] <J_Darnley> you will probably have to write your own
[14:48:55 CET] <wmp> :(
[14:53:30 CET] <andrey_utkin> wmp: opencv?
[14:55:11 CET] <andrey_utkin> I can connect you with a guy who has written opencv-powered motion detection on frames provided by ffmpeg. The existing code is proprietary, though, so cannot be shared, but you could contact the guy
[14:55:21 CET] <BAWSTi> Hey guys, is it normal that by using following command I don't get the file size information? 'ffmpeg -i inputFile.mp2'  And if this is the way it is supposed to act, how can I get the file size?
[14:56:51 CET] <wmp> andrey_utkin: now i looking for this: http://superuser.com/a/818412
[14:58:33 CET] <andrey_utkin> cool! thanks for sharing wmp. I wonder what "mp" in filter name stands for. Motion ...?
[14:59:23 CET] <andrey_utkin> BAWSTi, you can get size by "ls -l file" or "stat file"
[15:00:19 CET] <BAWSTi> andrey_utkin: so like ffmpeg -ls -l inputFile.mp2?
[15:00:43 CET] <J_Darnley> No!
[15:00:58 CET] <J_Darnley> Use the basic unix tool called ls
[15:01:37 CET] <J_Darnley> Or are you not on Linux or mac os?
[15:01:41 CET] <BAWSTi> oh, yea well that's gonna be a problem since that would require that much more work xD I'm actually using ffmpeg cli on android ;)
[15:02:04 CET] <J_Darnley> Well then, how would you get the size of any file?
[15:02:28 CET] <andrey_utkin> ffmpeg cli on android - what's that?
[15:02:49 CET] <BAWSTi> That's what I'm asking, honestly I just assumed it would be possible with ffmpeg since it shows so much information about a file, why not the size then? :
[15:03:38 CET] <J_Darnley> Because ffmpeg is far too big for such a simple use
[15:03:56 CET] <J_Darnley> because the filesystem will tell you the size in much less time than ffmpeg
[15:04:09 CET] <BAWSTi> There's an example of what it is @andrey_utkin https://github.com/WritingMinds/ffmpeg-android-java
[15:04:23 CET] <BAWSTi> alright, thanks for the help anyway J_Darnley :)
[15:07:32 CET] <J_Darnley> If you're programming, why the heck would you try and parse the output of some program?
[15:07:42 CET] <J_Darnley> Surely there is a function you should be calling with a filename or some file object
[15:10:10 CET] <andrey_utkin> J_Darnley, well, in the age of using jQuery for arithmetics...
[15:17:10 CET] <wmp> J_Darnley, andrey_utkin, louiz: its possible with ffmpeg :D
[15:17:12 CET] <wmp> ffmpeg -i upsy_krotkie.avi -filter_complex "[0:v] mpdecimate=hi=64*120:lo=64*30:frac=0.30 [a] ; [a] setpts='N/(25*TB)' [b] ; [b] fps=25 [c] ; [c] fps=25 [d]" -map "[d]" -c:v libx264 -f mp4 output_upsy_krotkietest.avi
[15:17:46 CET] <J_Darnley> I'm glad you managed to get something your satisfied with
[15:20:44 CET] <louiz> great
[16:51:38 CET] <limbo_> If I have some mp3 files that are probably encodes at30kbps, what's the easiest/best way to re-encode them so I don't lose quality or increase filesize?
[16:51:52 CET] <limbo_> It's about 7.1 GB of audio.
[16:53:22 CET] <DHE> any re-encoding degrades the audio unless it's a lossless codec (eg: FLAC)
[16:53:31 CET] <DHE> at 30 kilobits it probably sounds pretty bad already
[16:56:53 CET] <zeryx> hey guys I have some questionable output from a splitting process I'm running
[16:56:55 CET] <zeryx> Format mov,mp4,m4a,3gp,3g2,mj2 detected only with low score of 1, misdetection possible![mov,mp4,m4a,3gp,3g2,mj2 moov atom not found/tmp/vd/Hand_on_cereal_field.mp4: Invalid data found when processing input
[16:57:10 CET] <zeryx> I'm recieving this error message and I'm unsure why
[16:57:32 CET] <zeryx> when I run ffmpeg locally it doesn't complain, but when its run on my remote instance with the static ffmpeg binaries it fails; is this due to a missing library?
[16:58:08 CET] <DHE> might be an old version that doesn't detect the file
[16:58:18 CET] <DHE> or do you mean running remotely you're pointing at an HTTP url?
[16:58:31 CET] <zeryx> running remotely as in on a remote server executing the code & returning an exception
[16:58:36 CET] <zeryx> the file is the same
[16:58:44 CET] <zeryx> (I'm handling everything myself in that respect)
[16:58:54 CET] <zeryx> so you think it might be an older version of the binaries?
[16:59:15 CET] <DHE> run it without parameters. there should be a date and version number you can use
[17:02:01 CET] <zeryx> 2.8.5
[17:02:23 CET] <zeryx> https://gist.github.com/zeryx/3dbddc3dd1b6da69ee9e
[17:04:06 CET] <limbo_> DHE: why does it degrade the quality? Is there something about mp3 compression that I'm missing? also, will ffmpeg pick a sane default based on the input if I don't give it a quality setting.
[17:04:10 CET] <limbo_> ?
[17:05:30 CET] <kepstin> limbo_: all lossless codecs have an issue with "generation loss" where each time you re-encode, you lose some quality.
[17:05:35 CET] <kepstin> all lossy codecs*
[17:05:42 CET] <kepstin> lossless ones don't have this issue, obviously
[17:06:24 CET] <kepstin> limbo_: if you don't specify any quality setting, ffmpeg will use an arbitrary default that might not match what you want. You should almost always specify something.
[17:08:42 CET] <limbo_> kepstin: I don't understand what causes that loss though.
[17:09:13 CET] <kepstin> a lossy codec, when encoding, throws away some of the data to make the file smaller
[17:09:30 CET] <kepstin> if you decode then reencode, it will throw away more data
[17:15:11 CET] <zeryx> DHE I think it was actually just something output to the error stream
[17:15:14 CET] <zeryx> that wasn't actually an error
[17:15:34 CET] <zeryx> kinda dumb IMO
[17:16:42 CET] <zeryx> nah still errors out remotely
[17:16:44 CET] <zeryx> odd, works fine locally
[17:16:53 CET] <zeryx> [mov,mp4,m4a,3gp,3g2,mj2 @ 0x4f87100] moov atom not found/tmp/vd/Hand_on_cereal_field.mp4: Invalid data found when processing input
[17:17:47 CET] <kepstin> zeryx: compare your files to make sure they're actually the same? (e.g. sha1sum them both)
[17:18:02 CET] <kepstin> moov atom not found is common if you have a non-streamable mp4 file that's missing the end
[17:23:37 CET] <zeryx> gonna check that now, have to do it a bit roundabout since I have no direct access to the file
[17:24:46 CET] <zeryx> hah
[17:24:48 CET] <zeryx> the sum is different
[17:24:50 CET] <zeryx> thats.. odd
[17:37:27 CET] <zeryx> hah kepstin I figured it uot
[17:37:56 CET] <zeryx> the download link I was being provided wasn't actually a file download, wget somehow is able to differentiate a true file from an http page describing a file
[17:38:17 CET] <zeryx> but what I was using in my scala app wasn't able to do that, so it was downloading a borked file with a bunch of garbage in it
[17:38:59 CET] <kepstin> zeryx: the link probably had a redirect, and your scala app was maybe using the contents of the redirect page rather than following the redirect?
[17:44:00 CET] <zeryx> hmm still seems to be borked
[17:44:45 CET] <zeryx> its still different sha1sums
[17:45:36 CET] <zeryx> nah its that file specifically
[17:45:47 CET] <zeryx> I have another file that I put in there myself in the same directory
[17:45:49 CET] <zeryx> works fine
[20:41:28 CET] <pomaranc> hi, does anyone know if there is hardware h264 decoding available for linux+nvidia?
[20:41:50 CET] <c_14> vdpau
[20:43:39 CET] <pomaranc> c_14: yes, but https://trac.ffmpeg.org/wiki/HWAccelIntro says it's deprecated
[20:45:34 CET] <pomaranc> and looks like it cannot be used for transcoding
[20:47:13 CET] <c_14> In that case, I don't know.
[20:47:22 CET] <pomaranc> thanks anyway
[20:48:37 CET] <DHE> vdpau is for playback, not decoding for software
[20:52:48 CET] <DHE> also vdpau isn't deprecated
[20:53:38 CET] <DHE> oh I see what you mean...
[21:01:14 CET] <jkqxz> The standalone vdpau decoder is deprecated.  The vdpau hwaccel decoder is not.  For anything new, you should always use the latter.
[21:07:59 CET] <kepstin> my impression was that you couldn't use vdpau outside of X tho, so it's not useful for "headless" transcoding. is that true?
[21:09:40 CET] <DHE> that is correct. the decoded frame is only good for rendering on-screen and cannot be drawn back into RAM
[21:12:23 CET] <jkqxz> Yes you can - it's not particularly fast to extract it, but it does work.  (You can use "ffmpeg -hwaccel vdpau" for transcoding.)
[21:13:16 CET] <jkqxz> What you can't do is anything useful with the surface itself other than send it to the screen - there is no hardware transcode path.
[21:14:25 CET] <DHE> ah, okay. you have to do excessive pixel copying even to use NVENC
[21:20:53 CET] <J_Darnley> Surprise!  Using dedicated hardware a vast distance from the cpu and main memory is not very useful for general use.
[21:39:03 CET] <lmat> What is the way to upload examples?
[21:39:08 CET] <lmat> like audio files to share on irc?
[21:41:52 CET] <lmat> found it!  http://streams.videolan.org/upload/
[21:41:55 CET] <lmat> it wasn't ffmpeg, but videolan!
[21:42:45 CET] <lmat> doh! It just ate my file :-(
[22:03:15 CET] <somethingelse> Hi
[22:04:20 CET] <somethingelse> I'm trying to build libav* statically via MSVC2015 on windows and link them into a program
[22:04:40 CET] <somethingelse> I followed instructions and got the *.a files built (e.g. libavcodec.a)
[22:06:03 CET] <somethingelse> http://stackoverflow.com/questions/2096519/from-mingw-static-library-a-to-visual-studio-static-library-lib said I might be able to just use the .a's as .lib's
[22:06:48 CET] <somethingelse> However, while MSVC didn't complain about the .a files I added to its list of libraries, it doesn't seem to be able to find any libav* symbols, either
[22:07:03 CET] <somethingelse> Anyone has experience with this and knows what's up?
[22:08:50 CET] <somethingelse> (libav* was built via the MSVC toolchain)
[22:14:04 CET] <somethingelse> Oh wait
[22:16:57 CET] <somethingelse> Of course
[22:17:19 CET] <somethingelse> libav* needs extern "C"
[22:18:27 CET] <somethingelse> (It might be unfair, but many c libraries include 'extern "C"' guraded by a C++ #ifdef just to avoid those headaches)
[22:18:51 CET] <somethingelse> (Also, many c libraries don't expose structs as public apis)
[22:19:16 CET] Action: somethingelse fades out
[22:35:13 CET] <Lucifer333> hello
[22:35:42 CET] <Lucifer333> maybe its wrong channel, but i have ripped some DVD's as files on disk (yes they are playable and YES I DID BUY THEM)
[22:36:12 CET] <Lucifer333> does anyone have a DVD authering tool to edit, coalesce, cut out languages, etc
[22:37:24 CET] <llogan> you can use ffmpeg to do some of what I think you may want to do with your DVDs (but not DVD's)
[22:37:49 CET] <llogan> as long as the resulting files have been decrypted
[22:38:06 CET] <Lucifer333> yes everything is decrypted
[22:38:58 CET] <furq> Lucifer333: http://dvdauthor.sourceforge.net/
[22:39:01 CET] <kepstin> the '.vob' files are just mpeg-ps streams. it's as simple as concatenating all the split 1g files from the title you want, and you can use that as input directly to ffmpeg
[22:39:16 CET] <furq> assuming you actually want to write it back to a playable dvd
[22:39:19 CET] <furq> otherwise you can do everything with ffmpeg
[22:39:35 CET] <furq> aside from cutting out individual pgcs ;_;
[22:39:42 CET] <Lucifer333> no i dont, i want to convert to mp4 to make it playable on my ipad on the go
[22:39:58 CET] <furq> well then you probably shouldn't ask for a dvd authoring tool
[22:40:04 CET] <Lucifer333> I payed for the movie so i want to see it on my other devices
[22:40:20 CET] <Lucifer333> well the dvd files as is are weird, and split up in weird chucks
[22:40:27 CET] <Lucifer333> so they need to be "cleaned up"
[22:41:02 CET] <kepstin> they're split up into 1gb chunks because of file size compatibility issues, i think. but you just have to find the ones that belong to the movie and concatenate them.
[22:41:24 CET] <furq> if it's a tv show, or poorly-authored, then you might need to use a pgc demuxer such as pgcdemux
[22:41:25 CET] <Lucifer333> I read somewhere DVD's are mpeg2 streams
[22:41:50 CET] <Lucifer333> oh, the tool needs to be on linux, I dont run anything else
[22:41:57 CET] <Lucifer333> pgcdemux, lemmi see
[22:42:02 CET] <furq> pgcdemux is windows only afaik
[22:42:13 CET] <Lucifer333> maybe i can run it in wine
[22:42:18 CET] <llogan> see tools/dvdconcat
[22:42:32 CET] <furq> Lucifer333: you might not need to use it
[22:42:42 CET] <llogan> *dvd2concat
[22:42:51 CET] <furq> it's useful for tv shows where the episodes don't map exactly to the vobs
[22:42:54 CET] <kepstin> i wonder if the 'dvdunauthor' tool (included in the dvdauthor package) can do that
[22:43:11 CET] <Lucifer333> wow so many tools, thanks for the pointers,
[22:43:13 CET] <furq> other than that, the only thing ffmpeg won't do is preserve chapters
[22:43:17 CET] <Lucifer333> I need to do some research
[22:43:49 CET] <furq> http://dvdauthor.sourceforge.net/doc/dvdunauthor.html
[22:43:50 CET] <furq> what a great manpage
[22:44:51 CET] <kepstin> well, it has no bugs, that's good to know.
[22:48:03 CET] <Lucifer333> http://dvdauthor.sourceforge.net/doc/index.html
[22:48:05 CET] <Lucifer333> main page
[22:48:21 CET] <furq> those are the dvd author docs
[22:48:27 CET] <furq> i'm interested in dvdunauthor
[22:49:30 CET] <kepstin> dvdunauthor basically takes a dvd image and turns it into a directory structure that could be used as input to dvdauthor
[22:49:48 CET] <Lucifer333> ?
[22:49:48 CET] <kepstin> I forget exactly what it preserves.
[22:49:59 CET] <furq> http://sprunge.us/QVMN
[22:50:04 CET] <furq> someone really doesn't want me to find out what this does
[22:52:00 CET] <Lucifer333> do you know the CentOS repo
[22:52:05 CET] <Lucifer333> so i can install with yum?
[23:07:53 CET] <wirelesslinux> Hi,
[23:08:24 CET] <wirelesslinux> I'm trying to use ffmpeg with H264 nvidia support
[23:09:27 CET] <wirelesslinux> but when running ffmpeg, I block on error : dl_fn->cu_init(0) - failed with error code 0x3e7
[23:10:04 CET] <llogan> ...not that i have any experience with nvenc
[23:10:12 CET] <wirelesslinux> Ok, sorry, first time for me on irc :)
[23:13:14 CET] <wirelesslinux> http://pastebin.com/j5Z3T0yP
[23:15:38 CET] <kepstin> wirelesslinux: that "modprobe: FATAL: Module nvidia-uvm not found." sounds suspicious. do you have the binary nvidia drivers installed and loaded?
[23:16:36 CET] <wirelesslinux> How do I check it ? With lsmod | grep nvidia ?
[23:20:44 CET] <wirelesslinux> http://pastebin.com/19cK09hh
[23:21:20 CET] <wirelesslinux> And aptitude search uvm return an empty list.
[23:28:37 CET] <kepstin> hmm. dvdunauthor does /not/ split up the pgcs by chapter, but it does write timestamps for them into the dvdauthor xml file, at least.
[23:34:32 CET] <furq> yeah every tool i can find for this on *nix has been abandoned since 2004
[23:35:13 CET] <kepstin> i mean, you could probably write a shell script that interprets the ifo file and uses dd to extract a particular chapter or whatever :/
[23:35:44 CET] <kepstin> I can't remember if there was any work done on an input to ffmpeg using libdvdread or libdvdnav
[23:36:20 CET] <furq> yeah it is odd that ffmpeg seemingly has no concept of ifo files
[23:36:23 CET] <Plorkyeran> dvd things in general have been abandoned since 2004
[23:36:23 CET] <jookiyaya>  how i tried 10 different players, only one support  bluray java menu properly:  why is this so difficult
[23:36:53 CET] <Plorkyeran> they got to a working state and then have been bitrotting since
[23:38:29 CET] <furq> i should see if i can hack something together with libdvdread
[23:38:49 CET] <kepstin> jookiyaya: blurays in particular are very difficult
[23:39:00 CET] <kepstin> the dvd vm for navigation is a lot simpler :/
[23:39:18 CET] <jookiyaya> if bluray uses java for menu, what does dvd use
[23:39:22 CET] <kepstin> (dvdunauthor actually disasembles the nav vm instructions, which is kinda neat)
[23:39:29 CET] <kepstin> jookiyaya: custom very simple bytecode
[23:39:44 CET] <kepstin> and some still images and mpeg2 videos
[23:40:11 CET] <c_14> furq: there's the ifo_palette option
[23:40:15 CET] <c_14> or whatever it was called
[23:42:14 CET] <kepstin> c_14: that's used for setting the colours when remuxing dvd image subtitles, but it's kind of a sidechannel hack, and obviously doesn't let you read e.g. video chapters and whatnot :)
[23:43:00 CET] Action: c_14 wasn't really paying attention and his brain came up with the first result that seemed to match
[23:44:31 CET] <limbo_> kepstin: but what about the way mp3 encodes makes it degrade quality? jpg doesn't have this problem.
[23:44:40 CET] <furq> yes it does
[23:44:44 CET] <kepstin> limbo_: jpg has the same problem, yes...
[23:45:00 CET] <limbo_> no, it doesn't.
[23:45:04 CET] <furq> the example on the wikipedia page for generation loss is a jpeg
[23:45:09 CET] <furq> https://en.wikipedia.org/wiki/Generation_loss
[23:45:16 CET] <kepstin> (you can do some modifications to jpeg - rotations, certain crops - without decoding and reencoding, but a full reencode will lose quality)
[23:45:58 CET] <kepstin> i mean, if you're only doing it once or twice, you're probably not gonna notice
[23:46:11 CET] <furq> you must not have a facebook account
[23:46:56 CET] <furq> otherwise you'd see plenty of text screenshots of twitter feeds that have been reencoded by every viral image host on the internet before ending up on your decrepit uncle's wall
[23:49:10 CET] <growery> hi
[23:49:48 CET] <growery> how can i make ffmpeg hardware accelerated on a mac mini with osx el capitain?
[23:51:28 CET] <limbo_> for i in {1..504} ; do ffmpeg -i $i.jpg $((i+1)).jpg; done
[23:52:00 CET] <limbo_> I just did that on an imge to re-ecode it 505 times. I'm not seeing any quality loss after number 2.
[23:52:30 CET] <limbo_> how many times do I have to do it to get quality loss
[23:53:38 CET] <kepstin> limbo_: you might have just gotten lucky; if you re-encode with the exact same settings, it's possible that after a few generations it'll happen to pick the same encoding and give back the same result.
[23:53:42 CET] <jkqxz> When you reencode with exactly the same parameters you may well get exactly the same result - in this case you converged to a fixed point very quickly.  Try alternating JPEG and something else lossy instead.
[23:54:13 CET] <kepstin> alternately, something like cropping a pixel off the top each time will certainly mess it up :)
[23:54:32 CET] <limbo_> I know that, because I know how jpeg compression works.
[23:55:26 CET] <limbo_> can anyone explain why mp3s might have compession that stops them from being re-encoded without change losslessly?
[23:55:26 CET] <kepstin> limbo_: right - so in theory, if you re-encode a file with a the same lossy codec with exactly the same settings as it originally used, and the encoder is deterministic, you *might* get back the identical result.
[23:55:46 CET] <kepstin> use a different encoder? all bets are off.
[23:56:09 CET] <llogan> growery: FFmpeg has "OS X VideoToolbox support". whatever that is. might be worth a look.
[23:56:24 CET] <kepstin> use something like x264 in multithreaded mode? well, then it's not deterministic, so no way :)
[23:56:38 CET] <growery> llogan thanks i will search for it now
[23:56:47 CET] <furq> limbo_: set -q:v
[23:57:13 CET] <c_14> limbo_: because if you take a lossless source and encode it lossy and then decode it again you get something that isn't equivalent to your decoded original lossless source
[23:57:27 CET] <c_14> Therefore if you encode the newly decoded content it will encode to something different
[23:57:58 CET] <limbo_> but what about the codec used in mp3 makes that happen?
[23:58:04 CET] <furq> limbo_: http://imgur.com/a/NVVKn
[23:58:05 CET] <llogan> growery: also, this may be of slight interest to you: https://trac.ffmpeg.org/wiki/HWAccelIntro
[23:58:10 CET] <furq> that's what i get after 500 passes with -q:v 90
[23:59:03 CET] <llogan> why 90? IIRC, max is 31
[23:59:04 CET] <c_14> limbo_: the mp3 encoder throws away data when it encodes
[23:59:40 CET] <furq> because the max isn't documented anywhere i can find and everything else i've ever used for jpeg goes from 1-100
[23:59:47 CET] <c_14> As do all lossy encoders
[23:59:50 CET] <llogan> furq: heh. Karnov.
[00:00:00 CET] --- Sat Mar 12 2016