[Ffmpeg-devel-irc] ffmpeg.log.20151122

[burek]

[00:00:40 CET] <furq> either of those alone before the input works as well
[00:02:45 CET] <furq> http://sprunge.us/RMjZ
[00:05:58 CET] <furq> fwiw it works with mp4 as well, although seeking is broken
[00:07:13 CET] <furq> nvm seeking works fine in vlc
[00:09:52 CET] <pepee> well, if I change -c:a copy to '-c:a libvo_aacenc -ab 32k -ar 16000', I get a different output
[00:10:12 CET] <pepee> *with your command line
[00:10:43 CET] <pepee> is ffmpeg miscalculating the duration when using aac?
[00:11:08 CET] <furq> yeah i just tried with aac and now it's out by five seconds
[00:11:26 CET] <furq> i'm pretty sure it was doing the same thing with mp3 last time though
[00:11:37 CET] <Ripmind> can i somehow tell ffmpeg to convert all mkv files in the folder to mp4?
[00:11:41 CET] <Ripmind> with the same name?
[00:12:19 CET] <Ripmind> seems like ffmpeg -i *.mkv *.mp4 does not work at all :D
[00:12:51 CET] <Worf> for file in *.mkv; do ; ffmppeg -i ${file} ...blahblah ... ${file%.mkv}.mp4 ; done
[00:13:27 CET] <furq> for m in *.mkv; do ffmpeg -i "$m" -c copy "${m%.*}.mp4"; done
[00:14:01 CET] <furq> this is the point at which you tell us that you're on windows
[00:14:13 CET] <Ripmind> i am on windows
[00:14:19 CET] <Ripmind> but it works either way :)
[00:14:34 CET] <Ripmind> cygwin OP i guess
[00:14:40 CET] <furq> fun
[00:14:52 CET] <furq> can you ^C ffmpeg in cygwin
[00:14:58 CET] <furq> for some reason that does nothing in msys2
[00:15:08 CET] <Ripmind> is there any major diff beztween -i "$m" and -i ${file}?
[00:15:18 CET] <Ripmind> yes i can, furq
[00:15:48 CET] <c_14> "$m" is quoted
[00:15:55 CET] <Ripmind> can ffmpeg count the frames and give me back the amount?
[00:15:59 CET] <Ripmind> without encoding anything
[00:15:59 CET] <furq> you probably want it to be quoted
[00:16:00 CET] <Worf> $var is identical to ${var} ... but in some situations like if you want to add some string without a space after you need {}
[00:16:05 CET] <furq> you also definitely want -c copy
[00:16:21 CET] <c_14> Ripmind: ffprobe -count_frames eeeeeeeh something something read the manpage
[00:16:24 CET] <Worf> you can do ${var}blabla, but $varblabla isn't the same
[00:16:56 CET] <pepee> I install busybox when I need to, it's better than other shells :P
[00:17:04 CET] <Ripmind> thanks :)(
[00:17:25 CET] <Ripmind> why do i want -c copy furq?
[00:17:31 CET] <furq> because otherwise it'll reencode
[00:17:33 CET] <c_14> Ripmind: I think it's -show_streams or something
[00:17:57 CET] <Ripmind> i never heared abotu ffprobe tho to be honest
[00:17:58 CET] <Worf> actually you didn't specify if you want to reencode :)
[00:18:16 CET] <furq> i used my magic crystal ball
[00:18:43 CET] <Worf> ah ... good ... mine is broken
[00:18:59 CET] <Ripmind> i don't hav eany yet, where did you get it?
[00:19:49 CET] <furq> you know how l. ron hubbard is dead
[00:19:55 CET] <furq> that's how i got it
[00:22:32 CET] <Ripmind> c_14: btw, i checked soem more and ffmpeg is really using my gpu now, that's so nice :)
[00:26:06 CET] <Ripmind> do i need to calculate the number of frames by fps and duration?
[00:26:21 CET] <c_14> Ripmind: ffprobe -count_frames -show_streams
[00:27:09 CET] <c_14> If the video is vfr fps * duration won't be accurate
[00:27:34 CET] <Ripmind> true
[00:27:43 CET] <Ripmind> but vfr is not very common, is it?
[00:28:00 CET] <Ripmind> -count_frames does not return the number of frames
[00:28:08 CET] <c_14> It's listed as stream metadata
[00:28:23 CET] <c_14> should be anyway iirc
[00:28:48 CET] <Worf> can one recommend a filter for denoising videos recorded by smartphones (in suboptimal lightning)? atadenoise gives the best results of what those i tried, but i defenitely didn't try everything.
[00:28:54 CET] <c_14> Ripmind: and then look at nb_read_frames
[00:29:02 CET] <c_14> You can also tell it only to show that tag if you want
[00:29:47 CET] <c_14> ffprobe -count_frames -select_streams v -show_entries stream=nb_read_frames
[00:30:29 CET] <durandal_1707> Worf: there are better but slower
[00:30:40 CET] <durandal_1707> much slower
[00:31:06 CET] <Worf> durandal_1707: much slower is ok, the stabilizing and reencoding makes it much slower anyway
[00:32:07 CET] <Worf> currently i use vidstabdetect/vidstabtransform to counter my shaky hand, which isn't exactely fast
[00:32:08 CET] <Ripmind> c_14: Seems like that worked, thank you
[00:32:09 CET] <durandal_1707> but they are not present yet in lavfi
[00:35:33 CET] <Worf> durandal_1707: ok, anything you can recommend that is available for puny mortals like me? :)
[00:44:19 CET] <durandal_1707> Worf: nlmeans and BM3D
[00:45:33 CET] <pepee> so, what else can I try to copy the duration of the audio to the final video when converting 1 image + 1 audio file to video?
[00:46:12 CET] <pepee> -t works, but I would need to know the duration beforehand... is there a command to copy the duration from the audio file?
[01:05:34 CET] <Worf> durandal_1707: thanks, just looked them up and they look interesting. will have a closer look.
[01:06:51 CET] <darktoad> ''Anonymous also said the Islamic State group is planning an assault at the WWE Survivor Series event scheduled to take place in the Philips Arena in Atlanta Sunday at 7.30 p.m. EST, as well as attacks at multiple events in Paris.''
[01:07:17 CET] <darktoad> If they ever touch Big E from The New Day, I swear that i'am going in Syria to kill those noobs
[01:07:59 CET] <pepee> k
[01:09:39 CET] <pepee> interesting, avconv shows: [mp3 @ 0x21d6ae0] max_analyze_duration reached
[01:09:45 CET] <pepee> err, nm...
[01:10:36 CET] <waressearcher2> darktoad: was ist fur "Big E" ?
[01:29:50 CET] <aphorise> Is there a good cli tool / method to verifiy a file is complete - if it may  still be in a copy state / not finished?
[01:40:35 CET] <waressearcher2> aphorise: hallo
[01:47:57 CET] <Ripmind> does ffprobe really need to could EACH frame when i want to count them? is there no metadata for that? or some faster method? a 4gb movie file takes ages already o.O
[01:48:55 CET] <c_14> Ripmind: nope, no faster way (unless you're 100% sure the file is strict cfr)
[01:49:05 CET] <Ripmind> hmm
[01:49:26 CET] <Ripmind> tje 100% sure way would be to use fps * time?
[01:49:44 CET] <c_14> yep
[02:21:02 CET] <thebombzen> aphorise: if the container has a footer, then you could always try decoding it to /dev/null.
[02:21:43 CET] <thebombzen> ffmpeg will complain if it reaches the end of the file
[02:21:59 CET] <thebombzen> at least it will with sufficient verbosity. I don't know what that is, I'm afraid
[02:24:44 CET] <aphorise> thanks very much
[02:34:52 CET] <pepee> would converting vorbis -> aac lose too much quality?
[02:35:28 CET] <furq> too much quality for what
[02:37:04 CET] <pepee> with respect to the original audio, the output of a tts engine converted to ogg
[02:43:38 CET] <TD-Linux> you can be the judge, you have ears :)
[02:43:44 CET] <pepee> :/
[02:43:47 CET] <TD-Linux> probably the answer is no
[02:44:44 CET] <pepee> ok, thanks, I'll just use that, then
[02:45:06 CET] <furq> since you're concerned about speed, if you have the original wav it should be marginally faster to use that as the source
[02:45:53 CET] <c_14> It'll be even faster if you do both at the same time in one ffmpeg process because you'll save a fork+init
[02:46:16 CET] <furq> i also would expect a lossy transcode to be even worse at such low bitrates
[02:46:28 CET] <furq> even worse than it would be as a general rule, that is
[02:48:17 CET] <TD-Linux> well at these bitrates the differences aren't theoretical, he can just try it and see :) (also for choosing the bitrates in the first place)
[02:48:23 CET] <pepee> since I'm already making mp3/ogg audio files, I was thinking about making aac files too, and once needed, using that
[02:48:52 CET] <furq> i wouldn't bother with mp3 at all
[02:49:23 CET] <pepee> do all browsers support ogg?
[02:49:23 CET] <furq> also i noticed you're using libvo_aacenc which is a pretty bad encoder
[02:49:37 CET] <TD-Linux> sadly he's targeting the browser and certain vendors don't support ogg vorbis or ogg opus yet
[02:49:42 CET] <pepee> hmm, which one should I use?
[02:49:44 CET] <furq> all browsers support either aac or ogg
[02:49:56 CET] <furq> and libfdk_aac is the best aac encoder in ffmpeg
[02:50:03 CET] <pepee> wait, browsers support aac for audio files too?
[02:50:09 CET] <furq> http://caniuse.com/#feat=aac
[02:50:09 CET] <TD-Linux> pepee, only some but yes
[02:50:45 CET] <pepee> with what container(s)?
[02:50:52 CET] <furq> mp4
[02:53:25 CET] <furq> personally i'd go with aac and opus
[02:59:22 CET] <pepee> the static builds aren't compiled with libfdk-aac :(
[02:59:56 CET] <c_14> Just use the builtin encoder and a recent build from master
[03:00:20 CET] <furq> yeah i'm not sure why the static build includes libvo-aacenc
[03:00:25 CET] <furq> it's generally worse than the builtin aac encoder
[03:06:06 CET] <aphorise> I'm trying to build in cygwin & get: ' common.mak:57: recipe for target 'libavformat/tls_schannel.o' failed ... make: *** [libavformat/tls_schannel.o] Error ' - am I missing a dependecy?
[03:07:27 CET] <c_14> eeeh, probably disable schannel, not sure if that'll work in cygwin
[03:08:08 CET] <c_14> Unless you explicitly want it, in which case it might be a bug.
[03:11:19 CET] <aphorise> how do I do that?
[03:11:41 CET] <c_14> --disable-schannel
[03:12:13 CET] <aphorise> `./configure --target-os=cygwin --arch=x86_64 --disable-schannel`  ?
[03:12:18 CET] <c_14> yep
[03:12:35 CET] <c_14> Might need to fix the autodetection though...
[03:12:57 CET] <c_14> Can you pastebin more of the error?
[03:15:51 CET] <aphorise> c_14  https://paste.debian.net/334791/
[03:17:21 CET] <c_14> ye, looks to me like schannel won't build with cygwin
[03:17:46 CET] <aphorise> What is that related to?
[03:18:07 CET] <c_14> hmm?
[03:18:17 CET] <aphorise> what is `schannel` for?
[03:18:30 CET] <c_14> schannel? afaik Windows SSL api
[04:54:32 CET] <pepee> is there some sort of scripting language for ffmpeg? how come you can't configure some arguments depending on the input files?
[04:54:53 CET] <c_14> ffmpeg is complex enough without a scripting language
[04:59:52 CET] <pepee> hmm, I guess so
[07:18:30 CET] <tcq> Hey all. Does anyone know if transcode has a channel?
[07:52:22 CET] <waressearcher2> tcq: hallo
[07:57:35 CET] <tcq> Hi.
[07:57:41 CET] <tcq> Sorry, was in the other room.
[08:05:39 CET] <waressearcher2> tcq: wie geht's ?
[08:07:19 CET] <tcq> I have no idea what that means.
[08:14:37 CET] <jasom> tcq: it's scottish for tiny gates
[08:14:41 CET] Action: jasom ducks
[08:20:04 CET] <tcq> jasom, are you familiar with transcode?
[10:04:56 CET] <Fyr> does cutting and merging WAV or FLAC go without loosing information?
[10:05:25 CET] <fritsch> if you don't reencoded lossy
[10:05:26 CET] <fritsch> yes
[10:05:39 CET] <Fyr> ok
[10:05:49 CET] <fritsch> e.g. if your intermediate format can hold the decoded data
[10:06:19 CET] <Fyr> ffmpeg doesn't support multithreading when converting audio; I want to compress the first part of WAV into FLAC, then the second, the third etc.
[10:06:38 CET] <Fyr> then merge them into one FLAC file.
[10:07:02 CET] <fritsch> that sucks
[10:07:04 CET] <fritsch> :-)
[10:07:23 CET] <TD-Linux> that's possible with FLAC, but dunno about ffmpeg
[10:07:29 CET] <Fyr> I have 16 noded 12-core badass Intel Xeon cluster, but it converts audio with the same speed as my phone.
[10:07:32 CET] <fritsch> it's also possible for wav
[10:07:51 CET] <fritsch> but the very moment you use some formats that need "more frames" to reconstruct a single frame (like mp3)
[10:07:54 CET] <fritsch> you are doomed
[10:08:39 CET] <TD-Linux> Fyr, are you aware of libflac's speed levels?
[10:08:45 CET] <Fyr> if it doesn't loose frames, I can run a bash script to utilize all the cores.
[10:08:58 CET] <Fyr> TD-Linux, what levels?
[10:09:10 CET] <TD-Linux> you can trade off size for speed
[10:09:25 CET] <Fyr> I need smaller size. =)
[10:09:38 CET] <fritsch> then use 8, but run multiple encodes in parallel
[10:09:40 CET] <fritsch> :-)
[10:09:50 CET] <fritsch> e.g. one single encoder per thread you have
[10:09:54 CET] <fritsch> max parallelity
[10:09:57 CET] <fritsch> all cores used
[10:10:04 CET] <fritsch> nice for "multiple files"
[10:10:07 CET] <Fyr> I convert a 24-bit/192 kHz file. =)
[10:10:13 CET] <fritsch> one at a time?
[10:10:16 CET] <fritsch> or one on demand?
[10:10:42 CET] <fritsch> if you have 20 songs - use 20 threads that work in parallel without knowing of each other
[10:10:55 CET] <Fyr> it's not a song. =)
[10:11:00 CET] <Fyr> it's a signal. =)
[10:11:03 CET] <fritsch> okay
[10:11:06 CET] <Fyr> 1 hour-long
[10:11:12 CET] <fritsch> okay
[10:11:24 CET] <LiuYan> hi, i want to strip audio from video, however, I got an audio file in StartTime(Length)+Duration time length.  The command line issued is `ffmpeg -y -ss "05:52.5" -t "30.1" -i "Criminal.Minds.mp4"   -vn   -acodec copy "CriminalMinds-start.m4a"`, and the time length of output file is 00:06:22.60. FFMPEG version = 2.8.2
[10:11:26 CET] <fritsch> if you had 10 signals - that would also work
[10:11:37 CET] <fritsch> but I think you got the point: encoding audio in parallel is not so easy
[10:11:44 CET] <TD-Linux> Fyr, FLAC encodes in frames that you can stitch together afterwards, but ffmpeg itself can probably not do it, you might need a custom tool
[10:11:56 CET] <TD-Linux> you need to also make sure your encoder doesn't use variable frame sizes that would change the stitch points
[10:12:13 CET] <fritsch> and take care that each frame is really not related / needs information from frame n-1
[10:12:14 CET] <Fyr> ok
[10:12:58 CET] <TD-Linux> IIRC that's always the case with flac
[10:13:01 CET] <fritsch> write the input signal in the a buffer, take one chunk out per thread and organize the output
[10:13:34 CET] <fritsch> if you cannot hold the "complete hour in memory" - you need to write out from time to time
[10:14:24 CET] <fritsch> TD-Linux: if that's the case - then he can easily write a header, read 50 MB per thread or something and chunk out the frames behind the header
[10:15:19 CET] <Fyr> fritsch, how do I read a chunk from thread?
[10:15:37 CET] <fritsch> that's a C++ question right?
[10:15:38 CET] <fritsch> :-)
[10:16:08 CET] <fritsch> you "invent" threadsafe ringbuffer or something and an outbuffer
[10:16:38 CET] <fritsch> the threadsafe ringbuffer has an output_position_index and the thread reads that when reading it's data chunk
[10:16:50 CET] <fritsch> now the thread encode to flac
[10:17:00 CET] <fritsch> writes the outputbuffer and stores its position number
[10:17:19 CET] <fritsch> when all frames are finished, you write the header and concat the data in the order of output_position_index
[10:17:47 CET] <fritsch> you can make the output_buffer lockless the very moment you know how large the outputchunk is
[10:18:03 CET] <fritsch> and just hand over the target ptr to the encoder thread n
[10:18:22 CET] <Fyr> you mean programming avcode?
[10:18:29 CET] <fritsch> you need to care: input buffer might not hold the complete file, outputbuffer might be too small too
[10:18:33 CET] <fritsch> yes
[10:18:44 CET] <fritsch> custom solution for your personal usecase
[10:19:12 CET] <fritsch> as you get a signal over time, you can only start reading from the front
[10:19:49 CET] <Fyr> =(
[10:19:55 CET] <Fyr> I'm afraid I don't really want to bog on avcode swamp. =(
[10:20:00 CET] <fritsch> Signal: 10,9,8,7,6,5,4,3,2,1,0 -> first part goes to thread 0, 1 to next one, 2 to next one ... when 0 is finished it takes another chunk
[10:20:26 CET] <fritsch> http://christopherpeplin.com/2011/05/pflac/
[10:20:28 CET] <fritsch> ouh see
[10:20:32 CET] <fritsch> he had the same issue
[10:21:00 CET] <fritsch> yeah - read what he writes
[10:21:01 CET] <fritsch> nice
[10:24:13 CET] <Fyr> why do the developers not want to just add -threads option to FLAC encoding? =(
[10:25:22 CET] <LiuYan> Fyr: I'm not sure, because i tired, same result. The weird thing is: if I use same start_time and duration parameters to create GIF file, they all works fine.   >> My guess is that you should write the time stamp and the duration without point, only colons.
[10:25:32 CET] <LiuYan> s/tired/tried/
[10:26:24 CET] <Fyr> I use -ss and -t options all the time and never experienced errors. =)
[10:30:36 CET] <LiuYan> Fyr: I wrote 2 scripts to produce GIF and PNG series pictures, the all works fine. Now I want to strip audio only, so I copy & edit the one script for audio stripping, and now I got the unexpected time length of output audio: start_time in length + duration.
[10:31:42 CET] <Fyr> LiuYan, you should use option -report and post the log file onto pastebin.com
[10:32:16 CET] <Fyr> all of these words are bare without the output.
[10:32:34 CET] <LiuYan> Fyr: let me try the -report option ...
[10:41:01 CET] <LiuYan> Fyr: http://paste.fedoraproject.org/293241/14481851 , I saw all the options are parsed correctly, but still, in lin #79, I got this "size=    6098kB time=00:06:22.01 bitrate= 130.8kbits/s    "
[10:54:28 CET] <Fyr> LiuYan, <...> -ss XX:XX:XX -i XXXXXX.mp4 -t 30 <...>
[10:54:55 CET] <Fyr> -t parameter must follow the -i one.
[10:55:40 CET] <Fyr> LiuYan, the correct line will be:
[10:55:41 CET] <Fyr> ffmpeg -report -y -ss 00:05:52 -i "\xe7\x8a\xaf\xe7\xbd\xaa\xe5\xbf\x83\xe7\x90\x86.Criminal.Minds.S11E02.\xe4\xb8\xad\xe8\x8b\xb1\xe5\xad\x97\xe5\xb9\x95.HDTVrip.1024X576.mp4" -t 30 -vn -acodec copy CriminalMinds-start.m4a
[10:57:11 CET] <Fyr> try to use -c:a and -c:v, it looks like -acodec and -vcodec will be obsolete soon.
[11:03:44 CET] <LiuYan> Fyr: thanks, it works now.  But it still confusing me about the -t option, manual said: 'When used as an input option (before "-i"), limit the duration of data read from the input file.'. If the data read input file is limited to this $duration time length, why ffmpeg write more than $duration data out ...
[11:05:37 CET] <Fyr> it's confusing me too, but my guess is that audio and video have different seeking table, that's why it's not the same when cutting them separately.
[11:05:57 CET] <Fyr> or it's a bug.
[11:06:44 CET] <Fyr> I'm done sending ffmpeg bugs, if you think it's worth reporting, do it.
[11:11:39 CET] <Fyr> LiuYan, are you from China?
[11:13:29 CET] <LiuYan> Fyr: You mean you've already posted the bug report? >> I'm done sending ffmpeg bugs
[11:13:46 CET] <LiuYan> Fyr: yes, I'm from China.
[11:14:58 CET] <IntelRNG> LiuYan: I think he means he is not filling bug reports anymore
[11:15:48 CET] <LiuYan> IntelRNG: okay, sorry for my poor English reading skills.
[11:17:06 CET] <Fyr> yes, I'm sick of posting ffmpeg bug reports.
[11:17:06 CET] <LiuYan> Fyr: ok, I will report a bug. Thanks for the help!
[11:17:26 CET] <Fyr> LiuYan, could you recommend an Internet shop with good Chinese tea?
[11:18:35 CET] <Fyr> aliexpress/dhgate have plenty of them, but some of their green teas are not actually green.
[11:19:51 CET] <IntelRNG> Fyr: Is there any particular reason for you not to fill those reports any longer? Just curious here.
[11:20:07 CET] <Fyr> =)
[11:22:38 CET] <Fyr> I posted a bug, after a couple of weeks I found that it was neither fixed nor changed. I visited #ffmpeg-devel and asked them. They not very politely asked me for a donation. I made it, send them the confirmation. The bug still not fixed. =(
[11:23:20 CET] <Fyr> it was something simple, maybe option parsing and commenting/uncommenting some lines.
[11:23:54 CET] <Fyr> to be honest, they discussed the bug for 15 minutes.
[11:24:39 CET] <Fyr> then the conversation faded, they started to talk about other things.
[11:25:20 CET] <Fyr> after a month I came again and asked about the bug fixing. =)
[11:25:28 CET] <Fyr> they remembered me. xD
[11:26:43 CET] <Fyr> I mean, remembered that they asked me for a donation, which I made, and that they didn't solve the bug. =)
[11:26:59 CET] <LiuYan> Fyr: Sorry, I'm not familiar with tea (rarely drink tea). but for online shopping, i usually go several chinese websites.
[11:27:26 CET] <durandal_1707> Fyr: what bug?
[11:27:27 CET] <Fyr> LiuYan, do those websites send tea as well? =)
[11:27:47 CET] <Fyr> LiuYan, what do you drink, coffee?
[11:28:14 CET] <durandal_1707> Coca-Cola
[11:28:26 CET] <LiuYan> Fyr: I'll send the sites via pm.
[11:29:09 CET] <Fyr> durandal_1707, I posted that ffmpeg 2.8 doesn't support cover art or m4a.
[11:29:12 CET] <Fyr> 2.7 does
[11:29:35 CET] <Fyr> I posted it and still don't know if it's fixed or not.
[11:29:36 CET] <Fyr> =)
[11:29:43 CET] <Fyr> I don't use m4a anymore.
[11:29:56 CET] <Fyr> thanks to ffmpeg's developers. =)
[11:30:24 CET] <Fyr> OPUS is my sanctuary. =)
[11:30:25 CET] <durandal_1707> you mean coverart in m4a ?
[11:30:30 CET] <Fyr> yes
[11:35:02 CET] <Fyr> LiuYan, when I asked people on #qt-cn, they also said that they don't drink tea.   how is it possible that you don't drink it? I thought China's tea is world-famous.
[11:36:19 CET] <TD-Linux> <Fyr> why do the developers not want to just add -threads option to FLAC encoding? =(
[11:36:34 CET] <TD-Linux> most people use FLAC for music libraries, where they can just do each song in parallel
[11:38:13 CET] <Fyr> parallelization will bring effective utiliation of CPU and GPU anyway.
[11:39:00 CET] <Fyr> converting video with CPU when converting audio with GPU will save time.
[11:39:19 CET] <Fyr> especially if the input file has many audio channels.
[11:41:24 CET] <bencoh> "converting video with CPU when converting audio with GPU" that's quite funny if you think of it
[11:41:31 CET] <bencoh> next step is converting video with soundcard :p
[11:41:42 CET] <Fyr> they don't do anything, because ffmpeg is a monopolist. we should punish them with antimonopoly laws.
[11:42:08 CET] <Fyr> bencoh, converting audio without using CPU at all - it's the best thing ever.
[11:42:18 CET] <bencoh> :)
[11:42:50 CET] <Fyr> I used flacCL to convert 24-bit/192 kHz megalarge files without burning my CPU.
[11:43:30 CET] <Fyr> it's 200x times faster than using 3 CPU cores.
[11:51:18 CET] <Fyr> the most important thing, CPU is free, 0% of workload, so as video card. the temperature doesn't rise, you can use computer in whatever way you want.
[11:52:14 CET] <Fyr> meantime, 72 Gb file is being converted in background. =)
[12:40:16 CET] <TD-Linux> yeah, FLAC is one of the few formats where a fast GPU implementation is possible
[12:42:30 CET] <Fyr> moreover, it's already been implemented.
[12:42:42 CET] <Fyr> and proven to be useful
[13:03:26 CET] <LiuYan> Fyr: it seems there's a similar bug report: https://trac.ffmpeg.org/ticket/977
[14:47:58 CET] <c_14> LiuYan: If I had a patch, could you test it?
[15:02:53 CET] <Fyr> oh
[15:03:49 CET] <c_14> Well, it works for me so it's on the ml now.
[15:04:50 CET] <Fyr> c_14, do you know how to make the developers implement multithreading for audio conversion?
[15:05:08 CET] <fritsch> implement it yourself and send a PR?
[15:05:10 CET] <c_14> Make someone care.
[15:05:54 CET] <Fyr> this world is full of unperfectness. =)
[15:06:03 CET] <furq> imperfection
[15:06:10 CET] <Fyr> agreed
[15:10:36 CET] <Fyr> c_14, if I cut a WAV file into two pieces, convert each into FLAC, then merge them into one, and convert it back into WAV. will md5 hashes coincide?
[15:11:36 CET] <c_14> If the input WAV is 16 or 24bit, theoretically yes.
[15:12:15 CET] <Fyr> what about practice? does it loose frames?
[15:13:37 CET] <c_14> I just tested it and they match
[15:13:55 CET] <Fyr> awesome!
[15:13:57 CET] <Fyr> thanks
[16:35:23 CET] <Fyr> guys, could you provide me with link to manual of audio concatenation?
[16:36:02 CET] <Fyr> I cut a wav into pieces, but -f concat behaves in a strange way.
[16:37:32 CET] <c_14> https://trac.ffmpeg.org/wiki/Concatenate
[16:39:09 CET] <Fyr> thanks
[17:16:32 CET] <kruzin> hi, anyone here use ffmpeg on android?
[17:40:56 CET] <Fyr> me
[18:26:42 CET] Action: Fyr slaps kruzin around a bit with a large trout
[18:26:51 CET] <kruzin> lol Fyr
[18:26:58 CET] <kruzin> whats up?
[18:27:22 CET] <Fyr> you wanted to ask about ffmpeg for Android.
[18:27:31 CET] <kruzin> Yes Fyr
[18:27:42 CET] <kruzin> There's a small issue I'm facing
[18:27:58 CET] <kruzin> Heres my issue to be exact, everything works just fine in a new project I create, even on 64bit processors,  but when I try to run it in my existing project, somehow, the ffmpeg fails to execute
[18:28:09 CET] <kruzin> works like a charm on 32bit, 64 bit is where it just doesnt work
[18:28:40 CET] <Fyr> sorry, I don't know how to help. ='(
[18:28:46 CET] <kruzin> :/
[18:29:24 CET] <kruzin> no worries, thanks for trying :D
[18:29:28 CET] <JEEB> I think 64bit android had some funkyness
[18:29:41 CET] <kruzin> JEEB, cannot agree more.. :D
[18:30:08 CET] <kruzin> the funny thing though, in a standalone project, ffmpeg binary loads up just fine.
[18:30:23 CET] <Fyr> kruzin, guys on #ffmpeg-devel may help.
[18:30:24 CET] <kruzin> but, in my existing project, boom.. just doesn't wanna fire up
[18:30:38 CET] <JEEB> oh, ok. it wasn't 64bit it was intel-based android 6
[18:30:43 CET] <JEEB> https://trac.ffmpeg.org/ticket/4928
[18:30:49 CET] <JEEB> so probably unrelated
[18:31:50 CET] <JEEB> anyways, lemme check FATE
[18:32:05 CET] <kruzin> sure
[18:32:52 CET] <JEEB> ok, FFmpeg doesn't seem to have an android 64bit ARM instance, only aarch64 (iOS)
[18:33:09 CET] <kruzin> aarch64 is what the phone is using
[18:33:18 CET] <kruzin> Yureka Yu phone is my test phone for this
[18:33:29 CET] <JEEB> yeah, but it's tested with darwin OS X clang
[18:33:40 CET] <kruzin> :/
[18:33:42 CET] <JEEB> Libav seems to have a generic (?) aarch64 thing
[18:33:47 CET] <JEEB> which seems to be passing fine
[18:33:51 CET] <kruzin> Hmm
[18:33:54 CET] <JEEB> so it's specific to your thing most probably
[18:34:25 CET] <kruzin> I guess so too, I even tried it on a Lenovo which has a 64bit processor.
[18:34:48 CET] <kruzin> ffmpeg just fails
[18:35:18 CET] <kruzin> I bought a One Plus X, thinking it shouldn't work on that too.. surprisingly, it works.. then I noticed its 32bit. lol
[18:35:52 CET] <JEEB> no idea, I can just see that the test suite on Libav side works fine, so it most probably works fine on FFmpeg side as well.
[18:36:10 CET] <JEEB> so you'd have to look further into what exactly *you* are doing
[18:36:22 CET] <JEEB> esp. the "everything works just fine in a new project I create, even on 64bit processors" part
[18:37:04 CET] <kruzin> I believe my project's assets folder must be encountering issues with ffmpeg
[18:38:09 CET] <kruzin> I even tried to chmod ffmpeg's binary
[18:38:10 CET] <kruzin> lol
[18:38:58 CET] <JEEB> are you still able to run random binaries you distribute with android apps?
[18:39:10 CET] <JEEB> I thought you were limited to loading shared libs or so
[18:39:52 CET] <JEEB> but yeah, sounds like you have to debug something and check step-by-step what is going wrong on your side >_>
[18:40:09 CET] <kruzin> Yes
[18:40:19 CET] <kruzin> I must go through once again
[18:41:10 CET] <kruzin> JEEB, you this Daemon could guide? :/
[18:41:18 CET] <JEEB> no
[18:41:21 CET] <kruzin> Been struggling with this since about a week
[18:41:21 CET] <kruzin> haha
[18:41:25 CET] <JEEB> it's specific to android rather than FFmpeg
[18:41:29 CET] <kruzin> oh
[18:41:30 CET] <JEEB> and possibly your app
[18:41:38 CET] <JEEB> at least that's how it sounds
[18:41:51 CET] <kruzin> Feels so too.. something's up with my app
[18:41:52 CET] <kruzin> haha
[20:00:15 CET] <Fyr> c_14, when we discussed multithreaded FLAC conversion, I asked you if it were safe to concatenate FLAC files into one, you said yes. what command did you use to combine them?
[20:00:44 CET] <Fyr> I tried many times, ffmpeg looses some frames when concatenating.
[20:00:46 CET] <c_14> ffmpeg -f concat -i <(print "file $HOME/0.flac\nfile $HOME/1.flac\n") out2.wav
[20:01:50 CET] <Fyr> I used the same.
[20:01:54 CET] <Fyr> I converted the output FLAC file into s16le, wav, they all have different size.
[20:02:53 CET] <fritsch> 24 bit flac to s16 wav
[20:02:55 CET] <fritsch> ...
[20:02:56 CET] <Fyr> it also says: "[NULL @ 0000000000128060] sample/frame number mismatch in adjacent frames
[20:02:57 CET] <Fyr> [s16le @ 00000000001176c0] Application provided invalid, non monotonically increasing dts to muxer in stream 0: 1322496
[20:02:57 CET] <Fyr> >= 1322496"
[20:03:26 CET] <Fyr> I tried a sample 16-bit/44.1 kHz flac.
[20:03:47 CET] <Fyr> 1 minute-long, I cut it into two pieces, each 30 secods-log.
[20:04:07 CET] <Fyr> then merged them into WAV, FLAC and raw audio.
[20:04:24 CET] <Fyr> md5sums were the same in case of raw audio.
[20:04:44 CET] <c_14> I used this to generate the wav `ffmpeg -f lavfi -i aevalsrc='sin(1000*2*PI*t):c=2c:d=240' out.wav' and this to split it `ffmpeg -i out.wav -t 5 0.flac -ss 5 1.flac'
[20:04:48 CET] <Fyr> when combining WAV and FLAC, they were not the same.
[20:06:28 CET] <Fyr> and when concatenating FLAC files, it reported frame number mismatch. =/
[20:31:06 CET] <Fyr> is it possible to embed CUE into FLAC file?
[20:31:46 CET] <furq> do you mean in general or with ffmpeg
[20:31:51 CET] <Fyr> both
[20:32:05 CET] <furq> yes to the former
[20:32:09 CET] <furq> i don't know about the latter but i doubt it
[20:33:02 CET] <furq> you can do it inplace with metaflac
[23:24:14 CET] <Mallot_> let's say i've got two wide angle footages with a field of view overlap, how would i go about trying to stitch them together into one video?
[23:32:45 CET] <durandal_1707> Mallot_: crop and hstack?
[23:33:39 CET] <Mallot_> durandal_1707: the tricky part would be the seams - handling the distortions, is crop that sophisticated?
[23:34:33 CET] <durandal_1707> nope
[00:00:00 CET] --- Mon Nov 23 2015