[Ffmpeg-devel-irc] ffmpeg.log.20141130

burek burek021 at gmail.com
Mon Dec 1 02:05:01 CET 2014


[00:00] <wyatt8740> okay
[00:00] <wyatt8740> still confused about how mergeplanes works
[00:02] <wyatt8740> is anyone there?
[00:05] <wyatt8740> aha
[00:05] <wyatt8740> got it
[00:06] <wyatt8740> Sorry :\ the hexadecimal number thing was a little overcomplicated
[00:06] <wyatt8740> but I figured it out I think
[01:13] <Jan-> I make qscale bigger for less bitrate, right?
[01:13] <c_14> Depends on the encoder, but usually yes.
[01:13] <Jan-> -c:v prores_ks
[01:14] <Jan-> is there no way to actually specify a bitrate
[01:14] <c_14> -b:v usually works
[01:14] <Jan-> so why does everyone suggest qscale?
[01:15] <c_14> Because with -b:v if say, you have a section of video that is very plain and doesn't require much bitrate it will still get a lot of bitrate while if there's a section with lots of motion that might need more bitrate it won't get it.
[01:16] <c_14> With qscale/crf, it can give more bitrate to areas that need more and less to areas that need less.
[01:17] <Jan-> the thing is, with prores, it should be constant bitrate really
[01:17] <Jan-> although the apple decoder is quite tolerant
[01:17] <c_14> I don't know much about prores tbh.
[01:18] <Jan-> basic DCT codec
[01:18] <Jan-> used a lot in film and TV
[01:18] <Jan-> high bitrates, 444 etc
[01:23] <c_14> I think -b:v should work. I can't really give any recommendations though since I've never used it before.
[01:23] <Jan-> I'm honestly not sure if the apple encoder pads very simple frames to 50 megabits per second or whatever though
[03:15] <t4nk793> tell me please what should I di if I see copiler's output " attribute_deprecated - use encoder private options instead"
[05:08] <doki_pen_> I stupidly concatted a bunch of VOBs together and now the audio stream doesn't start up until about 10 seconds into the VOB. Anyway to get it into the output?
[05:09] <doki_pen_> it's stream 2, I try -map 0:2 but it doesn't exist at the begining
[05:22] <c_14> Try increasing probesize/analyzeduration ?
[05:46] <doki_pen_> c_14: no, I ended up just doing tail -c +<somebignumber> > out.vob
[05:46] <doki_pen_> until I cut the menus and crap out
[07:58] <Alina-malina> hmmm how to record the desktop with ffmpeg?
[08:08] <bk76> hi
[08:08] <bk76> anyone here?
[08:12] <Alina-malina> yes i am here
[08:13] <bk76> Alina can you tell me if ffmpeg can download video from youtube?
[08:14] <Alina-malina> which video you want to download?
[08:18] <bk76> this one https://www.youtube.com/watch?v=cN_JA_XnPgs
[08:22] <Alina-malina> ok downloading, wait i upload it somewhere
[08:22] <Alina-malina> and you can download it
[08:23] <bk76> but what about ffmpeg?
[08:23] <bk76> I have a lot more videos I want to download
[08:24] <Alina-malina> give me links i will download them:)
[08:24] <bk76> can I make youtube source for ffmpeg?
[08:25] <Alina-malina> bk76, http://dfiles.ru/files/3bib66zi8 here you go
[08:26] <bk76> Alina-malina: thx
[08:27] <Alina-malina> bk76, i doubt you can download with ffmpeg from youtube, if you have more p.m. me links i have automated stuff i can download and upload
[08:27] <Alina-malina> for you
[08:27] <bk76> Alina-malina: no thx, I'm looking for automated solution for myself :)
[08:28] <Alina-malina> bk76, you hve found:)
[08:29] <bk76> Alina-malina: too much manual work
[08:30] <Alina-malina> bk76, lol, there is no perfect solution that you just clap your hand and it just happen
[08:32] <bk76> Alina-malina: I'm programmer, there is nothing imposible for me
[08:32] <Alina-malina> bk76, lol me too
[08:33] <bk76> :)
[08:33] <bk76> can you downoad this: https://www.youtube.com/watch?v=7p4uAxXyaQ0
[08:33] <Alina-malina> sure
[08:42] <Alina-malina> bk76, http://dfiles.ru/files/ligq0pn7q here
[08:46] <bk76> Alina-malina: thx
[08:47] <Alina-malina> bk76, yw
[11:09] <bert__> So I'm trying to do on-the-fly transcoding of a mpegts with cropdetect and cropping combined, which is to say I want to parse the cropdetect output as -vf crop input. I can pipe STDERR to output.txt, but this gives me the whole stderr out. So question: is it possible to ONLY parse the relevant cropdetect output ("crop=...") into -vf crop?
[11:21] <relaxed> bert__: use awk
[11:31] <bert__> relaxed: yeah, thought of it, but I'm on a windows box and would really like to do it all in one .bat. Stuff like Gawk for Win seems a bit overkill.
[11:37] <relaxed> the whole point of a batch file is to do everything in one file, including using awk or whatever you decide on.
[11:45] <bert__> alright, so I'm installing gawk now and will try to figure this out. Strange there was only one guy on stackoverflow with a similar question. This doesn't seem that uncommon tbh.
[11:45] <bert__> Combining cropdetect and crop, that is.
[11:47] <relaxed> it should be trivial to script
[14:23] <holloway> Hey there! Does anyone know why my mp3s would be showing up as movie files in my forked-daapd? Thanks in advance for any assistance!
[14:23] <holloway> *forked-daapd itunes library that is
[14:32] <relaxed> holloway: did you encode the mp3s with ffmpeg?
[14:33] <holloway> no. Im testing a single album i downloaded
[14:33] <holloway> So im not sure what they were encoded using
[14:37] <holloway> according to logs forked-daapd uses ffmpeg libs and not libav
[14:37] <holloway> hence my presence here lol
[14:38] <relaxed> ffmpeg writes ablum art as if it were a video stream
[14:38] <relaxed> album*
[14:39] <relaxed> To remove it run,  ffmpeg -i input -map 0:a -c copy output.mp3
[14:41] <holloway> Okay Ill try that. Is there a way to process multiple files at all?
[14:41] <relaxed> on linux?
[14:42] <holloway> well im running it from the shell of a headless server thats arm based, the DNS321
[14:43] <holloway> ive got the new eminem album added to the library, so i could process one file but ill eventually need to do em all, and im unfamiliar with ffmpegs commandline capabilities
[14:43] <relaxed> for i in *mp3; do ffmpeg -i "$i" -map 0:a -c copy "${i%.*}".fixed.mp3; done
[14:46] <holloway> Oooh okay so that maps .mp3 onto each instance of i and runs everything?
[14:47] <holloway> Thanks for all your help, I hope it works lol. I have a feeling I may just have to update the database by loading the library locally
[14:47] <relaxed> holloway: yes, see http://mywiki.wooledge.org/BashGuide/TestsAndConditionals
[14:48] <eater> Hi! Is anybody aware of differences between ffmpeg h264 mpegts streaming and vlc udp streaming? My ffmpeg command is: ffmpeg -f decklink -i 'DeckLink Mini Recorder at 11' -threads 16 -acodec mp2 -ab 192k -ar 48k -ac 2 -vcodec libx264 -vb 12M -minrate 12M -maxrate 12M -bufsize 24M -g 50 -pix_fmt yuv420p -profile:v high -level 4.2 -f mpegts udp://172.21.7.4:5000. When I send stream directly from ffmpeg to appeartv device the stream is not de
[14:48] <eater> But if I send it ffmpeg -> vlc (udp stream legacy) -> appear tv it works perfectly... any ideas?
[14:48] <holloway> Thanks a lot relaxed!
[14:51] <relaxed> eater: did you look at "ffmpeg -h muxer=mpegts" ?
[14:57] <eater> no I didn't .. I'm total beginner with encoding/streaming etc :( can you give my any clue what is wrong or what should i set in mpegts muxer?
[14:59] <relaxed> I don't have a clue on your problem, just thought those options might be useful.
[15:01] <eater> ok, thank you relaxed! I will do some reading about mpegts muxer options.
[15:02] <c_14> eater: you could also try ffprobing the vlc udp output just to make sure it's the same format/codec
[15:05] <eater> c_14 ok I'll check with ffprobe format of that stream from vlc! thank you! Codec must the same because it's large h264 full hd stream, any transcoding on machine (it's very weak) with vlc would result in 100% cpu usage :)
[18:37] <rathodvikas10> hi i am using centos, i want to install ffmpeg bt cannot install lame
[18:38] <c_14> https://trac.ffmpeg.org/wiki/CompilationGuide/Centos
[18:38] <c_14> just ignore the parts about lame
[18:40] <rathodvikas10> ok
[18:41] <rathodvikas10> why lame is used to build ffmpeg
[18:52] <c_14> hmm?
[19:40] <ralphcor> Hi, I've a .m4a containing just a stereo AAC.  Right is silent, it should be mono.  I'd like a .m4a with mono AAC containing left, but with no recoding.  It's the "no recoding" I'm having trouble with.  Any clues?
[19:41] <ralphcor> Have tried   -filter_complex 'channelsplit=channel_layout=stereo[L][R]' -map '[L]' left.m4a -map '[R]' right.m4a   but that recodes.  And you can't -acodec copy because that's mixing filtering and streamcopy.
[19:41] <ralphcor> Is AAC open to having one channel split off in this way?
[19:41] <BtbN> That's not possible
[19:42] <ralphcor> With ffmpeg or underlying issue with AAC?
[19:43] <wyatt8740> I do not know how AAC compresses and encodes, but if it encodes both channels separately I'd say its possible. Implementation is another mattere
[19:43] <wyatt8740> *matter
[19:43] <JEEB> yeah, but with generic stream copying it most definitely is not possible
[19:43] <JEEB> you'd have to know the ins and outs of the format :P
[19:43] <JEEB> if you care enough, find the spec and start reading, ho
[19:44] <wyatt8740> you probably won't realize the difference if you choose a high enough bitrate for aac.
[19:44] <wyatt8740> with a recoding
[19:44] <wyatt8740> It sucks but sometimes you have to.
[19:45] <ralphcor> Agreed.  Would just seem nice not to, given it seems possible in theory.  Thought my ffmpeg-fu might be lacking.
[19:45] <wyatt8740> in theory, yes. but -acodec (or -c:a) copy is not per channel :\
[19:46] <JEEB> you would specialized code for anything like that
[19:46] <JEEB> basically since copying a track is just that
[19:46] <JEEB> trying to copy the underlying track as+is
[19:46] <ralphcor> Right.  So I thought filtering might do it.
[19:46] <wyatt8740> you could just use FLAC.
[19:46] <wyatt8740> if you care that much
[19:46] <JEEB> no, filtering is something you do on raw audio and video
[19:46] <JEEB> there are bit stream filters, but none that do this (this is under the stupid assumption that it would be possible to do what you want to do)
[19:46] <wyatt8740> was wondering about a filter, too. but that would require recoding
[19:47] <JEEB> so you either have to get really down and dirty yourself
[19:47] <ralphcor> That's what I've done so far, FLAC, but astats reports high 'peak count' and 'flat factors' compared to the original.
[19:47] <JEEB> or give up
[19:47] <wyatt8740> you could request a feature.
[19:47] <wyatt8740> I got hqx scaling implemented by a request.
[19:47] <JEEB> yes, but that is actually somewhat useful
[19:47] <JEEB> this is just "haxx0r the bit stream for me pl0x"
[19:48] <JEEB> doesn't exactly get anyone interested
[19:48] <wyatt8740> heh You're probably right
[19:48] <ralphcor> No, it's pointing out ffmpeg's manipulation of channels can only be done at a more coarse level than technically possible.
[19:48] <JEEB> uhh
[19:48] <ralphcor> wyatt8740: Where did you raise the request?  Issue tracker?
[19:48] <JEEB> the manipulation of channels is on raw audio data
[19:49] <wyatt8740> I believe it was there. was months ago
[19:49] <wyatt8740> lemme check
[19:49] <JEEB> also you are still talking under the assumption that what you want is possible
[19:49] <ralphcor> JEEB: Ah, you mean after decompression.
[19:49] <JEEB> yes
[19:49] <ralphcor> *decoding
[19:49] <wyatt8740> yep
[19:49] <wyatt8740> http://trac.ffmpeg.org/ticket/3404
[19:49] <ralphcor> Right.
[19:49] <JEEB> all filtering except for bit stream filters are on raw video or audio data
[19:49] <JEEB> which are under bsfs
[19:50] <JEEB> also I have no idea what astats you're looking at, but I wouldn't be surprised if you'd get some statistical differences due to the fact that there's no longer a goddamn silent side in there
[19:50] <JEEB> or you have PEBKAC'd the mono'ification
[19:51] <ralphcor> ffmpeg's astats chucks out per channel and then a total at the end.
[19:55] <ralphcor> http://wiki.multimedia.cx/?title=Understanding_AAC   tells me a bit about the format, ffmpeg can obviously decode it, perhaps using a library, is there an option to have it verbosely dump details as it goes?  Might shed light on whether it's got "single channel" or "channel pair" elements in it.
[19:58] <JEEB> just get the spec
[19:58] <JEEB> while multimedia wiki is a good thing
[20:55] <MadTBone_> I'm getting to know libav* data structures a little more, and I'm trying to see examples of when AVPacket->dts and ->pts differ.  I understand that the pts should never precede the dts, but in all of my test code they're the same.  In what situation would dts precede pts?  Is it encoded into the file somehow, or is it codec/(de)muxer dependent?
[20:56] <JEEB> take an AVC (H.264) or HEVC (H.265) stream encoded for high compression (lots of refs, lots of bidir [b-frames])
[20:56] <JEEB> then check it out
[20:56] <MadTBone_> JEEB, of course! bframes
[20:57] <MadTBone_> etc...
[23:20] <emhs> I'm trying to overlay a very wide backdrop behind a portrait-mode cell phone video. I'm using this command: ffmpeg -i Bunnyflake\ Backdrop.png -i Necessity.mp4 -strict -2 -filter_complex 'overlay=1167:0' -shortest Bunnyflake.mp4
[23:21] <emhs> This command works, except I can't see the backdrop, only the video: ffmpeg -i Bunnyflake.mp4 -i Bunnyflake\ Audio.wav -map 0:v -map 1:a -strict -2 Bunnyflake-Suit.mp4
[23:22] <emhs> Wait, wrong command
[23:22] <emhs> ffmpeg -i Necessity.mp4 -i Bunnyflake\ Backdrop.png -strict -2 -filter_complex 'overlay=-1167:0' Bunnyflake.mp4
[00:00] --- Mon Dec  1 2014


More information about the Ffmpeg-devel-irc mailing list