[Ffmpeg-devel-irc] ffmpeg.log.20191110

[burek]

[00:01:14 CET] <mrkotfw> DHE: Let me try that... I know that with this VOB, I can't extract text, only "picture-based" subtitle
[00:08:44 CET] <mrkotfw> Sorry, I'm having issues with this... I can't seem to pass the VOB file as the filename parameter, even if I specify the subtitle stream index
[01:17:24 CET] <fresheyeball> ok
[01:17:36 CET] <fresheyeball> now I need to get this .ts file to be an avi because opencv only supports avi
[01:17:51 CET] <fresheyeball> anyone know the incantation to get ffmpeg to do the conversion?
[01:21:18 CET] <pink_mist> ffmpeg -i foo.ts -c copy foo.avi # maybe?
[01:21:35 CET] <fresheyeball> pink_mist: I just got it
[01:21:42 CET] <pink_mist> assuming it's using a codec that avi supports
[01:21:43 CET] <fresheyeball> but badly by passing thorugh .mp4 first
[01:47:42 CET] <aleksandrdvorkin> hi
[01:47:55 CET] <aleksandrdvorkin> its really quite in #kodi
[01:48:03 CET] <aleksandrdvorkin> and i have a question
[01:48:31 CET] <aleksandrdvorkin> why is last command to make and install the pvr.iptvsimple addons for Kodi on this link https://github.com/kodi-pvr/pvr.iptvsimple
[01:48:34 CET] <aleksandrdvorkin> ends with make
[01:48:38 CET] <aleksandrdvorkin> then how do i install it
[01:48:58 CET] <pink_mist> this is #ffmpeg
[01:49:10 CET] <aleksandrdvorkin> i know
[01:49:14 CET] <pink_mist> then wtf
[01:49:27 CET] <aleksandrdvorkin> ok thanks for your prompt and kind responce
[01:49:28 CET] <pink_mist> just because they're quiet doesn't make this the right place to ask
[01:49:42 CET] <aleksandrdvorkin> ignoledged
[02:32:49 CET] <nicolas17> pink_mist: "I tried to buy a Renault in a Renault shop, but there were no salesmen available, so I came into this Ford shop. Why won't you sell me a Renault?!"
[02:33:19 CET] <pink_mist> lol, exactly :P
[02:34:01 CET] <Reinhilde> "Because we're a ford shop."
[02:34:02 CET] <nicolas17> it's a factoid in the ##c++ bot, super handy
[03:04:13 CET] <HexGlaze> Can I have some help? In 0.mkv there is audio, video, subtitle and font streams. In 1.mkv, there is a video stream. How would I go about copying everything but the video stream from 0.mkv, and then adding all those to the video stream from 1.mkv?
[03:05:25 CET] <HexGlaze> So far, with "map_metadata" I'm able to copy everything but the fonts.
[03:06:09 CET] <pink_mist> personally, I would use mkvmergegui to do that
[03:06:50 CET] <HexGlaze> >GUI
[03:06:52 CET] <HexGlaze> Nope
[03:07:37 CET] <pink_mist> so just use mkvmerge then and figure out how to do it from commandline
[03:07:51 CET] <pink_mist> I find it easier to use the gui for that though
[03:08:35 CET] <HexGlaze> I just really need to know what ffmpeg calls font streams...
[03:08:59 CET] <JEEB> those are not streams, but attachments
[03:09:07 CET] <HexGlaze> Well, attachments then
[03:13:02 CET] <HexGlaze> Almost got it with "-map 0:a -map 0:s"
[03:13:15 CET] <HexGlaze> Now, what are attachments prefixed as?
[03:17:56 CET] <HexGlaze> "[matroska @ 0x55bb64cc5b00] Attachment stream 5 has no filename tag." Hmm
[03:23:36 CET] <HexGlaze> Okay, got it
[03:24:26 CET] <HexGlaze> It's "ffmpeg -i original.mkv -i 60fps_videostream.mkv -map 1 -c copy -map_metadata 0 -map 0 -map -0:v -map_metadata:s:v 0:s:v -map_metadata:s:a 0:s:a -map_metadata:s:s 0:s:s -map_metadata:s:t 0:s:t out.mkv"
[03:29:24 CET] <HexGlaze> Oh wait, I'm not done
[04:16:23 CET] <HexGlaze> Okay, finished: "ffmpeg -i original.mkv -i 60fps_video.mkv -map 1:v -c copy -map_metadata 0 -map 0 -map -0:v out.mkv"
[05:04:14 CET] <montana> i encoded a 64kbps mono to  24kbps opus file  and it sounded fine  and i encoded that same 24 kbps opus file again and it sounds bad.  is this normal
[05:06:39 CET] <nicolas17> each lossy encoding will lose quality
[05:07:53 CET] <montana> i understand that but i only did it twice
[05:08:18 CET] <montana> i've seen people doing 100 times (for tesing)
[05:08:48 CET] <JEEB> yea the lower you go and the more noise/"detail" there is the harder it gets
[05:08:55 CET] <JEEB> although it is specific to the encoder
[05:09:44 CET] <JEEB> so if you want technical explanations then I recommend asking the people behind the opus encoder you're utilizing
[05:09:50 CET] <montana> https://hydrogenaud.io/index.php/topic,100067.0.html
[05:10:12 CET] <montana> wrong one:  http://bernholdtech.blogspot.com/2013/03/Nine-different-audio-encoders-100-pass-recompression-test.html
[05:11:22 CET] <montana> that website did it 100 times, but i only did it twice and it already sounds bad
[05:11:46 CET] <JEEB> as I said, I recommend asking the people behind the encoder you're utilizing :)
[05:13:38 CET] <nicolas17> they did it 100 times with a much higher bitrate
[05:13:56 CET] <montana> nicolas17 true
[07:01:48 CET] <montana> i encoded a 64kbps mono to  24kbps opus file  and it sounded fine  and i encoded that same 24 kbps opus file again and it sounds bad.  is this normal
[07:02:41 CET] <Jessidhia> sounds like https://en.wikipedia.org/wiki/Generation_loss
[07:03:20 CET] <JEEB> as I already said, ask the encoder's developers :P
[07:03:34 CET] <JEEB> if it's normal or not
[07:03:44 CET] <furq> i asked a question  and i got an answer  and i asked that same question again two hours later. is this normal
[07:04:26 CET] <montana> jeeb they are saying they cannot hear the difference in #opus:  can you hear the difference:   https://x0.at/zj6.mka  and https://x0.at/IQs.mka
[07:06:43 CET] <JEEB> no idea :P
[07:45:00 CET] <montana> why does  aac avcodec encoder sound so bad
[07:45:05 CET] <montana> is that one that ffmpeg use
[07:50:14 CET] <montana> how is aac so popular if it sounds this bad
[08:00:12 CET] <montana> is it normal that even vorbis sounds better than aac?
[08:01:44 CET] <kepstin> ffmpeg's builtin aac encoder isn't particularly good
[08:01:56 CET] <kepstin> probably still better than mp3, tho.
[08:02:03 CET] Action: kepstin hasn't actually checked
[08:02:24 CET] <montana> why can't ffmpeg have a decent aac encdoer
[08:02:51 CET] <kepstin> ffmpeg's builtin aac encoder is mostly useful for encoding for apple devices which need aac when you don't have an ffmpeg build with a better aac encoder
[08:03:22 CET] <Jessidhia> I remember there were some improvements done many years ago but I think it lost momentum after some library whose name I forgot was released
[08:03:24 CET] <montana> kepstin then what audio encoder do you use when you use ffmpeg
[08:03:34 CET] <kepstin> ffmpeg can be built with the 'fdk' aac encoder, but due to license conflicts, the result isn't redistributable. I build my own ffmpeg so i can use it when needed
[08:03:44 CET] <kepstin> but for my own use i normally use opus
[08:04:23 CET] <montana> kepstin i see
[08:04:42 CET] <montana> opus has horrible 2 pass though
[08:05:22 CET] <montana> http://bernholdtech.blogspot.com/2013/03/Nine-different-audio-encoders-100-pass-recompression-test.html
[08:06:14 CET] <kepstin> that's not 2 pass tho, that's 100 pass.
[08:07:01 CET] <kepstin> (and from a comment of a developer, they suggest that a lot of the actual loss is from conversion back and forth from 44.1 to 48 khz every time it's decoded/encoded)
[08:07:18 CET] <kepstin> er, well, that's not correct, but whatever
[08:07:55 CET] <kepstin> was due to sample misalignment with the frame size
[08:08:08 CET] <Jessidhia> ah, misalignment, that makes sense
[08:08:10 CET] <kepstin> which would also be an issue with other codecs
[08:08:23 CET] <montana> nero-aac sounds good even after 100 pass
[08:08:38 CET] <kepstin> probably a bug in an encoder/decoder library where it wasn't removing the delay when decoding or something like that
[08:10:37 CET] <kepstin> i expect that if that test was redone with the opus sample-aligned correctly, or with sample delays added to the other codecs so they aren't aligned in later passes (the latter would be more realistic for a multiple encode use case), the results would be more accurate/interesting
[08:10:41 CET] <Jessidhia> it sounded like multi-pass encoding but it's actually 100 generations? D:
[08:11:10 CET] <kepstin> usually when doing re-encoding audio, you're doing it because you changed something, and when you change something, that usually means you cut it, so samples won't realign.
[08:11:48 CET] <montana> ffmpeg -i test.mka -c:a libfdk_aac -b:a 32k output.m4a
[08:12:00 CET] <montana> even libfdk_aac sounds bad
[08:12:22 CET] <kepstin> aac-lc sounds pretty bad ad 32K, yeah
[08:12:34 CET] <kepstin> there's he modes, but they're not designed for music.
[08:12:40 CET] <Jessidhia> isn't that the lowest bitrate LC does
[08:12:58 CET] <kepstin> you can tweak the lowpass in fdk_aac to improve it on specific signals
[08:13:12 CET] <montana> what should i do to improve the sound then using 32kbps
[08:13:18 CET] <kepstin> use opus
[08:13:23 CET] <kepstin> or use more bits
[08:13:37 CET] <montana> i am using for testing
[08:13:54 CET] <montana> why can't i use he-aac
[08:14:15 CET] <kepstin> he-aac is actually a different codec, and it's not as well supported by decoders
[08:14:44 CET] <montana> wait it does work:  ffmpeg2 -i test.mka -c:a libfdk_aac -profile:a aac_he -b:a 32k output.m4a
[08:15:01 CET] <montana> that sounds better
[08:15:35 CET] <montana> okay using he-aac makes it better
[08:16:08 CET] <kepstin> if you're using fdk-aac lc, you want to tweak the value for the -cutoff option to adjust the lowpass, the default is very low.
[08:16:25 CET] <kepstin> adjusting it higher gives you more high frequencies, but also more audible artifacts
[08:16:29 CET] <montana> ffmpeg -i input.wav -c:a libfdk_aac -profile:a aac_he_v2 -b:a 32k output.m4a    : this does not work though
[08:16:49 CET] <montana> does "aac_he_v2"  work on your ffmpeg version?
[08:17:27 CET] <kepstin> haven't tried, i have no use for aac-he
[08:17:45 CET] <kepstin> since i encode web video, and browsers don't support it
[08:18:39 CET] <montana> can you try the commmand to see if it works
[08:18:47 CET] <montana> just for testing
[08:18:52 CET] <montana> ffmpeg -i input.wav -c:a libfdk_aac -profile:a aac_he_v2 -b:a 32k output.m4a
[08:31:37 CET] <mtcdood> I'm looking for advice on transcoding. If my source video footage was recorded in a constant bitrate and I want to transcode to a lower resolution would there be a benefit or drawback to using a constant bitrate for the reduced resolution too?
[08:32:11 CET] <mtcdood> at first glance it seems like there'd be no point in using a variable bitrate since data is already lost during fast motion sequences etc in the original
[08:32:49 CET] <montana> what vidoe/audio codec do you want use
[08:32:58 CET] <montana> that should be your first decision
[08:33:09 CET] <mtcdood> I want to provide footage to viewers over HLS in browsers
[08:33:17 CET] <mtcdood> so let me look at whether or not that limits me
[08:33:19 CET] <montana> HLS?
[08:34:39 CET] <mtcdood> HTTP live streaming
[08:34:40 CET] <ritsuka> constant bitrate would mean a lot of unneeded data
[08:35:12 CET] <ritsuka> or worse quality at the same bitrate
[08:35:33 CET] <mtcdood> even if the original video is already constant bitrate?
[08:36:09 CET] <furq> the source encoder settings make no difference at all
[08:36:26 CET] <mtcdood> ok
[08:36:43 CET] <furq> also yeah you should use vbr for hls
[08:37:31 CET] <mtcdood> in my very limited testing I was finding my ultimate filesizes for VBR much higher than expected for reduced qualities
[08:37:35 CET] <mtcdood> *resolutions
[08:37:36 CET] <furq> using cbr is just going to make the quality loss in high complexity scenes even worse
[08:38:46 CET] <ritsuka> if you set the bitrate yourself, the average bitrate will be the one you set
[08:39:05 CET] <ritsuka> if you are using constant quality or rate factor, you can limit the max bitrate
[08:39:15 CET] <mtcdood> yeah
[08:39:57 CET] <mtcdood> will limiting the maximum bitrate slow down encoding to get the same quality in a lower bitrate for high complexity scenes?
[08:40:50 CET] <furq> no
[08:40:55 CET] <furq> using a slower preset will do that
[08:41:11 CET] <mtcdood> ok
[08:42:00 CET] <mtcdood> then the consequences of limiting the maximum bitrate would be it just won't reach that quality during those high complexity scenes if the bitrate it would normally use at that preset is higher than the maximum bitrate selected?
[13:00:28 CET] <snooky> mooin
[13:20:31 CET] <CounterPillow> I wonder if ffmpeg supports sugandese locale
[13:22:49 CET] <furq> great content
[13:22:50 CET] <furq> thanks for sharing
[13:23:41 CET] <CounterPillow> furq please don't ruin this, someone in here recently fell for ligma
[13:24:19 CET] <furq> damn that's epic!
[13:25:49 CET] <CounterPillow> thanks it is
[15:31:51 CET] <montana> what is the difference between vorbis and libvorbis in ffmpeg
[15:36:03 CET] <durandal_1707> do not use native vorbis _encoder_
[15:37:12 CET] <CounterPillow> montana: vorbis is the builtin low quality vorbis encoder that you shouldn't ever use
[15:37:35 CET] <montana> then why is it even included in the first place
[15:38:59 CET] <CounterPillow> ¯\_(Ä)_/¯
[15:39:40 CET] <durandal_1707> it is experimental
[16:25:06 CET] <jemius> what? Why would the native encoder be worse than the ffmpeg version?
[16:25:13 CET] <jemius> or is it vice versa?
[16:33:13 CET] <furq> the native encoder is the one built into ffmpeg
[16:33:27 CET] <furq> as opposed to libvorbis which is the one xiph makes
[17:18:24 CET] <pink_mist> (so jemius was right, the native encoder cannot possibly be worse than the ffmpeg version, since they're the same thing :P)
[17:21:14 CET] <jemius> wohoo! Do I get a price now?
[17:27:23 CET] <furq> $7.99
[17:32:38 CET] <Lantizia> hey I've got 4 avi files I'd really like to "re-share" with the internet, I won't go into more detail than that as it'd be off topic :)  Suffice to say I think these avi files have had their metadata changes or file index rebuilt or something ... because even though they're the exact filesize (down to the byte) that they were downloaded as...
[17:32:57 CET] <Lantizia> when checked by the program to see if the checksums are good to share back up - it doesn't think they're the right files
[17:33:19 CET] <Lantizia> so I'm wondering - is there a way I can remove all data from an AVI except for the frames themselves?  leaving empty space in between wherever other data might be?
[17:34:09 CET] <Lantizia> oh and leave the audio too
[17:52:33 CET] <CounterPillow> you can remux them with a stream copy of the audio and video data and telling ffmpeg not to copy any of the other data, but last I tried that with something for extreme filesize optimisation purposes I noticed ffmpeg still wrote out certain metadata
[17:52:53 CET] <CounterPillow> but also what kind of idiot sells their porn in .avi format in 2019
[17:53:20 CET] <Lantizia> lol nothing as fun as porn
[17:53:48 CET] <Lantizia> just something I'd like to re-seed as it's very hard to find as it was only on VHS
[17:53:57 CET] <Lantizia> ed21bfe9954d344de4bf277870d9868475c7ce0e
[17:54:27 CET] <furq> you won't be able to do that if you modify the files in any way
[17:55:01 CET] <CounterPillow> yeah, the hashes won't match
[17:55:02 CET] <Lantizia> furq, i was hoping things like deluge would look at particular chunks maybe and know that *some* parts of it match?
[17:55:08 CET] <CounterPillow> nope
[17:55:12 CET] <CounterPillow> torrent protocol doesn't work like that
[17:55:20 CET] <Lantizia> hmm then why does it work if the file is partially downloaded?
[17:55:30 CET] <furq> the file positions of the chunks would have to match
[17:55:34 CET] <furq> if you remux then they probably won't
[17:55:34 CET] <Lantizia> you can throw it at a new client - tell it to re-check - and it'll know the percentage
[17:55:49 CET] <CounterPillow> because it checks the hashes of the individual chunks
[17:56:01 CET] <furq> but even then nobody would be able to download it from you because you'd be missing chunks
[17:56:12 CET] <Lantizia> yeah which is why i think this file has the audio/video data in the correct place - precisely the same filesize as what is on the torrent
[17:56:20 CET] <Lantizia> just want to remove all the other data
[17:56:21 CET] <CounterPillow> or has completely different chunks because the blanked out sections don't match the chunk borders
[17:57:08 CET] <Lantizia> furq, was i going to leave it seeding for a few months - hope some has it even if they're online for just 5minutes -fill in the tiny blanks
[18:08:48 CET] <jemius> ahm, how can I force ffmpeg to give me exactly one minute from the input stream? with -t it gives me 1:04
[18:10:59 CET] <furq> if you're using -c copy then you can't
[18:11:08 CET] <furq> if you're not the -t 60
[18:11:10 CET] <furq> then
[18:11:24 CET] <jemius> crap
[18:40:18 CET] <jemius> ffmpeg -i "concat:test.mkv|test2.mkv|test3.mkv" -c copy gesamt.mkv
[18:40:31 CET] <jemius> Why does this line not concatenate my videos? I get only the first one in gesamt.mkv
[18:42:26 CET] <furq> because you can't use the concat protocol with mkv
[18:42:29 CET] <furq> use the demuxer
[19:06:26 CET] <jemius> do you think it's worth upscaling a video with ffmpeg? I assume that my TV and other devices are just as good in upscaling the original material?
[19:09:03 CET] <furq> yeah they are
[19:09:25 CET] <furq> if you know your device is very bad at it then maybe it's worth it
[19:09:49 CET] <furq> but normally it's only useful for stuff that demands a specific scale, like youtube
[00:00:00 CET] --- Mon Nov 11 2019