[Ffmpeg-devel-irc] ffmpeg.log.20170816

[burek]

[01:07:08 CEST] <Justanick> kepstin: ffmpeg had enabled -mips32r6 and -mips32r5 which triggers the problem. only -mips32r2 is supported by the cpu.
[01:20:16 CEST] <sh4rm4^bnc> what's the recommended format for audio in mp4/mkv ?
[01:20:43 CEST] <sh4rm4^bnc> not for hi-fi, but it should be good enough so one can clearly distinguish voices from background noise
[01:20:59 CEST] <iive> aac
[01:21:08 CEST] <sh4rm4^bnc> i currently use aac 80kbit, but it's hard to hear the voices
[01:21:39 CEST] <sh4rm4^bnc> blends too much into the backgrounds
[01:21:56 CEST] <iive> mono?
[01:22:03 CEST] <sh4rm4^bnc> no
[01:22:17 CEST] <iive> 80kbps is just too low.
[01:22:32 CEST] <sh4rm4^bnc> well, i don't specifically specify stereo, if you mean that
[01:22:44 CEST] <iive> even opus uses 96kbps by default, and it is 2 generations after aac.
[01:22:51 CEST] <sh4rm4^bnc> i have mp3s with 80kbit that sound just fine...
[01:23:03 CEST] <iive> i don't believe that.
[01:23:53 CEST] <tdr> depending what they are, could be possible
[01:24:12 CEST] <sh4rm4^bnc> IIRC it was RZA-The Chase
[01:24:15 CEST] <tdr> a bet a spectrum analyzer would show some chopage tho
[01:24:27 CEST] <sh4rm4^bnc> is opus better than aac ?
[01:24:55 CEST] <iive> yes, it is improvement over vorbis.
[01:25:02 CEST] <iive> but afaik not mp4 support.
[01:25:16 CEST] <sh4rm4^bnc> ah, but it would work with mkv ?
[01:25:27 CEST] <iive> i guess.
[01:25:56 CEST] <sh4rm4^bnc> "-c:a opus" ?
[01:26:29 CEST] <sh4rm4^bnc> btw, if anyone knows a good subtitle website...
[01:26:32 CEST] <iive> that's still experimental
[01:26:50 CEST] <iive> but it works. you could try libopus
[01:27:01 CEST] <iive> too
[01:27:12 CEST] <sh4rm4^bnc> that would mean -c:a libopus ?
[01:28:05 CEST] <iive> yeh, external codec libraries use name starting with lib
[01:28:27 CEST] <sh4rm4^bnc> ok, thanks
[01:29:58 CEST] <CoreX> sh4rm4^bnc what kind of subtitle website like one where you can download a tool to extract subtitles or one that has already made subtitle files
[01:30:48 CEST] <sh4rm4^bnc> the latter
[01:31:47 CEST] <sh4rm4^bnc> tehre's www.opensubtitles.org which is OK, but its lacking some stuff
[01:32:54 CEST] <sh4rm4^bnc> and then there's subdb where you need some kind of tool to access it, or drag and drop your videos into the browser (!)...
[01:35:06 CEST] <CoreX> yeah subdb matches the hash up with the file
[03:07:55 CEST] <cryptodechange> Does xerror during an transcode only exit on an error with the source? or does it check frames for errors on the output too?
[03:10:40 CEST] <c_14> There's a few places scattered in the codebase that check xerror and if set exit
[03:10:48 CEST] <c_14> It mainly just turns several warnings into errors
[03:11:00 CEST] <c_14> There's no additional sanity checks added
[03:11:36 CEST] <DJ-Pyro> who has better drivers for windows?  nvidia or amd GPU
[03:20:12 CEST] <Cracki> for what
[09:04:33 CEST] <gene_> I'm trying to get an epoch timestamp to render on my video and literally nothing has worked
[09:05:02 CEST] <gene_> the issue comes from the %s in the command: text='%{localtime\:%s}':
[09:05:37 CEST] <gene_> if I put %H in there or something it works fine and it spits out the hour
[09:05:52 CEST] <gene_> but it seems like %s is some sort of mystical thing that breaks my recording
[09:06:43 CEST] <squ> do you have syntax format?
[09:07:08 CEST] <gene_> To give an example, this command works great: ffmpeg -f gdigrab -i desktop -vf "drawtext=fontfile=/Windows/Fonts/Arial.ttf: expansion=normal: text='%{localtime\:%a %b %d %Y}': x=10: y=20: fontcolor=black:  fontsize=30" out.mkv
[09:07:37 CEST] <gene_> but if I replace %a %b %d %Y with %s it fails completely -_____________-
[09:11:15 CEST] <gene_> When I put the %s in there it spits out something about side data (    Side data:       cpb: bitrate max/min/avg: 0/0/0 buffer size: 0 vbv_delay: -1), which is what's making me believe %s has some sort of special meaning in ffmpeg
[09:15:18 CEST] <gene_> hmmm... I'll try asking in the devel channel
[09:19:32 CEST] <Eldarko> I know i've asked this question yesterday, but i still can not fine the answer: Could you guys tell me why this line does not save files with name pattern provided? What is missing? Thank you!   ffmpeg -i " + url + " -c copy -f hls -hls_time 10 -hls_flags second_level_segment_duration -use_localtime 1 -use_localtime_mkdir 1 -nostdin -loglevel 16 -hls_segment_filename %Y%m%d/%H/%M/segment_%Y%m%d%H%M%S_%%010t.ts -method PUT http://12
[13:02:11 CEST] <chamath> Hi guys, I'm trying to add a wipe transition between videos like 12th second of this video.
[13:02:14 CEST] <chamath> https://www.youtube.com/watch?v=SJ1CcaawxmA
[13:03:30 CEST] <chamath> I have tried following command
[13:03:53 CEST] <chamath> https://thepasteb.in/p/y8h656B8OXjTO
[13:04:23 CEST] <chamath> Bit it does not have that gradient kind of effect to wipe line
[13:04:30 CEST] <chamath> Any ideas?
[14:20:36 CEST] <lavalike> do you happen to know if there's any ffmpeg-supported aac encoder that will encode to 2.1? FR+FL+LFE (the source is EAC3 5.1 640k in my case)
[14:21:23 CEST] <lavalike> both -c:a aac and -c:a libfdk_aac create 3 channels in the shape of FL+C+FR in my tests, but maybe I'm missing something
[14:22:23 CEST] <lavalike> I think it might be impossible for fdk given this https://github.com/nu774/fdkaac/blob/master/README#L298-L307 haven't found evidence for the native encoder though
[14:24:21 CEST] <atomnuker> lavalike: aac has crap support for non-standard (according to the 1997 specs) channel layouts
[14:27:19 CEST] <lavalike> ah, hmpf!
[14:28:17 CEST] <lavalike> I will try to test the other supported layouts, I think 3.1 could do the job, do you have other suggestions? maybe a different codec?
[14:32:00 CEST] <atomnuker> yes, vorbis or opus ought to work
[14:33:18 CEST] <atomnuker> (actually only opus would work, and you'll need to set -mapping_family 255)
[14:51:20 CEST] <lavalike> ah that's an interesting idea
[14:56:15 CEST] <lavalike> it's even being supported by bleeding edge safari, neat
[15:03:56 CEST] <kerio> yay opus
[15:15:04 CEST] <lavalike> this test opus audio track is hilarious https://hpr.dogphilosophy.net/test/
[15:26:15 CEST] <lavalike> Audio: opus (libopus) ([255][255][255][255] / 0xFFFFFFFF), 48000 Hz, 2.1, flt, 384 kb/s (default) exciting
[15:44:33 CEST] <certaindestiny> Hi all, I need a little help adding a watermark at the top right corner of my command syntax. Anybody able to help in this regards?
[15:44:35 CEST] <certaindestiny> https://pastebin.com/UmVQDHee
[16:06:09 CEST] <thebombzen> certaindestiny: you don't actually use the watermark.png input at all
[16:06:37 CEST] <thebombzen> also certaindestiny consider using the vstack and hstack filters rather than an overlay chain. much faster
[16:06:41 CEST] <thebombzen> !filter vstack
[16:06:41 CEST] <nfobot> thebombzen: http://ffmpeg.org/ffmpeg-filters.html#vstack
[16:06:43 CEST] <thebombzen> !filter hstack
[16:06:43 CEST] <nfobot> thebombzen: http://ffmpeg.org/ffmpeg-filters.html#hstack
[16:08:34 CEST] <certaindestiny> @thebombzen @nfobot. I am not familiar with that filter. How can i transform my current command to use hstack and vstack and also include a watermark?\
[16:15:09 CEST] <Dark-knight> quick question what is deinterlace?
[16:15:43 CEST] <durandal_1707> certaindestiny: read docs, stack filters take same dimensions inputs
[16:15:58 CEST] <durandal_1707> Dark-knight: use google
[16:16:32 CEST] <Dark-knight> ...aright, I had irc open, so I thought I'd ask
[16:17:40 CEST] <certaindestiny> @durandal_1707, If i was able to understand the exact syntax from the doc ffmpeg provided i would not have asked the question.....
[16:25:33 CEST] <durandal_1707> certaindestiny: your inputs are of same size?
[16:25:47 CEST] <certaindestiny> For this particular instance, yes
[16:28:06 CEST] <certaindestiny> I think i finally found something that might explain it. Please correct me if i am wrong. I basically need to create 2 vstacks, (or hstack) and stack those to hstacks?
[16:28:50 CEST] <durandal_1707> just use [0:v][1:v]hstack[a],[2:v][3:v]hstack[b],[a][b]vstack....
[16:29:39 CEST] <certaindestiny> Alright great, and how would you go about adding the watermark to the top right cornor of the entire output?
[16:36:26 CEST] <certaindestiny> @durandal_1707 When using the command you provided it gives me the error,  Unable to find a suitable output format for
[16:36:56 CEST] <certaindestiny> Has this been released in ffmpeg3.3.3 ?
[16:43:52 CEST] <certaindestiny> Never mind, Fixed that. Its a bad idea to make a typo
[17:09:59 CEST] <sh4rm4^bnc> is there a flag to extract and re-embed subs when re-encoding ?
[18:09:53 CEST] <fede_> Hello
[18:10:26 CEST] <fede_> I have a camera ip and i convert with ffmpeg to mjpeg
[18:11:08 CEST] <fede_> I can see my cam in localhost on http://localhost:8090/monitoring1.mjpg
[18:11:16 CEST] <fede_> But i can´t see in external ip
[18:11:32 CEST] <fede_> I try ACL allow all in config file but doesnt work
[18:11:42 CEST] <fede_> Any one can help me?
[18:14:04 CEST] <SpeakerToMeat> I'm using a '-filter_complex "[1:a][2:a][3:a][4:a][5:a][6:a]amerge=inputs=6[aout]" -map "[aout]"' but then of course I can't do a "-c:a copy", is there any way to mix these together without having them recoded and specifying to use the original file parameters?
[18:23:54 CEST] <SpeakerToMeat> fede_: So you're using ffmpeg as http server¡
[18:23:56 CEST] <SpeakerToMeat> ?
[18:24:14 CEST] <SpeakerToMeat> via -listen ?
[18:26:40 CEST] <SpeakerToMeat> fede_: If so, afaik from what I see the ip it binds to is the one you specify in the url at the parameters, so if your specified url is http://localhost:8090/monitoring1.mjpg in the command line, it'll bind to localhost and not be accessible formt he outside, change localhost to your ip address to bind to it
[18:34:16 CEST] <kepstin> SpeakerToMeat: to change the number of channels in an audio stream with a compression codec requires recompression with most codecs. No real way around that.
[19:21:33 CEST] <cryptodechange> any options for logging the output on ffmpeg? 2> /path/to/out.txt hides the terminal output
[19:29:52 CEST] <relaxed> cryptodechange: ffmpeg 2>&1 | tee ffmpeg.log
[19:31:27 CEST] <cryptodechange> ty!
[19:38:22 CEST] <furq> cryptodechange: -report also works but it forces -v verbose
[20:05:43 CEST] <cryptodechange> aha, it makes a huge list in the log of each frame, I suppose I can't really avoid that?
[20:14:58 CEST] <rjp421> "Too many packets buffered for output stream 0:1."   "2 frames left in the queue on closing"
[20:19:57 CEST] <rjp421> ^ using the latest linux x64 static binary, cmd and log: https://pastebin.com/raw/M6TBEtYS
[20:20:38 CEST] <rjp421> i have to keep running the cmd and eventually it will start encoding
[20:29:45 CEST] <SpeakerToMeat> I have something strange, I'm muxing video and audio into a quicktime wrapper (prores 422 video, pcm audio)... the original video is 24 fps, 24 tbr tbn and tbc, but the output video (which is desynced to the audio that should fit perfectly), is 24 fps and tbr, but tbn and tbc are 12288, why would this change? could it be affecting the sync? how do I avoid it?
[20:39:51 CEST] <SpeakerToMeat> if I specify -r 24 for the video stream input, the output stream (this it all -c:v copy) then tbc reads 24 on the output too, but tbn is still at 12288
[22:04:12 CEST] <paveldimow> Hi, ffmpeg does not support amf0, is there any way I can "extract/convert/read" this data? Maybe any other tool ?
[22:12:47 CEST] <mightyiam> What would make pngstream -> vp9 very fast, please? For example, what is faster, if the png is compressed or not? Specifically, my pngs have very little info in them. Primarily some black lines...
[22:23:00 CEST] <kepstin> mightyiam: the slowest part of that is probably the vp9 encoding; play with encoder settings before worrying about the png decoding speed
[22:23:42 CEST] <mightyiam> kepstin, thanks. I have played with encoder settings quite a bit. Any tips?
[22:24:41 CEST] <kepstin> mightyiam: show us what you have right now?
[22:25:28 CEST] <mightyiam> kepstin, https://pastebin.com/wXVkh7SX
[22:25:33 CEST] <kepstin> mightyiam: general ideas are make sure you have a recentish git build of libvpx so you have the new multithreading code, and tweak the -speed option
[22:27:13 CEST] <kepstin> (-cpu-used is an alias for -speed, iirc)
[22:28:06 CEST] <kepstin> also, your options list is kinda weird there - in most languages, you'll need to have the parameter and value in separate strings so it'll be parsed correctly
[22:28:19 CEST] <kepstin> depends how you're constructing the ffmpeg command tho
[22:29:11 CEST] <mightyiam> This is through https://github.com/fluent-ffmpeg/node-fluent-ffmpeg
[22:29:56 CEST] <furq> you probably want to use row-mt if you're using a recent enough libvpx
[22:29:59 CEST] <mightyiam> I bet I have an oldish libvpx
[22:30:01 CEST] <furq> that's supposed to give a big speed boost
[22:30:26 CEST] <furq> or rather it's supposed to actually utilise all your cores
[22:30:41 CEST] <mightyiam> 1.6.1
[22:31:21 CEST] <mightyiam> Seems to be latest release
[22:33:23 CEST] <Asuran> thats stable release
[22:33:29 CEST] <Asuran> git means newer
[22:33:34 CEST] <Asuran> but unstable perhaps
[22:33:39 CEST] <Asuran> depends on project
[22:34:17 CEST] <furq> yeah row-mt still needs a git build
[22:34:25 CEST] <furq> there's not been a new release since january
[22:34:33 CEST] <furq> i figured it'd have made it into a release by now
[22:34:50 CEST] <furq> if you're actually getting 100% utilisation then i don't think it really matters
[22:35:16 CEST] <furq> but tile-columns is legendarily bad at multithreading
[22:35:34 CEST] <Asuran> will av1 be same?
[22:35:35 CEST] <Asuran> i fear it
[22:38:44 CEST] <mightyiam> Thank you guys
[22:51:25 CEST] <furq> Asuran: probably at first but the vpx guys know it's a problem
[22:51:36 CEST] <furq> there just isn't much impetus to fix it because vp9 is basically just for youtube
[00:00:00 CEST] --- Thu Aug 17 2017