[Ffmpeg-devel-irc] ffmpeg.log.20170721

burek burek021 at gmail.com
Sat Jul 22 03:05:02 EEST 2017


[00:34:15 CEST] <cryptodechange> Wow, trying to encode some 1080p anime
[00:34:31 CEST] <cryptodechange> crf=17, over twice the bitrate of some films I've done
[00:34:34 CEST] <cryptodechange> bluray source
[00:34:54 CEST] <cryptodechange> It's a blu ray release of an old release, very grainy
[00:34:57 CEST] <cryptodechange> Probably upscaled
[00:35:03 CEST] <JEEB> well yea, grain kills the man :P
[00:35:07 CEST] <JEEB> (and encoders)
[00:35:25 CEST] <cryptodechange> So, filter?
[00:35:26 CEST] <JEEB> also that crf is low for most 1080p animoo I've poked my fingers at
[00:35:38 CEST] <JEEB> no, don't filter other than downscale if it's an upscale
[00:35:47 CEST] <JEEB> unless there's something really wrong with it
[00:35:48 CEST] <dystopia_> add -tune animation
[00:35:54 CEST] <JEEB> dystopia_: lol
[00:36:01 CEST] <cryptodechange> I'm pretty much using the animation preset
[00:36:07 CEST] <JEEB> I would have expected tune grain or something 'cause he said it was grainy
[00:36:08 CEST] <JEEB> :P
[00:36:21 CEST] <cryptodechange> Tune grain would've preserved the grain though?
[00:36:28 CEST] <JEEB> that's what it's doing anyways
[00:36:32 CEST] <c_14> tune grain preserves more grain
[00:36:40 CEST] <JEEB> anyways, re-calibrate your CRF range and downscale if it's upscaled
[00:36:45 CEST] <cryptodechange> = more bitrate, which is what I want to avoid.
[00:36:48 CEST] <JEEB> no need to keep it at 1080p if the content's not that
[00:36:58 CEST] <cryptodechange> It's the DBZ blurays, so most definitely upscaled
[00:37:30 CEST] <cryptodechange> so I want to downscale it to 480p?
[00:37:51 CEST] <c_14> that's probably overkill
[00:38:08 CEST] <cryptodechange> Raw, they are ~3.8gb
[00:38:15 CEST] <dystopia_> first aired in 1996
[00:38:24 CEST] <dystopia_> i wonder what jpn broadcast standard was then
[00:38:38 CEST] <cryptodechange> crf=17 was producing up to 25mbps
[00:38:40 CEST] <JEEB> well yea, blu-rays are gonna be what they are unless the mastering studio wanted to save space (and put more stuff on a single disc)
[00:39:02 CEST] <JEEB> and this is what encoders actually do, you start the journey into looking up the actual detail level in that thing :|
[00:39:18 CEST] <JEEB> most likely someone else has already done that but you need to base your thing on good enough data
[00:39:20 CEST] <cryptodechange> 10 episodes at 3.8gb each
[00:39:44 CEST] <JEEB> that doesn't fucking matter as I noted. that sounds like the most usual BD bit rate I've seen :P
[00:39:47 CEST] <cryptodechange> If I stored them all raw, it'd be 1TB :D
[00:39:47 CEST] <JEEB> ~40mbps
[00:40:41 CEST] <JEEB> anyways, time for you to get your lazy ass and do some research and experimentation. and to be honest, if it was analog art then if it was re-scanned it could be surpriisngly high. but if it wasn't, then you've just got better bit rate compared to DVDs (for the source) and you can downscale it
[00:41:11 CEST] <JEEB> there's also things that can help you guesstimate the actual pre-upscale resolution if it was upscaled in a sane way
[00:41:39 CEST] <dystopia_> i guess 640x480 would be a good place to go
[00:42:14 CEST] <cryptodechange> The quality is not bad for 1080p
[00:42:15 CEST] <cryptodechange> https://imgur.com/a/YsweE
[00:42:38 CEST] <JEEB> dystopia_: it really depends on how much was actually remastered (and how much was filtered to hell)
[00:42:45 CEST] <cryptodechange> upscaled 1080p*
[00:42:56 CEST] <JEEB> dat blur
[00:43:23 CEST] <JEEB> anyways, have fun properly thinking it out :P
[00:43:24 CEST] <cryptodechange> Though around the outlines
[00:43:30 CEST] <c_14> It's not even that grainy
[00:43:43 CEST] <cryptodechange> It's the motion grain
[00:54:38 CEST] <kepstin> from 1996? that predates most digimation stuff in japan, it might actually be a new scan from film
[00:54:52 CEST] <kepstin> in which case, it's actually native hi def :/
[00:55:29 CEST] <kepstin> weird that they cropped it to 16:9 ratio tho, the original video would have been 4:3
[00:56:00 CEST] <c_14> 16:9 is in, 4:3 is way old news
[00:56:09 CEST] <Threads> dont you know ?, 16:9 is the new standard for people who dont know anything
[00:56:39 CEST] <kepstin> it's kind of sad that the only way to store HD 4:3 stuff on bluray is to pillarbox it
[00:58:32 CEST] <Threads> just leave it as nobody will care
[00:58:40 CEST] <Threads> fuck just give me the 4:3 dvd
[00:59:08 CEST] <c_14> the thing with the bluray is that someone could have potentially if they reeeally felt like it rescanned the film in high definition
[00:59:14 CEST] <cryptodechange> I should hold back in telling you guys I'm cropping it further to match my 21:9 monitor  >_>
[00:59:23 CEST] <cryptodechange> because I like staring at eyes only
[01:00:03 CEST] <cryptodechange> So in the event of being rescans/native hd, would I want to try and remove the grain instead of downscale?
[01:00:06 CEST] <furq> kepstin: emphasis on "might"
[01:00:15 CEST] <furq> http://i.imgur.com/AeA6K7N.png
[01:00:25 CEST] <furq> check out this pristine upscale of a show from 1993
[01:00:38 CEST] <c_14> oh my good
[01:00:39 CEST] <kepstin> I personally own a few really nice looking HD anime blu-rays that were rescanned from film, they're gorgeous.
[01:00:42 CEST] <furq> and yes, of course it's pan&scan
[01:00:44 CEST] <cryptodechange> his face is an accurate representation
[01:00:49 CEST] <Threads> give it 10years we will see film from 1980 remastered to 4k
[01:00:55 CEST] <furq> we're already seeing that
[01:01:03 CEST] <furq> film from 1980 is no problem, you can just rescan it
[01:01:32 CEST] <Threads> furq the simpsons on fxx was cropped like mad
[01:01:37 CEST] <kepstin> any animation done digitally from the mid-2000s is just gonna be awful, it's natively SD resolution and there's no way you can make it look good.
[01:01:41 CEST] <furq> 35mm film will theoretically go way behind 8k
[01:01:52 CEST] <furq> assuming you still have a good master ofc
[01:01:54 CEST] <c_14> beyond?
[01:01:57 CEST] <furq> er, yeah
[01:02:00 CEST] <furq> im good at typeing
[01:02:12 CEST] <cryptodechange> What's your opinion on this, kepstin? https://i.imgur.com/8bYRB65.jpg
[01:02:44 CEST] <cryptodechange> My eyes are untrained, but is there blocking around the outlines?
[01:02:47 CEST] <kepstin> cryptodechange: hard to tell, if that is an upscale, it's very well disguised
[01:02:59 CEST] <furq> please learn how to use your video player's screenshot function
[01:03:27 CEST] <furq> but yeah that looks like it could be a rescan
[01:03:30 CEST] <cryptodechange> Sorry, it was a habit to prtscr and paste into imgur
[01:03:46 CEST] Action: cryptodechange even pastes into paint to crop at times
[01:04:09 CEST] <kepstin> there's some telltale smearing in some of the lines around the eyes that point to filtering, but it's hard to say
[01:04:25 CEST] <furq> yeah the lines generally look a bit unnatural
[01:04:29 CEST] <cryptodechange> The somewhat grainy blue sky is super noticeable when the video is playing
[01:04:42 CEST] <cryptodechange> being not so 'somewhat'
[01:04:59 CEST] <kepstin> i'm not familiar with that era of dragonball, but it might have actually been done on 8mm rather than 16mm film, which could explain the thicker lines and extra grain.
[01:05:42 CEST] <cryptodechange> Some of the 'distance' shots are humourous, as you can clearly see it's a still image when they talk
[01:06:10 CEST] <kepstin> yeah, that's a fun thing about these anime rescans - you can see detail that would have been invisible on an SD television.
[01:06:16 CEST] <kepstin> as so was left out
[01:06:26 CEST] <Threads> http://www.kanzenshuu.com/production/kai-remastering/
[01:06:54 CEST] <kepstin> ah, so it was 16mm
[01:07:22 CEST] <cryptodechange> This is from the non-kai bluray, but I assume it was a similar process, without the reworking of any of the content
[01:07:46 CEST] <furq> In fact, the first 194 episodes of the original Dragon Ball Z TV series were condensed down into a mere 98 episodes, resulting in nearly 100 episodes worth of material being cut out.
[01:07:52 CEST] <cryptodechange> Maybe the DBZ blurays were the precursor to kai, they released them to ensure extra funding
[01:07:56 CEST] <furq> 100 episodes worth of two dudes going AAAAAAAAHHHHHHH and glowing
[01:08:31 CEST] <kepstin> my favourite modern film rescan is the 1960s batman tv series, of course
[01:08:35 CEST] <cryptodechange> I want the AHHHHHHHHHHHHHH
[01:08:48 CEST] <kepstin> that is absolutely glorious on bluray.
[01:09:14 CEST] <cryptodechange> I've still got the interlaced/telecined 90s Batman and X-men animated shows to encode
[01:09:44 CEST] <furq> So its no surprise why certain scenes were edited to remove some blood or male genitalia.
[01:09:48 CEST] <furq> well this is no good any more
[01:09:48 CEST] <Threads> ohh wow they left the batman in 4:3
[01:10:48 CEST] <cryptodechange> so now to look into degrain filters and try that out
[01:10:48 CEST] <kepstin> of course. if they cropped it you wouldn't be able to read all the tiny sign gags in the background
[01:11:24 CEST] <kepstin> (the background gags in that show are hilarious, particularly since many would have been invisible in the original broadcast)
[01:15:29 CEST] <furq> and to tie it all together
[01:15:30 CEST] <furq> https://www.youtube.com/watch?v=JJdXMrFRDGk
[03:01:44 CEST] <cryptodechange> This is a pretty good resource for denoising
[03:01:45 CEST] <cryptodechange> http://www.dirk-farin.net/projects/nlmeans/index.html
[03:05:03 CEST] <leif> Does anyone know what format I would use if I wanted to use the rawvideo video codec and s16 audio codec?
[03:05:09 CEST] <furq> nut
[03:05:23 CEST] <leif> There seems to be a rawvideo format, but for audio the closest thing I can find is the wv format.
[03:05:36 CEST] <furq> do you mean format or codec
[03:05:48 CEST] <furq> the audio codec for s16 pcm is pcm_s16le
[03:06:07 CEST] <furq> or be, but nobody uses that any more
[03:06:47 CEST] <leif> furq: I mean, I know the codec I want is pcm_s16le, but I am trying to figure out what format I should put the codec into.
[03:08:03 CEST] <leif> Basically, I'm trying to convert a file to rgb24 (which looks like it needs to be in the rawvideo codec), and s16 audio, which is thepcm_s16le format.
[03:08:09 CEST] <leif> err...codec*
[03:08:20 CEST] <leif> But I'm having trouble figuring out what format that can go into.
[03:08:45 CEST] <furq> if you mean both together then nut
[03:09:09 CEST] <drv> you can use the s16le format to get raw audio
[03:09:30 CEST] <leif> drv: Cool, thanks.
[03:09:43 CEST] <leif> furq: Honestly together or seperate is fine.
[03:10:06 CEST] <leif> But ya, drv answered the question. (I was searching for formats with raw in the name...)
[03:10:19 CEST] <leif> But ya, the s16le one works great. :)
[03:10:24 CEST] <drv> yeah, not sure why there's a single rawvideo and lots of raw PCM formats
[03:11:59 CEST] <leif> drv: Ya. I mean, I get that there's a lot of different types of 'raw' audio...but there are also lots of different types of 'raw' video too. :D
[03:12:03 CEST] <leif> Anyway, thanks again.
[04:12:29 CEST] <redrabbit> hi
[04:12:39 CEST] <redrabbit> im trying to transcode my dvbviewer streams to hevc / heaac instead of avc/mp3 defaults
[04:13:14 CEST] <redrabbit> http://dpaste.com/3VJ3SS2.txt here is the config line, there must be something wrong with it; probably some option that only works with the original codecs ?
[04:14:54 CEST] <furq> what's the error
[04:15:35 CEST] <furq> i'm guessing it's something to do with -profile baseline and -tune film because neither of those exist in x265
[04:16:04 CEST] <redrabbit> that's a good question, i can't/dont know how see the error because its ran by a service
[04:16:16 CEST] <redrabbit> ok, im going to try to remove this
[04:16:56 CEST] <furq> also you probably won't get very good results from changing the video codec
[04:17:09 CEST] <furq> unless you have a fast enough cpu to run x265 in realtime with a slow profile
[04:17:48 CEST] <furq> x265 doesn't really outperform x264 at the same encoding speed
[04:18:19 CEST] <furq> (or the same cpu utilisation for realtime input)
[04:18:23 CEST] <redrabbit> its for low bw application
[04:19:02 CEST] <redrabbit> the 1st part i wanted to get right was the he-aac
[04:19:23 CEST] <furq> well that bit will definitely make a difference
[04:21:21 CEST] <redrabbit> furq: the h265 part works now
[04:21:31 CEST] <redrabbit> you were right about -profile baseline and -tune film
[04:21:50 CEST] <redrabbit> there must be something with the aac part as well
[04:22:04 CEST] <redrabbit> i had to revert to lame to try the hevc
[04:22:09 CEST] <furq> if you didn't build ffmpeg yourself then you don't have fdk-aac
[04:22:27 CEST] <furq> it's not gpl compatible so binaries with it aren't distributable
[04:23:01 CEST] <redrabbit> i got it ther https://ffmpeg.zeranoe.com/builds/win64/shared/?C=M&O=A
[04:23:21 CEST] <furq> yeah that doesn't have fdk
[04:23:33 CEST] <redrabbit> so there is no way to get it without diy compile
[04:23:35 CEST] <furq> and the builtin aac encoder doesn't support he-aac
[04:23:41 CEST] <furq> although it should still be better than lame at 64k
[04:23:59 CEST] <redrabbit> certainly
[04:24:30 CEST] <redrabbit> compling it myself is a bit out of my league i guess
[04:24:43 CEST] <furq> i take it you can't use opus
[04:24:48 CEST] <redrabbit> i wish there was builds somewhere
[04:25:01 CEST] <redrabbit> i dont know what is opus
[04:25:01 CEST] <furq> blame fraunhofer for their stupid license
[04:25:12 CEST] <furq> opus is a better audio codec that outperforms he-aac at 64k
[04:25:17 CEST] <furq> but it's not as widely compatible
[04:25:20 CEST] Action: redrabbit blames
[04:25:32 CEST] <redrabbit> i need a TS compatible format
[04:25:42 CEST] <furq> opus will mux into mpegts iirc
[04:25:49 CEST] <furq> but whether your devices will decode it is another issue
[04:26:05 CEST] <redrabbit> oh its for a pc
[04:26:32 CEST] <furq> that should probably work then
[04:26:43 CEST] <redrabbit> i have a very weak BW for a couple week vacation and im trying to enhance quality i get
[04:26:46 CEST] <furq> -c:a libopus
[04:27:57 CEST] <redrabbit> so i replace -acodec libmp3lame with -c:a libopus ?
[04:28:01 CEST] <furq> right
[04:30:15 CEST] <redrabbit> it must not like some of the options, fails
[04:33:13 CEST] <redrabbit> aac works though
[04:33:46 CEST] <redrabbit> its an improvement
[04:34:29 CEST] <furq> 64k is plenty for lc-aac if you're downmixing to mono
[04:36:29 CEST] <redrabbit> yep its mono
[04:36:40 CEST] <redrabbit> a big step from 48k mp3
[04:36:41 CEST] <redrabbit> :)
[04:37:08 CEST] <redrabbit> (the original quality preset)
[04:52:05 CEST] <cryptodechange> oh wow
[04:52:27 CEST] <cryptodechange> is it just me or does nlmean work wonders?
[04:52:42 CEST] <cryptodechange> https://imgur.com/a/scY02
[04:53:07 CEST] <cryptodechange> using -vf nlmeans=s=6
[05:03:31 CEST] <sikilikis> hi everyone. quick question. If it better to compile ffmpeg with --enable-librtmp or to just use the native rtpm?
[05:04:51 CEST] <sikilikis> I've been using the --enable-librtmp option and it works, but I'm using ffmpeg to stream 24/7 to youtube and the stream will randomly get disconnected
[05:05:43 CEST] <sikilikis> forget the error but its some generic rtmp error. It's like writeN error 32 or something. I've no idea if using the native rtmp will be better, or perhaps trying to compile the latest version of librtmp-dev
[05:05:51 CEST] <redrabbit> sweet. after some optimisations i get quite a good picture with the high efficiency codecs
[05:05:54 CEST] <PixelPerfect> ...Do you mean rtmp?
[05:05:59 CEST] <PixelPerfect> sikilikis
[05:06:05 CEST] <sikilikis> yeah. Did I spell it wrong?
[05:06:15 CEST] <PixelPerfect> >rtpm
[05:06:17 CEST] <redrabbit> still need to get what failed with opus
[05:06:23 CEST] <sikilikis> oh oops. But yes
[05:06:28 CEST] <redrabbit> some extra audio quality is nice
[05:08:19 CEST] <sikilikis> anyway yes I'll randomly get like rtmp send error 104, send error 32, and send error 9
[05:08:33 CEST] <sikilikis> stream might go days before it gets that error
[05:08:37 CEST] <sikilikis> sometimes hours
[05:08:55 CEST] <sikilikis> I had to wrap my ffmpeg command inside a simple bash script with a loop just so it can recover
[05:10:11 CEST] <furq> sikilikis: librtmp isn't really developed any more afaik
[05:10:29 CEST] <furq> and the native rtmp support works just fine for me, although i'm not streaming 24/7
[05:10:48 CEST] <PixelPerfect> Yeah, I'd wonder if that's just net hiccups or DHCP release renew
[05:11:06 CEST] <sikilikis> dhcp renew?
[05:11:14 CEST] <furq> cryptodechange: denoising ought to give great results on that source because there's not really any fine detail for it to confuse with grain
[05:11:35 CEST] <PixelPerfect> sikilikis as in, your router automatically reassigning IP addresses when they expire
[05:11:39 CEST] <furq> nlmeans is pretty good in general though
[05:12:20 CEST] <sikilikis> hmm. actual ip adress right? not the lan ip?
[05:12:27 CEST] <furq> no, the lan ip
[05:12:31 CEST] <PixelPerfect> or both
[05:12:34 CEST] <PixelPerfect> but likely lan
[05:12:37 CEST] <sikilikis> oh. No that hasn't changed
[05:12:59 CEST] <PixelPerfect> Your DHCP lease can still expire even if you're reassigned the same IP
[05:13:00 CEST] <sikilikis> I didn't set up a static ip, but I've been using the same ip to ssh to
[05:13:14 CEST] <sikilikis> ah I see. You think setting a static ip might fix things?
[05:13:15 CEST] <furq> yeah you probably want to set up a static lease regardless
[05:13:17 CEST] <furq> or a static ip
[05:13:32 CEST] <sikilikis> I can try that then. thanks
[05:13:35 CEST] <PixelPerfect> It's one less variable in the equation
[05:13:54 CEST] <sikilikis> would you also recommend using the native rtmp? sounds like I should if librtmp isnt in development
[05:14:00 CEST] <PixelPerfect> Honestly, I think programs like OBS just have fallover wrapping already, which is why you don't have to bother with it
[05:14:02 CEST] <furq> it's worth a try
[05:14:10 CEST] <PixelPerfect> Go for it
[05:14:22 CEST] <furq> i know librtmp causes a couple of option parsing bugs that builtin doesn't have
[05:14:29 CEST] <furq> no idea about the actual meat of it though
[05:16:12 CEST] <cryptodechange> where can I find more information regarding a specific filter? Mostly in regards to ranges and default values
[05:16:34 CEST] <cryptodechange> ie. nlmean, I am not sure what the default strength value is, or its range
[05:18:19 CEST] <furq> http://vpaste.net/UX337
[05:19:27 CEST] <furq> usually also https://ffmpeg.org/ffmpeg-filters.html but not all the filters are fully documented on there
[05:26:00 CEST] <cryptodechange> ty!
[07:10:20 CEST] <kepstin> cryptodechange: 'ffmpeg -h filter=nlmeans' will show you all the option defaults and ranges
[07:10:40 CEST] <kepstin> er, that's what furq pasted
[07:10:41 CEST] <kepstin> ok
[08:24:38 CEST] <dynek> hello all!
[08:25:09 CEST] <dynek> I'm trying to read an rtsp stream into an flv file using vaapi but it ends-up empty: https://pastebin.com/LNtD4w4K
[08:25:21 CEST] <dynek> any hint will be much appreciated - I've been trying to get this va-api thingie work for quite some time :-(
[10:00:15 CEST] <thebombzen> wait, I'm confused about something
[10:00:34 CEST] <thebombzen> if Matroska only has a timebase of 1k, how do you put 24 fps, 30 fps, 60 fps etc. video inside it correctly?
[10:00:57 CEST] <thebombzen> after encoding a 60 fps video and putting inside matroska (using lavf's muxer), ffprobe reports 60 fps. But how does that work with a 1k timebase?
[10:03:32 CEST] <rsegecin> HI
[10:03:35 CEST] <rsegecin> ops
[10:03:39 CEST] <rsegecin> Hello
[10:06:36 CEST] <rsegecin> is anyone out there?
[10:07:28 CEST] <thebombzen> I'm sure someone is here
[10:07:39 CEST] <thebombzen> if you have a question, you should just ask it though
[10:07:58 CEST] <thebombzen> if nobody knows the answer who's on at the moment, nobody will say anything
[10:08:05 CEST] <thebombzen> if someone knows, and sees it, they'll respond
[10:08:17 CEST] <rsegecin> thanks
[10:08:19 CEST] <thebombzen> but if you just say "hello" most people aren't nice enough to say "hello" back or to give you this little bit of info
[10:09:44 CEST] <rsegecin> I'm trying to run this code:
[10:09:45 CEST] <rsegecin> loopback-capture | ffmpeg -i pipe:0 -f s16be -ar 44100 -acodec pcm_s16be pipe:1
[10:10:24 CEST] <rsegecin> loopback-capture captures and prints in the stdout the chunk of a wav file
[10:10:25 CEST] <thebombzen> what's wrong?
[10:10:27 CEST] <thebombzen> oh
[10:11:02 CEST] <thebombzen> what seems to be the issue? (also, fyi, you can use -i - to read from standard in and - to write to standard out. you don't need to use pipe:0 or pipe:1)
[10:11:39 CEST] <thebombzen> also your current command looks like it's printing raw PCM samples to the terminal on standard output
[10:11:45 CEST] <thebombzen> are you sure you want to do that? or is that not the correct command?
[10:11:57 CEST] <rsegecin> but ffmpeg doesn't do anything because I think "loopback-capture" doesn't spit the beggining of the header wav file
[10:12:15 CEST] <thebombzen> what do you mean by "doesn't do anything"
[10:12:28 CEST] <thebombzen> do you mean it sits there and hangs or do you mean it fails to detect the input file format?
[10:13:07 CEST] <thebombzen> I can't help you without the exact command and full uncut output that is. I can't see what's on your screen
[10:14:09 CEST] <rsegecin> https://pastebin.com/d2EARtPX
[10:14:55 CEST] <rsegecin> by doesn't do anything I mean ffmpeg should print in stdout
[10:15:53 CEST] <rsegecin> what I want that ffmpeg do is to convert a wave file format from type WAVE_FORMAT_EXTENSIBLE to PCM
[10:17:09 CEST] <thebombzen> well if loopback-capture is printing samples without a WAV header, then FFmpeg isn't able to autodetect the format
[10:17:12 CEST] <thebombzen> it looks like this is happening
[10:17:38 CEST] <thebombzen> because if it printed a WAV header, FFmpeg would be able to read that, and autodetect the format of the input.
[10:18:05 CEST] <thebombzen> if loopback-capture is outputting raw samples, you have to tell FFmpeg what sort of stuff it's getting
[10:18:36 CEST] <rsegecin> I think ffmpeg  needs the header but as its a continious stream from my output device I don't have the information of the total size because I don't know wether it's going to stop
[10:19:01 CEST] <thebombzen> if it doens't have the header, it's fine. It just means it cannot auto-detect the stuff like sample rate or sample format, so you have to provide them yourself
[10:19:07 CEST] <thebombzen> try this: loopback-capture | ffmpeg -f s16le -ar 48000 -ac 2 -i - -c copy -f wav -
[10:19:22 CEST] <rsegecin> ok
[10:19:43 CEST] <thebombzen> what this does is it has ffmpeg read signed 16-bit little endian samples from standard in, at 48 kHz in stereo, attaches a wav header, and then writes the WAV to standard out.
[10:20:24 CEST] <rsegecin> hmm I need PCM in the stdout
[10:20:34 CEST] <thebombzen> Well if you have pcm in the input then why do you need FFmpeg
[10:20:53 CEST] <rsegecin> shouldn't I need to add pipe:0 and pipe:1 as well?
[10:21:11 CEST] <thebombzen> no, the -i - tells ffmpeg to use stdin, you don't need -i pipe:0
[10:21:21 CEST] <thebombzen> similarly writing to "-" tells it to write to stdout, no need for pipe:1
[10:21:33 CEST] <thebombzen> but loopback-capture is outputting raw PCM samples. if you need raw pcm samples, why do you need FFmpeg at all?
[10:21:49 CEST] <thebombzen> is there any reason you cannot pipe loopback-capture directly to the program reading the PCM samples?
[10:21:55 CEST] <rsegecin> loopback-capture is outputting chunks of wave file
[10:22:01 CEST] <thebombzen> It definitely isn't.
[10:22:13 CEST] <thebombzen> FFmpeg can read the wave header
[10:22:38 CEST] <thebombzen> now,  you can force FFmpeg to interpret the input as a WAV
[10:22:47 CEST] <thebombzen> add "-f wav" before the -i
[10:23:02 CEST] <thebombzen> so it's ffmpeg -f wav -i - (or as you like it, ffmpeg -f wav -i pipe:0)
[10:23:46 CEST] <rsegecin> ok I'll try the command you showed me before
[10:25:21 CEST] <thebombzen> the command I said before outputs wav
[10:26:07 CEST] <thebombzen> rsegecin: if you're sure loopback-capture is outputting chunked wav and you need pcm, try this:
[10:26:30 CEST] <thebombzen> loopback-capture | ffmpeg -f wav -i - -f s16be -
[10:26:41 CEST] <thebombzen> now, this will convert it to signed 16-bit big-endian samples
[10:26:50 CEST] <thebombzen> you probably want little-endian samples
[10:26:55 CEST] <thebombzen> because little-endian is fairly standard
[10:27:06 CEST] <thebombzen> (CD Audio is little-endian, i.e. s16le, not s16be)
[10:27:50 CEST] <rsegecin> can we go on private?
[10:28:32 CEST] <rsegecin> I got an error
[10:28:33 CEST] <rsegecin> [wav @ 0000000001c924a0] invalid start code Pres in RIFF header pipe:: Invalid data found when processing input
[10:29:01 CEST] <thebombzen> no
[10:29:09 CEST] <thebombzen> I'm not going to private message you about an issue in a public channel
[10:29:17 CEST] <thebombzen> if it says "invalid start code" that means it could not read the WAV header
[10:30:58 CEST] <thebombzen> so it's likely writing raw PCM samples
[10:31:07 CEST] <rsegecin> yes, loopback-capture doesn't print the header because it's getting in "real time"
[10:31:38 CEST] <thebombzen> but if loopback-capture is printing raw PCM samples, why do you need FFmpeg?
[10:31:49 CEST] <thebombzen> why can't you just pipe loopback-capture into whatever is reading the PCM samples?
[10:31:56 CEST] <thebombzen> What are you actually trying to do?
[10:32:25 CEST] <rsegecin> hold on may be it's a miss communication
[10:32:31 CEST] <rsegecin> let me explain
[10:35:50 CEST] <rsegecin> loopback-capture is printing on it's stdout chunk of wave data that is getting from my PC output device, it doesn't contain the header of a wav file because as it happens while the song is playing from my pc I don't have the file size that it's going to be
[10:37:38 CEST] <rsegecin> ffmpeg should be able to read the chunk of data and convert it to PCM type and print it on stdout
[10:40:09 CEST] <rsegecin> so as soon I execute this both commands the ffmpeg should be printing PCM data values to me converted from SubFormat generated by loopback-capture
[10:40:35 CEST] <thebombzen> rsegecin: you realize that WAV audio is just raw pcm samples with a header attached to it?
[10:41:02 CEST] <rsegecin> no it's not I thought it was it's not
[10:41:06 CEST] <thebombzen> It definitely is
[10:41:18 CEST] <rsegecin> http://www.onicos.com/staff/iz/formats/wav.html
[10:41:31 CEST] <thebombzen> (usually)
[10:41:35 CEST] <thebombzen> (it can contain other things)
[10:41:46 CEST] <rsegecin> there are various types of wav format
[10:42:20 CEST] <thebombzen> Yes, but 99% of all WAVs are just PCM audio with a header
[10:42:41 CEST] <thebombzen> the audio generated by your loopback capture device will be PCM
[10:44:01 CEST] <Fig1024> I actually implemented C++ loop-back capture from Windows audio playback devices. Most common format is floating point samples, -1.0 to +1.0
[10:44:22 CEST] <rsegecin> ok so I will try to not use ffmpeg and pipe it up with the rest of the system
[10:44:34 CEST] <rsegecin> and see what I've got
[10:44:34 CEST] <thebombzen> Yea. Although you should check the sample format.
[10:44:50 CEST] <thebombzen> Because your loopback capture is probably going to output 16-bit little endian samples
[10:44:57 CEST] <thebombzen> not big-endian
[10:45:02 CEST] <thebombzen> (little-endian is more standard)
[10:45:17 CEST] <thebombzen> which means that if you're expecting big-endian samples, you'll get static
[10:45:30 CEST] <rsegecin> hmmm ok let me check and I'll brb
[10:51:00 CEST] <rsegecin> I get an error sayng "pa_stream_write() failed: Invalid argument"
[10:56:03 CEST] <rsegecin> anyhow do you guys know if there's is a way to create a wav header file without it's content's size, so may be ffmpeg would understand that it's a stream and should convert the audio wav formats on the fly
[10:59:57 CEST] <rsegecin> https://pastebin.com/NNrCudKD
[11:00:50 CEST] <rsegecin> this time I didn't use ffmpeg hopping for the wav content be in PCM type
[11:02:14 CEST] <rsegecin> pulse audio that is playing back on a orange pi connected through ssh
[11:03:56 CEST] <rsegecin> I've this command ffmpeg -i loopback-capture.wav -f s16be -ar 44100 -acodec pcm_s16be pipe:1 | plink -v 192.168.11.5 -l rinaldi -pw star13r "cat - | pacat --playback --format s16be --rate 44100 --volume 30000" that it works
[11:04:23 CEST] <rsegecin> And I can listen
[11:05:17 CEST] <rsegecin> but my wav file now needs to come from my PC output device
[11:07:05 CEST] <Fig1024> so you need loopback from audio playback device, not sure if ffmpeg supports it. I knows Windows audio devices support such an interface, but dunno about other operating systems
[11:08:20 CEST] <rsegecin> I know that ffmpeg doesn't support getting the input from an Windows output device, only in linux
[11:08:51 CEST] <rsegecin> so that's what loopback-capture is doing
[11:09:10 CEST] <Mavrik> hmm
[11:09:15 CEST] <Mavrik> wav by definition needs content size
[11:09:23 CEST] <rsegecin> loopback-capture is generating correctly the wav file but I don't know how to output it as PCM format
[11:09:28 CEST] <Mavrik> As such it's a strange format to stream data in.
[11:09:46 CEST] <rsegecin> hmmmm
[11:10:45 CEST] <Mavrik> Are you sure you don't really just want to send raw PCM data?
[11:10:51 CEST] <Mavrik> Or mux it into something else?
[11:11:42 CEST] <Fig1024> on Windows you can create a DirectShow capture filter that acts as loopback from output device. This directshow filter can then by used by FFMPEG. To create this filter, you go to Windows Sounds -> Recording tab, then enable "Stereo Mix"
[11:12:29 CEST] <rsegecin> I'd like to send PCM data directly from loopback-device but I don't know how, but I think ffmpeg should be able to handle this
[11:14:34 CEST] <rsegecin> Fig1024 I see my two devices there but I don't have the stereo mix feature
[11:17:45 CEST] <Fig1024> right click on the device list (not device item), popup menu should appear with "show disabled devices" - should be checked
[11:19:14 CEST] <rsegecin> yes it's checked
[11:19:35 CEST] <rsegecin> I see that there might be a way of installing it
[11:24:07 CEST] <rsegecin> I'll have to restart the computer
[11:24:12 CEST] <rsegecin> brb
[11:24:29 CEST] <thebombzen> rsegecin: if you're looking to forward pulseaudio over an SSH connection, you should check out Xpra Winswitch
[11:24:47 CEST] <thebombzen> https://xpra.org/
[11:29:08 CEST] <rsegecin> omg stereo mix appeared
[11:29:34 CEST] <rsegecin> so much trouble I went through
[11:29:55 CEST] <thebombzen> [05:24:29] <thebombzen> rsegecin: if you're looking to forward pulseaudio over an SSH connection, you should check out Xpra Winswitch
[11:30:10 CEST] <thebombzen> you missed that b/c reboot probably
[11:30:11 CEST] <rsegecin> I'll try to do the streaming with LineinCode  now
[11:30:23 CEST] <thebombzen> but you should check out Xpra Winswitch if you're looking to forward PulseAudio over an SSH Connection
[11:30:25 CEST] <thebombzen> it can do that
[11:30:42 CEST] <thebombzen> it's much easier than trying to manually capture and stuff
[11:31:23 CEST] <thebombzen> instead, it essentially implements the PulseAudio sound server protocol on the server, and forwards the playing audio through an SSH tunnel to the client rather than send it to the speakers on the server.
[11:33:08 CEST] <rsegecin> hmmm I'm trying to understand
[11:33:14 CEST] <thebombzen> http://xpra.org/
[11:33:43 CEST] <Fig1024> big question is whether you want the audio to go thru speakers or not
[11:34:20 CEST] <rsegecin> ohh is it just like an tem view for linux?
[11:35:02 CEST] <rsegecin> xpra
[11:35:26 CEST] <thebombzen> just go to the home page, they explain it better than I could
[11:35:36 CEST] <thebombzen> Fig1024: if the goal is to play audio through the speakers on the server, then everything we've discussed would have been pointless.
[11:36:15 CEST] <thebombzen> You keep doing that, interjecting into a conversation with random questions and stuff that we've already discussed if you scroll up.
[11:36:17 CEST] <rsegecin> from what I understood it's a remote display
[11:36:36 CEST] <thebombzen> Yea, but it can also forward sound and printing.
[11:36:52 CEST] <thebombzen> So you could play a song on the server and the audio will be forwarded through the client.
[11:36:53 CEST] <rsegecin> I need the audio streamed from windows
[11:37:18 CEST] <thebombzen> you're using pulseaudio....
[11:38:00 CEST] <rsegecin> the device I'm trying to reach is an arm that I can't run some programs, they weren't made to be compile in arm
[11:38:07 CEST] <rsegecin> like spotify
[11:38:19 CEST] <rsegecin> so I need to stream from my pc
[11:38:19 CEST] <thebombzen> well what are you actually trying to do then
[11:40:09 CEST] <rsegecin> Stream from my Windows output device to my ArmBian device to play
[12:01:31 CEST] <rsegecin> I'll create a thread in the forum and post the code in github
[12:01:59 CEST] <rsegecin> I've to go now thank you all anyway
[13:24:15 CEST] <hexhaxtron> How can I convert a 3D Bluray MKV to MP4?
[13:43:51 CEST] <dynek> thebombzen: when you talked about Matroska (around 2h ago) where you mentioning my pastebin?
[13:44:03 CEST] <thebombzen> completely unrelated question
[13:45:17 CEST] <rsegecin> I couldn't make the thread in the forum yet but here it's the code to the project https://github.com/rsegecin/WLStream
[13:45:58 CEST] <rsegecin> see you guys later
[13:46:05 CEST] <thebombzen> speaking of my matroska question, apparently the codec timebase is 1/240
[13:46:19 CEST] <thebombzen> but the format and Opus timebase is 1/1000
[13:46:33 CEST] <thebombzen> how do you have a H.264 timebase of 1/240 inside matroska if it only does powers of 10?
[15:11:47 CEST] <c7j8d9> Is is possible to add --enable-vp9-highbitrate to and existing ffmpeg.exe or would I need to rebuild?
[15:14:12 CEST] <Fig1024> pretty sure those are build options
[15:14:16 CEST] <kepstin> c7j8d9: where do you get this "--enable-vp9-highbitrate" option from? That's not an ffmpeg or ffmpeg configure option, and googling it doesn't seem to find anything relevant
[15:14:36 CEST] <furq> it's a libvpx configure option
[15:14:39 CEST] <furq> and also it's highbitdepth
[15:15:01 CEST] <c7j8d9> yes sorry miss spell
[15:16:24 CEST] <kepstin> c7j8d9: you should recompile ffmpeg if you switch that option on libvpx.
[15:16:43 CEST] <c7j8d9> got it. thanks kepstin
[15:17:00 CEST] <c7j8d9> never hurts to check for a shortcut
[15:34:50 CEST] <kerio> is nut the best format
[15:40:05 CEST] <iive> yes
[15:46:32 CEST] <dviola> hi
[15:47:30 CEST] <dviola> I'm trying to encode a small video for sending it over https://web.whatsapp.com/ i.e. ffmpeg -i foo.gif out.mp4 but after trying to send it I always get "unsupported format"
[15:47:42 CEST] <dviola> any ideas what format I should use?
[15:48:00 CEST] <dviola> I tried googling this and I get not much results, there are other mp4 files I can send just fine
[15:48:46 CEST] <Mavrik> What's ffmpegs output when you run the encode?
[15:50:44 CEST] <dviola> http://ix.io/yC3
[15:54:23 CEST] <furq> dviola: add -pix_fmt yuv420p
[15:55:12 CEST] <dviola> ok
[15:56:09 CEST] <dviola> that worked, thank you :)
[16:27:30 CEST] <zerodefect> Using the C-API, is there a handy function to clear an AVFrame containing pixels to a pre-defined set of colors?
[16:28:35 CEST] <zerodefect> say clear to red, black, blue or green.
[16:29:42 CEST] <devinheitmueller> zerodefect: You can probably do that with the chromakey filter.
[16:30:02 CEST] <zerodefect> Tbanks for the hint. I'll check it out.
[16:31:13 CEST] <zerodefect> and set the 'similarity' to 1.0?
[16:32:35 CEST] <devinheitmueller> I think you would only want to do that if the colors are an *EXACT* match.  If the color is inserted by some digitial process then that would be appropriate.  Otherwise you would need the similarity to be lower in order to approximate the color.
[16:33:10 CEST] <devinheitmueller> With real video from a camera, youre never going to have pixels which are red.  They will be some shade of red.
[16:34:13 CEST] <devinheitmueller> Oh wait, I have it backwards.  The value 0.01 means exact match.
[16:34:46 CEST] <devinheitmueller> If youre processing real video then youll likely have to play with that value.  If the color is injected through some digital process, then 0.01 would be appropriate.
[16:36:22 CEST] <zerodefect> Ok. Cool thanks.
[16:38:32 CEST] <zerodefect> It's a bit of shame to setup a graph/filter to retrieve one AVFrame of a certain color. :(
[16:39:51 CEST] <devinheitmueller> Better than writing an application from scratch.  :-)
[16:40:50 CEST] <zerodefect> Certainly! :)
[16:51:24 CEST] <zerodefect> devinheitmueller: Looks like I may need to use a pix_fmt that has alpha transparency (looking at the code).
[16:51:56 CEST] <devinheitmueller> That wouldnt be shocking.  The goal is typically to overlay with some other video/image, which would typically make use of the alpha channel.
[16:53:04 CEST] <devinheitmueller> You would probably colorspace convert to a format with an alpha channel, pass through the filter, and then the resulting image/video would have the alpha channel cleared for the areas that matched the requested color.
[16:53:36 CEST] <devinheitmueller> At least thats how chroma keying works in general.  I have never used ffmpegs implementation.
[16:55:36 CEST] <zerodefect> Ok. That gives me enough to give it a go. Thanks.
[17:59:08 CEST] <lukas_gab> Hi everyone!
[17:59:27 CEST] <lukas_gab> I want, to play rtsp stream from ip camera using ffplay.exe
[17:59:50 CEST] <lukas_gab> In VLC I can play this stream, but in ffplay.exe no
[18:00:01 CEST] <lukas_gab> please take see on this output
[18:00:02 CEST] <lukas_gab> https://pastebin.com/CLX9Jqzg
[18:00:14 CEST] <lukas_gab> how can I play stream in ffplay.exe ?
[18:00:47 CEST] <shincodex> what the crack
[18:00:48 CEST] <lukas_gab> Finally, I want mix 5 streams in ffmpeg i restream trugh ffserver, but now I test connection i ffplay.exe
[18:00:52 CEST] <shincodex> watch out for probe
[18:00:58 CEST] <shincodex> it sucks sometimes on virtual mjpeg streams
[18:01:03 CEST] <shincodex> you have to force the format
[18:01:23 CEST] <DHE> lukas_gab: might try adding "-rtsp_transport tcp" to the commandline
[18:01:44 CEST] <lukas_gab> ok, I try use this flag
[18:02:33 CEST] <Dhruv> Hello there, can anyone tell me if we can scale and crop video in the same ffmpeg command?
[18:02:49 CEST] <lukas_gab> WTF ?!
[18:02:59 CEST] <lukas_gab> @DHE - you are my God now
[18:03:04 CEST] <lukas_gab> it's work
[18:03:13 CEST] <lukas_gab> thank's a lot man
[18:04:23 CEST] <lukas_gab> I try to make some rtsp proxy server - becouse when I play grid with 5 stream HD, my terminal myst resize this stream in realtime and on Core2Quad 2,8Ghz is very laggy make calcuate 5 HD Streams
[18:04:35 CEST] <lukas_gab> and take this 20Mbps on my WiFi ...
[18:05:14 CEST] <hron84> Hi! I'd like to loop infinitely an image merged with an audio stream as flv. However, ffmpeg -stream_loop -1 -i 'file.mp3' -loop 1 -i 'image.png' does not work, any option i tried with them makes the stream end at least after the audio stream ends. But I need repeating audio stream under the image.
[18:05:17 CEST] <lukas_gab> ok, I go to try mix this stream in ffmpeg.exe, meybe in future I'll need your help
[18:05:26 CEST] <lukas_gab> but now Thank You !
[18:05:41 CEST] <lukas_gab> btw ffmpeg it's cool stuff
[18:06:10 CEST] <lukas_gab> in past I made some corporation soft to record and store users desktop
[18:12:00 CEST] <lukas_gab> @DHE - now I see little slow down the stream. I try to use "-rtsp_transport udp", to make low lag, but I have this same errors
[18:12:12 CEST] <lukas_gab> how can I play this stream on UDP layer ?
[18:15:49 CEST] <Dhruv> can anyone tell me if we can scale and crop video in the same ffmpeg command?Has anyone tried that? In my case it seems it ignores scale command totally, here is my command ffmpeg-ss 00:00:01 -i /sdcard/input.mp4 -to 00:00:07 -vf scale=720:720 -filter:v crop=2160:2160:840:0 -preset ultrafast /sdcard/output.mp4
[18:16:00 CEST] <Dhruv> can anyone tell me if we can scale and crop video in the same ffmpeg command?Has anyone tried that? In my case it seems it ignores scale command totally, here is my command ffmpeg -ss 00:00:01 -i /sdcard/input.mp4 -to 00:00:07 -vf scale=720:720 -filter:v crop=2160:2160:840:0 -preset ultrafast /sdcard/output.mp4
[18:20:47 CEST] <iive> lukas_gab: error are probably caused by something readodering or dropping packets, some shaper or priority queue. If you don't have loss, then tcp should have about same latency as udp.
[18:23:48 CEST] <lukas_gab> this is wird but on otherr soft udp work
[18:24:08 CEST] <lukas_gab> ok, but I have 1Gbps LAN cord betwean switch and cameras
[18:24:39 CEST] <lukas_gab> Now I work on RDP Desktop but on other machine TCP is god for me, zero latency
[18:38:06 CEST] <Blubberbub> I'm a little bit confused by Frames and Packets. The Documentation says that "Reading data from an AVFormatContext is done by calling av_read_frame, which will return an AVPacket... But what is AVFrame then, if read_frame generates a packet?
[18:38:29 CEST] <Anne> folks, I was wondering if any one knows why a video encodec by h264, with an AVI container does not play in media player, and how can I resolve this issue
[18:38:39 CEST] <Anne> I have to to WMP for this vido
[18:38:43 CEST] <hron84> Blubberbub: it sounds like #ffmpeg-devel would be more helpful for you :-)
[18:39:16 CEST] <hron84> Anne: did you installed a h264 codec for WMP?
[18:39:31 CEST] <Anne> no, how can I do that ?
[18:39:36 CEST] <iive> Blubberbub: packet contains complressed bitstream, frame contains raw image
[18:39:46 CEST] <iive> the codec turns one into the other.
[18:39:48 CEST] <hron84> I think WMP does not know the video codec by default.
[18:40:01 CEST] <Anne> Also is there any way to just make the codec available in windows for all the player ?
[18:40:54 CEST] <iive> wmp might be able to play it, if you use fourcc that it knows.
[18:40:57 CEST] <hron84> Anne: this is what you looking for i think: http://www.mediaplayercodecpack.com/
[18:41:10 CEST] <hron84> Anne: however, VLC can play nearly an video and audio format and codec.
[18:42:11 CEST] <Anne> so there is another video player which is important for my application, which is not playing this video
[18:42:23 CEST] <Anne> VlC is the only one playing it
[18:42:26 CEST] <Blubberbub> iive, but there are some demuxers that already return decompressed data which is then also stored in a packet, right? like alsa or lavfi?
[18:42:36 CEST] <hron84> Anne: then use the codec pack i linked before
[18:43:00 CEST] <hron84> that would install the codec for all applications that use DirectShow / WMP for playing videos
[18:43:29 CEST] <lukas_gab> Ok, now I want to combine some streams in one file - coud you can me tell why I have some error ?
[18:43:30 CEST] <lukas_gab> https://pastebin.com/1PhfMp8d
[18:43:32 CEST] <iive> Blubberbub: yes, and there is a raw codec that turns one into the other by doing a simple copy :D
[18:43:51 CEST] <Anne> hron84, let me install it
[18:44:02 CEST] <Blubberbub> iive, ah okay. That explains things. Thank you :)
[18:44:51 CEST] <Blubberbub> how does it know if something is raw, though?
[18:46:20 CEST] <iive> fourcc
[18:47:04 CEST] <hron84> How can I apply an image to this stream? ffmpeg -re -stream_loop -1 -i Music.mp3 -f flv rtmp://127.0.0.1:1935/live
[18:47:16 CEST] <hron84> -loop 1 -i image.png does not work
[18:47:23 CEST] <hron84> Maybe I need some -filter_complex
[18:47:43 CEST] <hron84> but i am terrible using this stuff.
[18:59:03 CEST] <relaxed> hron84: try, ffmpeg -loop 1 -i image.png -re -stream_loop -1 -i Music.mp3 -f flv rtmp://127.0.0.1:1935/live
[19:04:33 CEST] <Anne> hron84 WMP not playing
[19:05:14 CEST] <hiru> what's the correct way to use -ss and -to to cut a video file? using them before or after the -i string?
[19:06:05 CEST] <hiru> using it before the -i should make ffmepg read only that portion of the input file instead of the whole file making the process faster, right?
[19:06:09 CEST] <hron84> relaxed: I tried, stream ends
[19:06:31 CEST] <hron84> hmm, wait, i mistyped soemthing
[19:07:56 CEST] <hron84> relaxed: i fixed the command, picture streaming continue, but no audio
[19:08:40 CEST] <hron84> i want audio to streamed with the image in loop
[19:10:15 CEST] <hron84> hiru: order is needed. -t and -ss can be only applied after inputs are specified. So, ffmpeg -i something.mp4  -c copy -t 30 output.mp4 makes an output with the first 30 seconds of the input and this is the only valid way to specify these optioms
[19:10:55 CEST] <hron84> s/needed/important/
[19:11:28 CEST] <hiru> in the doc there's this part -> "When used as an input option (before -i), seeks in this input file to position."
[19:12:08 CEST] <hiru> "When used as an output option (before an output url), decodes but discards input until the timestamps reach position."
[19:14:34 CEST] <relaxed> hron84: maybe, ffmpeg -re -stream_loop -1 -i Music.mp3 -loop 1 -i image.png -map 0 -map 1 -f flv rtmp://127.0.0.1:1935/live
[19:15:06 CEST] <cryptodechange> does deblock occur before or after filters?
[19:15:19 CEST] <cryptodechange> So when I use the hlmean filter, does it apply deblocking after?
[19:18:19 CEST] <furq> if you mean x264 deblocking then after
[19:19:02 CEST] <hron84> relaxed: it plays the music only once again
[19:19:21 CEST] <hron84> I cannot imagine what happens
[19:19:44 CEST] <hron84> i copy-pasted the command line you specified from here
[19:26:07 CEST] <hron84> relaxed: the solution was: ffmpeg -y -loglevel 0 -re -stream_loop -1 -i Music.mp3 -f mp3 - | ffmpeg -y -re -i pipe:0 -loop 1 -i image.png -f flv rtmp://127.0.0.1:1935/live
[19:32:52 CEST] <relaxed> hron84: maybe he gets confused by the different loops in one command
[19:33:54 CEST] <hron84> relaxed: yeah.
[19:41:59 CEST] <thomedy> okay im reading throught the source code on ffmpeg because im writing my own player right now
[19:43:25 CEST] <thomedy> here is my question..
[19:43:26 CEST] <thomedy> https://www.ffmpeg.org/doxygen/2.5/avio_8c_source.html#l00218
[19:43:35 CEST] <JEEB> unless you really need to re-invent the wheel I recommend you basing your efforts on either libvlc or libmpv
[19:43:47 CEST] <thomedy> are you tlking to me
[19:43:53 CEST] <JEEB> yes
[19:44:03 CEST] <dsc> "punk"
[19:44:05 CEST] <thomedy> okay then let me tell you what i need and you tell me if im wasting my time
[19:44:11 CEST] <JEEB> surefine
[19:44:12 CEST] <thomedy> i ws just runnning command line
[19:44:28 CEST] <JEEB> your keyboard seems a bit borked .)
[19:44:30 CEST] <thomedy> because i have several different videos ... i want to keep them seperate i
[19:44:33 CEST] <thomedy> yes it is
[19:44:42 CEST] <thomedy> my son rippped off a keys so its temp
[19:44:44 CEST] <thomedy> ha ha
[19:44:56 CEST] <thomedy> okay....
[19:45:34 CEST] <thomedy> i dont want to combine files...  they need to stay seperate but i do need them to play in back to back without any pause...
[19:45:56 CEST] <thomedy> i tried ffplay and i get a hiccup in time so i figured i would redo my own player and just call the next frame myself
[19:46:05 CEST] <thomedy> recommendations?
[19:46:09 CEST] <JEEB> ffplay is just a proof of concept / example :)
[19:46:19 CEST] <JEEB> test mpv with the playlist loop options
[19:46:20 CEST] <JEEB> *option
[19:46:31 CEST] <thomedy> what do you mean im sorry
[19:46:37 CEST] <thomedy> ill google mpv
[19:46:42 CEST] <JEEB> https://mpv.io
[19:46:55 CEST] <JEEB> it's a nice player based on the FFmpeg libraries and custom V/A outputs
[19:47:06 CEST] <JEEB> least retarded mplayer fork there is :)
[19:47:18 CEST] <JEEB> https://mpv.io/manual/master/#options-loop-playlist
[19:47:24 CEST] <JEEB> and it has a command line option to loop the playlist
[19:47:49 CEST] <thomedy> okay well here is to cutting time...
[19:48:11 CEST] <thomedy> i dont know how to cross my fingers in the chat room but know that i am crossing my fingers because the faster i can get done the better off
[19:48:14 CEST] <thomedy> your awesopme
[19:48:15 CEST] <thomedy> thanksw
[19:51:19 CEST] <thomedy> so far so good
[19:52:22 CEST] <JEEB> thomedy: also #mpv is for mpv support, and unlike ffplay it's a proper player :)
[19:55:17 CEST] <thomedy> i agree and mpv is the next room your awesome... im trying to read before i ask questions rtfm
[19:55:19 CEST] <thomedy> you get it
[19:55:30 CEST] <JEEB> sure :)
[19:56:05 CEST] <JEEB> anyways, learning FFmpeg's APIs or contributing to it can be fun, but "I just want a player" is usually something that has already been made, and in that case it's better to contribute to it
[20:04:40 CEST] <thomedy> there is the slightest of jump in files... ever so slight but it might just be my video... but this will get me to the next step
[20:05:03 CEST] <thomedy> if i find that i need to write an app i will but im hoping i can use this for now
[20:05:16 CEST] <thomedy> seriously JEEB you have saved me an enourmous amount of time
[20:05:41 CEST] <JEEB> yay
[20:11:17 CEST] <cryptodechange> So experimenting with nlmean, I feel s=6 is a good compromise
[20:11:25 CEST] <cryptodechange> but I noticed a bit of discoloring in small details
[20:11:25 CEST] <cryptodechange> https://imgur.com/a/lnktv
[20:12:15 CEST] <cryptodechange> Some blocking still around the outlines, but it was encoded with deblock=0,0, so I'll try with 1,1
[20:12:27 CEST] <cryptodechange> Would decreasing the patch size help with the discolor?
[21:38:44 CEST] <Blubberbub> if ffplay is only a "proof of concept / example" what is missing? what are the downsides? (regarding playback and not user-control-ability or user "friendlyness")
[21:50:26 CEST] <DHE> a user interface would be nice
[21:50:49 CEST] <Blubberbub> thats the thing i don't need :D
[21:51:49 CEST] <DHE> personally I consider ffplay to be a baseline for ffmpeg. it's not great, but it's a true ffmpeg experience. see how ffmpeg filters will react, etc.
[21:55:12 CEST] <Blubberbub> i have to admit: for a moment i was worried that my pts might overflow when i run my app for too long... then i did the math... :D:D
[21:55:58 CEST] <Blubberbub> Also i feel like every time i want to code something with ffmpeg i find implementing a state machine to be the easiest & cleanest thing i can think of
[22:03:16 CEST] <JEEB> Blubberbub: basically it's not optimized in any way or form. it's a basic thing fully based on SDL and there is no attempt to make anything work "the best it could". it's just that, a test app.
[22:03:48 CEST] <JEEB> not necessarily bad as such, but if you want a media player it's not the one you would go and recommend to people
[22:05:23 CEST] <yegortimoshenko> can i use ffmpeg to embed album art into a flac file? currently i use metaflac but i want to redistribute this so i'd like to use as few dependencies as possible
[22:09:45 CEST] <durandal_1707> no
[22:11:24 CEST] <Blubberbub> JEEB, could you name a few optimizations that are missing?
[22:12:33 CEST] <JEEB> proper optimized pipe line for video/audio rendering that can support properly most of the operating systems in use (Win, MacOS, *nix) and that lets you do most of the rendering on something that is good at it (the GPU)
[22:13:10 CEST] <JEEB> ffplay's idea is mostly just to use SDL as a sand bag, which works to a level
[22:13:57 CEST] <JEEB> and don't get me wrong, SDL can be used as a base to create a basic graphical application - it's just that when you go barebones you get barebones
[22:14:49 CEST] <Blubberbub> ah ok. i think i see what you mean now. Thank you
[22:14:51 CEST] <JEEB> even if on just *nix, you can just take a look at the differences between, say, ffplay and mpv
[22:17:09 CEST] <leif> Does anyone know what would cause the error: `Invalid pts (0) <= last (0)`?
[22:17:32 CEST] <leif> I _think_ its an encoding error as it only shows up when I'm encoding to mpeg4
[22:17:48 CEST] <JEEB> it means that you fed two things with the same PTS
[22:17:59 CEST] <JEEB> given that last was zero and now you're at zero again
[22:18:12 CEST] <leif> JEEB: I see. I was trying to speed up the video with (PTS-STARTPTS)*0.5
[22:19:17 CEST] <Blubberbub> so you need to drop some frames?
[22:19:53 CEST] <JEEB> leif: not sure if what's happening is correct, but in theory it could be integer PTS becoming the same if the time base is small enough :P
[22:20:00 CEST] <leif> Blubberbub: mmm...that would work. What filter does that?
[22:20:16 CEST] <leif> JEEB: that makes sense.
[22:21:42 CEST] <Blubberbub> select filter might be what you want?
[22:22:38 CEST] <leif> Blubberbub: Ah yes, that should work. Thanks.
[22:32:45 CEST] <Blubberbub> is there a generic method to add 2 audio samples with the same sample format together that also does clipping? (clipping is better than overflowing - right?) I only found AVFloatDSPContext - but that is only float and i would prefer to also be able to use some of the integer formats
[22:49:13 CEST] <leif> Is there any good way to simulate pausing in a stream?
[22:50:48 CEST] <leif> I guess I could just use selectframe and just set the PTS to be really long.
[00:00:00 CEST] --- Sat Jul 22 2017


More information about the Ffmpeg-devel-irc mailing list