[Ffmpeg-devel-irc] ffmpeg.log.20170812

[burek]

[00:00:52 CEST] <c_14> where are you putting the option?
[00:02:08 CEST] <c_14> should go before -i apparently
[00:02:15 CEST] <c_14> changed from global options to demuxer-private options
[00:02:28 CEST] <alexpigment> oh really?
[00:02:32 CEST] <alexpigment> lemme try that
[00:02:45 CEST] <c_14> since 2011 apparently
[00:03:14 CEST] <alexpigment> same deal
[00:03:17 CEST] <alexpigment> unrecognized option
[00:05:09 CEST] <alexpigment> https://pastebin.com/B1NQPyCp
[00:09:41 CEST] <c_14> That doesn't look like DV1394
[00:09:44 CEST] <c_14> What are you trying to do?
[00:10:06 CEST] <alexpigment> ?
[00:10:13 CEST] <alexpigment> this is atsc
[00:10:20 CEST] <alexpigment> not dv
[00:10:54 CEST] <c_14> yeah, that option just doesn't do what you want it to do?
[00:10:59 CEST] <c_14> What's the problem you're trying to solve?
[00:11:21 CEST] <alexpigment> i'm trying to make sure this video is tagged to match exactly a specification
[00:11:33 CEST] <alexpigment> and the "standard" is being reported as Component by MediaInfo
[00:11:37 CEST] <alexpigment> I'd like it to say NTSC
[00:12:04 CEST] <alexpigment> sure, it's metadata, but there's an option for ffmpeg to set that, and that option is currently unrecognized
[00:12:22 CEST] <alexpigment> on top of that, there's a deprecated option (tvstd), which explicitly tells you to use that option instead
[00:12:52 CEST] <c_14> the only thing that comes close to that is -target, but I'm not sure you want that
[00:13:03 CEST] <alexpigment> yeah i think target is pretty limited
[00:13:17 CEST] <c_14> Everything other than that you'll have to do manually
[00:13:19 CEST] <JEEB> it can't be before -i because it's an option for mpegtsenc
[00:13:27 CEST] <JEEB> if you mean the standard thing
[00:13:44 CEST] <c_14> Is it?
[00:13:48 CEST] <JEEB> yes
[00:13:55 CEST] <JEEB> mpegts_flags system_b
[00:13:56 CEST] <c_14> I can only find -standard as an option for the dv thingy
[00:13:57 CEST] <JEEB> it seems
[00:14:07 CEST] <JEEB> https://github.com/FFmpeg/FFmpeg/blob/master/libavformat/mpegtsenc.c#L1923
[00:14:24 CEST] <JEEB> no wait
[00:14:26 CEST] <JEEB> ATSC is default
[00:14:38 CEST] <JEEB> alexpigment: then you might want to check WTF mediainfo is reading from its source
[00:14:56 CEST] <JEEB> otherwise it's hard to know what the flying fuck you want to match
[00:15:39 CEST] <ars> hello guys
[00:16:00 CEST] <alexpigment> JEEB: so I guess I'm going to have to dig through the MediaInfo source code? I'm pretty green on any sort of programming
[00:16:04 CEST] <JEEB> yes
[00:16:12 CEST] <JEEB> that's the problem with mediainfo
[00:16:13 CEST] <JEEB> :P
[00:16:27 CEST] <alexpigment> for what it's worth, i'm able to set this option easily with x264
[00:16:37 CEST] <JEEB> so what exactly is it that you want
[00:16:42 CEST] <JEEB> I understood you wanted ATSC MPEG-TS?
[00:16:46 CEST] <alexpigment> yeah
[00:16:55 CEST] <JEEB> well that you're already getting so it's something else
[00:17:00 CEST] <JEEB> esp. if you're talking about x264
[00:17:06 CEST] <JEEB> which doesn't do MPEG-TS to begin with
[00:17:13 CEST] <alexpigment> but those are usually flagged with a tv standard which is the generic NTSC/PAL etc, even though those are technically innacurate
[00:17:32 CEST] <JEEB> the moment you said you could set it with x264 it is not MPEG-TS level
[00:17:47 CEST] <JEEB> please switch over and say what you actually want to set. "constant frame rate"?
[00:17:49 CEST] <JEEB> in libx264
[00:18:09 CEST] <alexpigment> I'm not understanding what you're saying
[00:18:20 CEST] <alexpigment> I'd like mediainfo to report my file as "NTSC" for the Standard
[00:18:22 CEST] <ars> I have problem. error while loading shared libraries: libva.so.1: cannot open shared object file: No such file or directory. Looks like it not linked statically. But i compile like: configure --enable-libmfx --enable-nonfree --enable-gpl --disable-shared --enable-static
[00:18:26 CEST] <alexpigment> there is a parameter to specify this in FFMPEG
[00:18:28 CEST] <alexpigment> it is called -standard
[00:18:34 CEST] <alexpigment> it is unrecognized
[00:18:38 CEST] <alexpigment> according to ffmpeg
[00:18:39 CEST] <JEEB> no, that is something else specific to a certain input module
[00:18:56 CEST] <JEEB> 1) you just said you could set whatever you want with x264 easily
[00:19:04 CEST] <alexpigment> yes
[00:19:10 CEST] <JEEB> 2) "if 1), then it is not MPEG-TS"
[00:19:16 CEST] <JEEB> because x264 doesn't do MPEG-TS
[00:19:22 CEST] <JEEB> thus, most likely waht you need is a libx264 parameter
[00:19:25 CEST] <alexpigment> i'm not using MPEG-2 in an MPEG-TS container. I'm unable to set the equivalent flag
[00:19:34 CEST] <alexpigment> *now using
[00:19:54 CEST] <JEEB> ok, unfortunately I don't know the flying fuck about the mpeg2 encoder
[00:20:05 CEST] <JEEB> nor do I know which flag mediainfo is looking for then
[00:20:09 CEST] <JEEB> since each video format is different
[00:20:21 CEST] <alexpigment> fair enough. happy friday JEEB :)
[00:20:31 CEST] <JEEB> in any case, look for parameters in the mpeg2 video encoder
[00:20:41 CEST] <JEEB> while looking at what the flying fuck mediainfo is looking for
[00:20:54 CEST] <JEEB> that way you can find out what you need and if the mpeg2 video encoder can write that
[00:21:10 CEST] <alexpigment> source code is all gibberish to me. i come in here because you guys are good at knowing what that gibberish usually means
[00:21:25 CEST] <alexpigment> JEEB: sounds good. thanks for your time
[00:21:31 CEST] <JEEB> we can't know what Random Application X wants exactly
[00:21:34 CEST] <alexpigment> go have a beer and forget this conversation happened
[00:33:39 CEST] <JEEB> ok, quickly scrolled through examples of strings "frame" and "time" in libavcodec/mpegvideo_enc.c
[00:33:46 CEST] <JEEB> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/mpegvideo_enc.c;h=918728391d46a6c2651c909eec4f5c16e996b4c7;hb=HEAD#l740
[00:34:13 CEST] <JEEB> and a few other usages of time base
[00:34:17 CEST] <JEEB> that's all one gets
[00:34:35 CEST] <JEEB> so I recommend looking up whatever the fuck mediainfo is reading :D
[00:35:38 CEST] <JEEB> if you can override the time base, I recommend trying that in case that helps for the video encoder (in the weird case where it will utilize the time base for some header field in another mpegvideo related file)
[00:36:03 CEST] <JEEB> alexpigment: also do you even know if you can get this field shown with MPEG-2 Video?
[00:36:14 CEST] <JEEB> or are you basing on the fact that it is showing it with AVC
[00:36:24 CEST] <JEEB> and don't even knwo if it will show up with MEPG-2 Video at all
[00:36:56 CEST] <thousand> Hi folks, Sorry for jumping to question. I'm seeking a one liner command for quickly adding some start and end music to my video podcast. I have music.mp3 and video.mp4. How can I do this? Thanks in advance.
[00:38:35 CEST] <JEEB> well it may be fucked with a rusty spork http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/mpeg12enc.c;h=f45598a087022b2438213fdd22e2ef5ada6a5dc2;hb=HEAD#l138
[00:40:41 CEST] <JEEB> alexpigment: so if you were setting --fps XX with x264, the closest is to make sure the video encoder's time base is correct. that should write the frame rate value as per that file I linked
[00:41:58 CEST] <JEEB> but yea, check mediainfo source code to make sure what you're hunting
[00:42:38 CEST] <JEEB> and now, I've wasted enough time and effort into this bullshit of "I want to see this thing exactly this way in mediainfo" without having the slightest idea of what the fuck mediainfo is showing
[00:59:32 CEST] <jkqxz> alexpigment:  Sounds like you want to set video_format in the sequence display extension.  Edit <http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/mpeg12enc.c;h=f45598a087022b2438213fdd22e2ef5ada6a5dc2;hb=HEAD#l356> to write 2 rather than 0.
[00:59:55 CEST] <alexpigment> sorry, was away for a few minutes
[01:00:01 CEST] <jkqxz> (<https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-H.262-200002-S!!PDF-E&type=items> table 6-6 for possible values.)
[01:00:44 CEST] <alexpigment> JEEB: I appreciate your efforts, but seriously, there's no need to get angry about it. I know it's a field that other MPEG-2 videos have and show in MediaInfo
[01:01:33 CEST] <alexpigment> jkqxz: OK, is there a way to specify that in the command line parameters for FFMPEG while encoding with mpeg2video?
[01:02:54 CEST] <JEEB> jkqxz: so that's different from the frame rate header the encoder is already writing?
[01:03:17 CEST] <alexpigment> it's definitely decoupled from the frame rate
[01:03:19 CEST] <jkqxz> Yes, it's unrelated to the framerate.
[01:03:22 CEST] <JEEB> ok
[01:03:23 CEST] <alexpigment> because i have the correct rate
[01:03:45 CEST] <alexpigment> it's just showing "Component" in mediainfo because it's not being explicitly specified as NTSC
[01:03:48 CEST] <jkqxz> It's a single field which ffmpeg hardcodes, which you can edit the line I linked above to change.
[01:04:11 CEST] <JEEB> if it has various parameters I think if someone cares enough it should be an AVOption or something like that
[01:04:15 CEST] <alexpigment> ok, so it has to be done in the source code as a global change?
[01:04:19 CEST] <jkqxz> No, it's showing component because it's hardcoded to that.  No idea who thought that was a good idea (maybe because it's zero) - five is "unspecified", which it should be.
[01:04:31 CEST] <JEEB> alexpigment: anyways I just fucking hate mediainfo because it ends up with people like you coming up and having no idea of what they actually want
[01:04:44 CEST] <JEEB> other than "it shows up like that in mediainfo"
[01:04:55 CEST] <alexpigment> JEEB: apologies. I tried to be clear in stating that I want it to say NTSC instead of Component
[01:05:15 CEST] <alexpigment> so that it matches the *other MPEG-2 videos I'm using as references*
[01:05:17 CEST] <JEEB> yes, but that doesn't tell anyone on which level that field is
[01:05:22 CEST] <JEEB> or where it is
[01:05:37 CEST] <JEEB> jkqxz just happened to know MPEG-2 Video headers I guess
[01:05:46 CEST] <JEEB> and that the field was there
[01:05:53 CEST] <JEEB> which mediainfo might be parsing for that
[01:06:16 CEST] <JEEB> anyways if the default is a weird value please raise a bug on trac
[01:06:20 CEST] <alexpigment> I would hope they're parsing something :) there's nothing "component" about my command line
[01:06:45 CEST] <alexpigment> JEEB: defaulting to anything else is not ideal either
[01:06:46 CEST] <jkqxz> I only know it off-hand because it was an obvious thing to make settable with the mpeg2_metadata bsf, but that isn't merged anywhere yet.
[01:07:04 CEST] <alexpigment> it needs to be a feature request, I believe
[01:07:05 CEST] <JEEB> oh right, you did that one :P
[01:07:22 CEST] <JEEB> that makes sense then for you to know the fields in memory
[01:07:45 CEST] <JEEB> now the real question is whether that field actually does anything in actual use
[01:07:50 CEST] <JEEB> most likely: not
[01:07:53 CEST] <alexpigment> JEEB: i agree
[01:07:59 CEST] <alexpigment> most likely not
[01:08:32 CEST] <alexpigment> but when you're putting a file out there with an unknowable amount of destinations and combinations of hardware and software, it's a good idea to have your bases covered in case something *does* care
[01:08:54 CEST] <alexpigment> which is my general rule with encoding videos. be as explicit as possible just in case
[01:09:05 CEST] <JEEB> then fucking start learning those header values at some point
[01:09:19 CEST] <alexpigment> JEEB: your attitude is seriously caustic
[01:09:38 CEST] <alexpigment> i appreciate your efforts. please calm down about this
[01:09:40 CEST] <JEEB> maybe, I'm just tired of people like you coming with only "mediainfo shows me X"
[01:09:57 CEST] <JEEB> and that seriously burns me down and even knowing that I still go through code because I'm a fucking idiot
[01:10:05 CEST] <alexpigment> JEEB: then consider me "one of those people" and set your client to block my posts
[01:11:05 CEST] <JEEB> but seriously, if you are working with this shit you should have a basic understanding of this stuff better than me if you're working with MPEG-2 Video
[01:11:12 CEST] <JEEB> in broadcast
[01:11:31 CEST] <alexpigment> JEEB: i know video encoding. i know *nothing* about source code, programming, development, anything
[01:11:38 CEST] <alexpigment> consider the fact that my skillset doesn't align with yours
[01:11:51 CEST] <JEEB> where did I say coding
[01:11:52 CEST] <alexpigment> consider that i'm not being lazy, but that I'm ignorant
[01:11:53 CEST] <JEEB> I said goddamn fields
[01:12:01 CEST] <JEEB> in your goddamn video stream
[01:12:11 CEST] <alexpigment> please let me know how i can find such fields
[01:12:12 CEST] <JEEB> unless you were just uanble to tell me that it was a mpeg-2 video header called X
[01:12:19 CEST] <alexpigment> open my video up in a text editor?
[01:12:41 CEST] <alexpigment> again, consider my ignorance here. i don't know where to look to provide the information you're talking about
[01:12:42 CEST] <JEEB> if nothing else, hex editor. there are multiple stream parsers for video formats as well
[01:12:48 CEST] <alexpigment> like?
[01:12:58 CEST] <JEEB> elecard etc
[01:12:59 CEST] <alexpigment> i'm interested in learning
[01:13:13 CEST] <JEEB> also there's the patch set jkqxz is making that's on the libav-devel ml
[01:13:14 CEST] <alexpigment> ok, i'll look into that on monday
[01:13:17 CEST] <JEEB> which can print header values
[01:13:23 CEST] <jkqxz> The standard is by far the best reference for this sort of thing.
[01:13:27 CEST] <JEEB> yes
[01:13:33 CEST] <JEEB> having the spec on hand is the nr1 thing
[01:13:41 CEST] <JEEB> it even was linked by jkqxz
[01:14:16 CEST] <alexpigment> noted
[01:14:35 CEST] <alexpigment> anyway, i appreciate your help
[01:15:19 CEST] <alexpigment> JEEB: seriously, i don't want to make an enemy here. I don't know if I'm actually that annoying or if you're having a bad day. But if it helps ease things over, I'll gladly paypal you a beer (as 2017 as that sounds)
[01:16:45 CEST] <JEEB> unless I have a vivid memory of  you each case of me herping a derp is separate. I just saw a person who didn't even look at encoder/muxer options and poked random parameters and couldn't specify what field he wanted to set where.
[01:17:48 CEST] <alexpigment> JEEB: you're right. i'm googling for answers when I can't find them in the encoder/muxer options. it's my fault for assuming that there are more parameters that what's there
[01:18:08 CEST] <JEEB> there was a help thing to list options for demuxers, muxers etc
[01:18:21 CEST] <JEEB> I don't remember it by heart so you'll have to find it out :P
[01:18:29 CEST] <alexpigment> yes, i'm familiar with it
[01:18:29 CEST] <JEEB> it listed all the module-specific AVOptions
[01:18:35 CEST] <alexpigment> -h encoder=mpeg2video
[01:18:52 CEST] <JEEB> in this case there was no such thing, and you will want to make a ticket to make that field so you long for to be an avoption
[01:19:00 CEST] <JEEB> and then you will have to hope that someone cares enough :)
[01:19:19 CEST] <alexpigment> right. i'm going to log a feature request
[01:19:20 CEST] <JEEB> (or you make an additional option yourself which shouldn't be too hard)
[01:19:42 CEST] <jkqxz> JEEB:  <http://sprunge.us/fFgX>?
[01:19:44 CEST] <alexpigment> I just don't even know where to begin with that
[01:20:08 CEST] <JEEB> jkqxz: that comment is fucking weird. what does components even mean there
[01:20:19 CEST] <JEEB> maybe one would understand by reading the spec
[01:20:31 CEST] <JEEB> but yes, unless components means something funky that looks like a better default value
[01:20:32 CEST] <jkqxz> Yep, one does!
[01:20:57 CEST] <alexpigment> component i think means that it comes from component video. i don't know why that's useful though...
[01:21:11 CEST] <JEEB> well ye, but what's the meaning of that in the context of mpeg-2 video
[01:21:19 CEST] <JEEB> if there is such a special meaning
[01:21:24 CEST] <JEEB> if not, then "unknown" is better
[01:21:39 CEST] <alexpigment> yeah, although one better is just being able to specify NTSC or PAL
[01:21:49 CEST] <alexpigment> which are the only two meaningful options
[01:21:57 CEST] <alexpigment> (to me, anyway)
[01:22:11 CEST] <JEEB> alexpigment: I recommend looking into a patch that adds another AVOption into another module, which should be a small thing
[01:22:18 CEST] <JEEB> and then trying to mimic that :P
[01:22:56 CEST] <alexpigment> JEEB: that sits with my brain better. I guess I need to get an IDE though?
[01:23:01 CEST] <JEEB> nope
[01:23:03 CEST] <jkqxz> Only NTSC or PAL?  You forgot SECAM!
[01:23:05 CEST] <JEEB> just a sane text editor is enough
[01:23:28 CEST] <alexpigment> jkqxz: yes, secam. good old secam. is that even a thing anymore??? :)
[01:23:37 CEST] <JEEB> in VHS at least
[01:23:46 CEST] <JEEB> and you still find TVs sold with SECAM support
[01:23:51 CEST] <alexpigment> yeah, i have an NTSC/PAL/SECAM VCR actually
[01:24:10 CEST] <alexpigment> i have a lot of VCRs, as weird as that sounds
[01:24:15 CEST] <voip_> Hi guys i have error: error while loading shared libraries: libva.so.1: cannot open shared object file. Can i copy libva.so.1 from PC where i compiled ffmpeg ?
[01:25:32 CEST] <jkqxz> voip_:  Possibly, and it will likely need some of the other libva-foo libraries as well.  But, libva is just a dynamic loading shim, so you won't be able to actually load any driver without more stuff as well.
[01:26:23 CEST] <voip_> jkqxz, thnk you. Because i dont want compile one more time and setup Intel DSk
[01:26:30 CEST] <voip_> sdk
[01:26:52 CEST] <alexpigment> JEEB: your suggestion about finding something similar and copying it is actually quite helpful. i know it probably seems dumb, but to a non-developer, the idea of submitting a patch seems like something that would take some training to do.
[01:26:56 CEST] <Sgeo> I have some Indeo3 videos whose dimensions are not divisible by 4. Why does ffmpeg have that restriction on Indeo3?
[01:27:16 CEST] <Sgeo> And are there any non-ffmpeg based viewers I can try?
[01:27:17 CEST] <alexpigment> Sgeo: that's pretty common for video encoders
[01:27:37 CEST] <alexpigment> or are you talking about decoding the video?
[01:27:42 CEST] <Sgeo> Decoding
[01:27:52 CEST] <JEEB> Sgeo: file a bug report on the trac and post a sample if only possible
[01:28:10 CEST] <JEEB> clearly someone had the idea that there was a limitation like that
[01:28:20 CEST] <JEEB> but if valid files exist that are like that then that's a bug
[01:28:25 CEST] <Sgeo> Um, my samples might be a copyright violation. I took them from a game.
[01:28:44 CEST] <JEEB> almost any sample is under some sort of copyright :P
[01:28:55 CEST] <JEEB> that's why people tend to try to cut them as much as possible
[01:29:11 CEST] <JEEB> as it is specifically for sample
[01:29:14 CEST] <JEEB> reasons
[01:30:55 CEST] <Sgeo> If I use video codec and audio codec copy on something ffmpeg considers invalid, will it still copy the video data?
[01:31:49 CEST] <Sgeo> If not, then I might not really have a valid sample unless I upload a 300MB file
[01:35:09 CEST] <JEEB> what container is it in?
[01:35:45 CEST] <Sgeo> .mov
[01:35:47 CEST] <JEEB> if it's AVI or something and it errors out on the first frame you might be able to use virtualdub to copy things over if ffmpeg doesn't support it (it should be able to with -c copy)
[01:35:56 CEST] <JEEB> ok, but it should still be able to :)
[01:36:10 CEST] <JEEB> (if it can write that format altogether)
[01:36:55 CEST] <Sgeo> So, the game stores its video data in one large .MOV file which apparently contains many movies. What I did was repeatedly write a skip atom at the beginning to make the .mov effectively one or another movie inside it, then use ffmpeg
[01:41:01 CEST] <mfg> Hi. Does anyone here know about HLS encryption? If I wanted to have a go at sample encryption is there anyone that can help get me started. I've done fixes to the http and hls demuxers before but I don't know anything about h264. I'm trying to asses if this is within my capabilities.
[01:51:34 CEST] <Sgeo> JEEB, thank you, created ticket
[01:55:45 CEST] <Sgeo> Aww, virtualdub doesn't speak .mov
[08:12:30 CEST] <froyo1> Is there a free online video encoding service that lets you input custom ffmpeg commands?
[08:14:54 CEST] <Cracki> froyo1, it's probably faster to encode your videos on your own computer
[08:15:40 CEST] <Cracki> cpu time is never free. you either pay for it with money or with other things.
[08:16:25 CEST] <froyo1> @Cracki Not when it's a netbook with an Intel Atom CPU...
[08:16:47 CEST] <froyo1> CloudConvert lets you do exactly this for free, except it's limited to 25 conversion mins
[08:16:59 CEST] <froyo1> So I was wondering if alternatives existed
[08:17:19 CEST] <froyo1> It's fine if it's paid too, if I can't find free ones
[08:18:09 CEST] <Cracki> any vserver or shell provider, if you can handle ffmpeg yourself
[08:19:02 CEST] <Cracki> in any case, I wouldn't know of a web service that takes custom parameters
[15:50:02 CEST] <beauty> hello
[15:52:24 CEST] <beauty> How get supported video stream of  bitrate X in an Nginx server? Nginx server's network bandwidth is Y.
[15:52:36 CEST] <BtbN> what?
[15:53:45 CEST] <beauty> If I know nginx server's bandwidth and every channel video's bitrate, could I get the maxnum supported video stream?
[15:55:04 CEST] <BtbN> Not sure what you're asking. Just calculate how many parallel streams fit in its available bandwidth
[15:55:14 CEST] <beauty> yes
[15:55:56 CEST] <beauty> I want to calculate the maxmum parallel video stream in an nginx server
[15:56:10 CEST] <BtbN> so... divide the total bandwidth by the one of a single stream?
[15:56:34 CEST] <beauty> how to calculate????
[15:57:06 CEST] <BtbN> uhm. With any calculator of your choice, if it's that hard to not do it manually.
[15:58:01 CEST] <beauty> I can't understand you
[15:58:08 CEST] <beauty> could you make it clear
[15:58:47 CEST] <BtbN> It's just two numbers. Can't be that hard to do a division.
[15:59:37 CEST] <beauty> but practice, it's not bandwith/bitrate
[15:59:50 CEST] <BtbN> It is
[16:01:01 CEST] <beauty> the distance between bandwith/bitrate and practice value  is large
[16:01:34 CEST] <BtbN> Make sure you use proper units
[16:03:02 CEST] <beauty> what's proper unit
[16:03:41 CEST] <BtbN> the same ones on both sides
[16:04:42 CEST] <beauty> ?????????
[16:05:02 CEST] <beauty> every ts is the same bitrate and size
[16:05:27 CEST] <BtbN> I'm not able to make any sense out of what you're saying. Good luck.
[16:07:24 CEST] <beauty> en
[19:36:22 CEST] <thebombzen> does libvpx-vp9 not thread?
[19:36:33 CEST] <thebombzen> Apparently using -threads:v 8 doesn't work
[19:39:23 CEST] <thebombzen> neither does -threads 8
[19:39:28 CEST] <JEEB> the encoder if new enough should have some threading
[19:39:36 CEST] <JEEB> decoder I have no idea
[19:40:01 CEST] <thebombzen> I built ffmpeg this morning, the encoder is from the Arch repos (1.6.1 I think)
[19:40:24 CEST] <thebombzen> now, the video is small spacially (400x300), so should that affect it?
[19:40:31 CEST] <thebombzen> because it should still have frame-based threading
[19:41:19 CEST] <JEEB> now that is something I don't know if libvpx has or enables by default
[19:41:24 CEST] <thebombzen> it looks like this should work: ffmpeg -i input.mp4 -c:v libvpx-vp9 -b:v 0 -crf:v 31 -deadline best -c:a libopus -b:a 128k -threads 8 output.webm
[19:41:25 CEST] <JEEB> for some reason a lot of vendors hate frame threading
[19:41:52 CEST] <thebombzen> it's only 404x336 so if that's one slice, I can see it ignoring threading if it has no frame threading
[19:42:01 CEST] <thebombzen> and the grapevine tells me that libvpx's threading is terrible
[19:42:28 CEST] <thebombzen> but I feel it should still use more than one thread, especially if I turn the "extrEEEEEme" setting on (i.e. -deadline best)
[19:42:47 CEST] <thebombzen> this is a 40 second 404x336 video, that took several minutes to encode
[19:50:16 CEST] <Easyfab> and with -row-mt 1 did libvpx-vp9 have better threading ?
[19:55:30 CEST] <Easyfab> thebombzen, for me with -row-mt 1 -threads 8 threading is enabled
[19:56:23 CEST] <thebombzen> I get "unrecognized option -row-mt"
[19:57:06 CEST] <Easyfab> how old is your libvpx  / ffmpeg version ?
[19:58:42 CEST] <thebombzen> libvpx is 1.6.1, ffmpeg was built this morning
[19:58:58 CEST] <thebombzen> row-mt is in the documentation, it just complains about it
[20:05:17 CEST] <thebombzen> Easyfab: so this is true if I test with git master, and also with 3.3.3
[20:05:23 CEST] <thebombzen> both using libvpx v1.6.1
[20:05:54 CEST] <thebombzen> in case you want a log, http://sprunge.us/KKcK
[20:06:34 CEST] <thebombzen> if you want a nightly build log, http://sprunge.us/QcPU
[20:11:56 CEST] <Easyfab> row-mt was added after 1.6.1 so you must build libvpx with recent source and not using the 1.6.1 default package.
[20:15:47 CEST] <JEEB> Easyfab: let me guess, that feature is in a version that they used for a recent chromium ;)
[20:16:00 CEST] <JEEB> so it is tagged as a version used in chromium but has no official release
[20:18:06 CEST] <Easyfab> v1.6.1 was tagged in january and row-mt was added in february -> https://github.com/webmproject/libvpx/commit/5881601488ef4278d7ca2b06e2ad4d44b3063620
[20:18:39 CEST] <Easyfab> no new stable version since january
[20:18:40 CEST] <JEEB> yea, I would just guess taht the code is already in production in chromium ;)
[20:18:51 CEST] <JEEB> since that's the classic way google operates
[20:19:06 CEST] <JEEB> it tags things for stability for chromium, but doesn't really care about "external" releases
[20:19:11 CEST] <JEEB> even though it might as well make one the other
[20:19:26 CEST] <JEEB> since if you're pushing libvpx to millions with chromium you might as well call it a release :D
[20:19:44 CEST] <JEEB> and then those who don't know about these tendencies in google end up using funky old versions :)
[20:33:19 CEST] <Easyfab> for me (i7-2600k) without row-mt  speed=0.0437x and with row-mt  speed=0.168x    not quite the same speed :)
[21:07:46 CEST] <Ingvix> If I burn subtitles to hevc ffmpeg turns it to h264. How do I get it to keep it as hevc?
[21:09:03 CEST] <JEEB> the default it picks is some default set for the container format, you generally shouldn't be trusting that selection but rather specifying what you want
[21:09:21 CEST] <JEEB> as in, you should be picking one of the encoders that encodes HEVC, libx265 probably being the most popular
[21:09:45 CEST] <Ingvix> I see
[21:11:01 CEST] <Ingvix> so I add "-c:v libx265" to the command and that should do it?
[21:11:11 CEST] <JEEB> after the input, yes
[21:11:22 CEST] <JEEB> that will then use whatever are the default parameters for that encoders
[21:11:22 CEST] <Ingvix> yes, thanks
[21:11:24 CEST] <JEEB> *encoder
[21:15:13 CEST] <Cracki> don't rely on default parameters, they're supposed to suck
[21:15:24 CEST] <Cracki> at least bitrate is supposed to.
[21:15:40 CEST] <JEEB> at least x264 and x265 have their defaults IIRC what the libraries' defaults are
[21:16:12 CEST] <JEEB> for x264 while CRF 23 is a bit on the higher compression side, I've seen a lot of people be just happy with it (which is why it's the default in the library)
[21:16:43 CEST] <JEEB> oh, and you get people asking for "best settings" while there is no magical number that's "best"
[21:17:15 CEST] Action: JEEB just ends up recommending encoding ~2500 frames or so with different parameters and picking the highest CRF that still looks good
[21:17:53 CEST] <JEEB> but yes, going back to that comment, if it was any of the internal encoders - totally :D 200kbps default generally doesn't end up good. at least with video. that said, libx264 and libx265 wrappers should be OK in that sense
[21:19:32 CEST] <Cracki> i love 200 kbps
[21:19:41 CEST] <Cracki> it tells you when someone didn't pay attention
[21:20:02 CEST] <Cracki> 200k also applies to audio, didn't it... at least for one audio encoder, I once got 200k
[21:20:13 CEST] <JEEB> yes, it probably is just a global avcodec default
[21:20:16 CEST] <Cracki> <3
[21:24:46 CEST] <Ingvix> um, is 200kbps high or low in this case? It wasn't until recently that I got together with ffmpeg and video converting
[21:25:36 CEST] <JEEB> depends on various things but don't worry - libx264 and libx265 override that default :P
[21:25:48 CEST] <JEEB> so they're not affected by the global default thingamajig
[21:26:09 CEST] <paperManu>   bnnyyyy
[21:30:20 CEST] <Ingvix> just to be sure: if I want to specify which audio track to use in the output file I need to add "-map 0:[audio stream I want]" and "-c:a copy"?
[21:31:40 CEST] <Ingvix> and I only want one in the output
[21:41:28 CEST] <Ingvix> and then I have to define the video stream too, it seems
[21:41:38 CEST] <Ingvix> map it, I mean
[22:16:54 CEST] <dorvan> hi all, I have to stream as "live view" an rtsp/rtmp input to an html visualization output, at moment  I'm using a mp4 without problems for input but i can't make a good "live view" to be viewed by web, hsl and dash are excluded due big lags, ffserver seems it's no more maintened, so can be a problem adopt it. I'm here after month of tests and research on internet, now I'm searching for a...
[22:16:56 CEST] <dorvan> ...definitive solution.  tips to make a dinamic socket file as output are also welcome.
[22:22:05 CEST] <BtbN> HLS and DASH is the best you'll get for viewing a live stream in a browser.
[22:22:13 CEST] <BtbN> everything else needs Flash
[22:24:59 CEST] <JEEB> in theory you could utilize an mpeg-ts parser or something from HLS.js and fuck around with HTTP if XHR lets you access the data while the connection is still going
[22:25:48 CEST] <BtbN> using fragmented mp4 would be more natural
[22:25:57 CEST] <BtbN> but then you essentially have DASH anyway
[22:26:41 CEST] <JEEB> sure, you could just be feeding segments into this thing. and since DASH segments don't require to start with an IRAP you could even make the latency small'ish
[22:26:44 CEST] <BtbN> The MediaSourceExtensions only take in mp4. So even if you managed to stream mpeg-ts to JavaScript. You'd have to make up chunks there, and end up with the same delay that causes.
[22:27:39 CEST] <JEEB> but yea, this all bases upon XHR being able to feed you data while you keep the connection going
[22:27:43 CEST] <JEEB> which I'm not sure is something doable
[22:31:09 CEST] <BtbN> you can, but it's not overly efficient
[22:31:30 CEST] <BtbN> As it will keep everything it downloaded so far in a buffer until the request finishes
[22:33:34 CEST] <dorvan> JEEB: I've already used hls/ts with hls.js... but lag is high...
[22:33:45 CEST] <JEEB> that's not what I was talking about
[22:34:03 CEST] <JEEB> also your latency in MPEG-DASH/HLS depends on a whole lot of things in your chain to be honest
[22:34:09 CEST] <JEEB> such as length of segments etc
[22:34:27 CEST] <JEEB> esp. DASH lets you do funky stuff because segments don't have to start with random access samples
[22:34:51 CEST] <JEEB> so you can still keep your compression efficiency, but not have as long latency
[22:35:19 CEST] <JEEB> but not talking about that - some sort of streaming input is the least bad lag-wise
[22:35:26 CEST] <JEEB> s/lag/latency/
[22:35:33 CEST] <JEEB> you will still have lag even if you don't have latency
[22:35:57 CEST] <JEEB> (for example if you have a random access sample every ~10s then at most you have to wait for that much to get a picture but your latency can still be much lower)
[22:36:57 CEST] <dorvan> ok, excluding the destination media (html), the problem it make a "remote source" input to a "local source" dinamic file as output...
[22:38:11 CEST] <dorvan> where on a h264 file the needs are header format and content format (content format can be dynamic through keyframs, but header will be static)
[22:38:30 CEST] <dorvan> this usually it's solved by a server....
[22:39:39 CEST] <dorvan> the best, i think, can be trasform a remote streaming in an local streaming device like /dev/video with v4l  (as intermediate output)
[22:39:48 CEST] <dorvan> or something like this
[22:41:03 CEST] <dorvan> rstsp/rtmp -> local device/socket <- read to output
[22:41:31 CEST] <dorvan> BtbN: what do you think about?
[22:41:41 CEST] <dorvan> JEEB: what do you think about?
[22:42:51 CEST] <BtbN> what?
[22:43:07 CEST] <dorvan> the process need to be the same to make more complex operation in a chain, like motion detection, or face recognition
[23:02:15 CEST] <dorvan> https://github.com/bkw/node-dronestream
[23:03:47 CEST] <dorvan> BtbN: what is not clear?
[23:04:27 CEST] <Cracki> I think you're unaware of the complexity of your task
[23:04:49 CEST] <Cracki> there are several technologies that can be a part of the solution
[23:05:26 CEST] <BtbN> in short, you're not going to be able to do low-latency streaming to a webbrowser, without utilizing either flash or WebRTC, which both is a giant mess
[23:05:47 CEST] <Cracki> (I advised to use webrtc in #opencv)
[23:06:19 CEST] <Cracki> I also advised to try mjpeg, since it's only happening locally anyway
[23:06:26 CEST] <BtbN> streaming to a browser via WebRTC is... not exactly straight forward
[23:06:41 CEST] <BtbN> I haven't figured out how to even do it at all
[23:06:47 CEST] <Cracki> I've seen "content replacement" repeated jpeg "streams"
[23:07:35 CEST] <dorvan> yes, mjpeg seem the less complex way, but it's not the best performance solution.
[23:07:56 CEST] <Cracki> try it
[23:08:16 CEST] <Cracki> unless you've tried it, you have no performance numbers
[23:08:23 CEST] <dorvan> Cracki already tried.
[23:08:28 CEST] <Cracki> and?
[23:09:02 CEST] <kerio> mjpeg is the best
[23:09:07 CEST] <kerio> you <img>
[23:09:09 CEST] <kerio> and it works
[23:09:11 CEST] <dorvan> Cracki: i had performance problem... so now since 3 weeks I'm on a mp4 output file take by ffmpeg
[23:09:18 CEST] <BtbN> if you don't care about the massive bandwidth
[23:09:21 CEST] <Cracki> be more specific please
[23:09:32 CEST] <Cracki> "performance problem"
[23:09:33 CEST] <BtbN> mp4 cannot be streamed
[23:10:03 CEST] <Cracki> "bandwidth" on localhost is relative
[23:10:11 CEST] <dorvan> kerio: 4mp frames at 30fps on many input streams... have you tips to tuning mjpeg?
[23:10:19 CEST] <Cracki> downsample
[23:10:36 CEST] <dorvan> BtbN: yes with some ffmpeg flags and a buffer
[23:10:55 CEST] <BtbN> that's fragmented mp4
[23:11:01 CEST] <BtbN> essentially a lot of very short mp4 files
[23:11:02 CEST] <Cracki> if you display multiple streams, no person viewing that is gonna profit from 30 fps. use 15 fps.
[23:11:15 CEST] <dorvan> BtbN: no, monolithic. not fragmented
[23:11:34 CEST] <BtbN> "monolithic" mp4 cannot be streamed due to the nature of the format
[23:11:52 CEST] <kerio> do browsers support fragmented mp4
[23:11:53 CEST] <BtbN> well, it can be stream-played when the moov atom already exists
[23:12:00 CEST] <BtbN> DASH is fragmented mp4
[23:12:08 CEST] <kerio> without all that bullshit machinery
[23:12:14 CEST] <kerio> :^)
[23:12:17 CEST] <dorvan> BtbN: can be streamed, there is a specific page on ffmpeg docs to make it..
[23:12:30 CEST] <BtbN> dorvan, you mean the page about fragmented mp4?
[23:12:50 CEST] <Cracki> dorvan, you need to be a lot more specific in your statements.
[23:13:09 CEST] <dorvan> BtbN: we have used 3 different solution in tests... HLS, DASH, a a trick with mp4 supported by some ffmpeg flags
[23:13:19 CEST] <BtbN> fragmented mp4 can still be a single file. Basically a lot of short mp4 files stitched together
[23:13:33 CEST] <dorvan> BtbN: no non the page about fragmented mp4?
[23:13:42 CEST] <dorvan> BtbN: no non the page about fragmented mp4!
[23:13:57 CEST] <Cracki> well which one then?
[23:14:32 CEST] <dorvan> Cracki: about what?
[23:14:36 CEST] <Cracki> nevermind
[23:14:58 CEST] <Cracki> you've been given several alternatives.
[23:15:39 CEST] <Cracki> and I still think jpeg/mjpeg can work for your application, but unless you become more forthcoming with details, you can't be helped.
[23:15:40 CEST] <dorvan> Cracki: I've no real answers... only tips about already tried solutions... without considering my answers..
[23:16:25 CEST] <Cracki> if it didn't work, maybe that didn't have anything to do with the technology, but with how it's been configured
[23:16:38 CEST] <dorvan> Cracki: if you read me i can provide my position about. not like BrbN position..
[23:17:30 CEST] <Cracki> ¯\_(Ä)_/¯
[23:17:34 CEST] <dorvan> Cracki: you know how to use a local device with v4l in ffmpeg like a usb cam? dev /dev/video0 for example...
[23:18:10 CEST] <dorvan> Cracki: do you know what happen in /dev/video0 ?
[23:18:54 CEST] <dorvan> Cracki: I need to have the same pointer for a remote stream.
[23:19:17 CEST] <dorvan> input format it's not a problem... conversion it's not a problem...
[23:20:17 CEST] <dorvan> output it's a problem... and it's the like Mjpeg stream with <img> like say kerio
[23:20:55 CEST] <Peste_Bubonica> Hi folks. I'm trying to encode x264 to x265, using ffmpeg 3.3.3 and hevc_nvenc
[23:21:10 CEST] <BtbN> *h
[23:21:23 CEST] <Peste_Bubonica> this is the command line:  ffmpeg -i input.mkv -vcodec hevc_nvenc -pix_fmt yuv420p10 -preset hq -2pass 1 -vb 8000k -acodec copy output.mkv
[23:21:27 CEST] <BtbN> x264/4 is not a codec
[23:21:36 CEST] <Peste_Bubonica> and this is the error: [hevc_nvenc @ 0xc18e20] Cannot init CUDA
[23:22:01 CEST] <BtbN> Is this on Windows or Linux?
[23:22:03 CEST] <kerio> your issue is that ffmpeg can't init CUDA
[23:22:03 CEST] <dorvan> BtbN: probably you are a Bot! :-)
[23:22:08 CEST] <Peste_Bubonica> Linux
[23:22:22 CEST] <BtbN> Is your nvidia driver working properly otherwise?
[23:22:29 CEST] <dorvan> Peste_Bubonica: h264 h265 ..... are the codec
[23:22:49 CEST] <Peste_Bubonica> BtbN, I think so. I can play games here for example
[23:22:57 CEST] <Diag> on linux?
[23:22:57 CEST] <Diag> psh
[23:23:02 CEST] <Peste_Bubonica> driver version is: 381.22
[23:23:07 CEST] <Cracki> games don't mean cuda/nvenc can be used
[23:23:22 CEST] <BtbN> Well, them working means the driver is working
[23:23:27 CEST] <JEEB> esp. since some distros even separate the packages
[23:23:30 CEST] <Cracki> should...
[23:23:31 CEST] <JEEB> against nvidia's wishes
[23:23:32 CEST] <Cracki> what GPU is it?
[23:23:38 CEST] <Peste_Bubonica> a GTX 1070
[23:23:45 CEST] <JEEB> so you might have the drivers installed, but not the CUDA stuff
[23:23:53 CEST] <Cracki> ok recent enough :P
[23:23:55 CEST] <BtbN> It's most likely something with your driver installation, yeah
[23:24:09 CEST] <BtbN> The cuda library is not missing, that would throw a diffrent error
[23:24:20 CEST] <BtbN> It just fails on the most basic thing it can do
[23:24:30 CEST] <JEEB> ok
[23:24:41 CEST] <Peste_Bubonica> I can check for something specific to my distro about cuda support.
[23:24:46 CEST] <Cracki> see if you can find a tool to "detect cuda devices"
[23:24:59 CEST] <BtbN> ffmpeg can be such a tool
[23:25:03 CEST] <Cracki> ah good
[23:25:05 CEST] <BtbN> https://github.com/FFmpeg/FFmpeg/blob/release/3.3/libavcodec/nvenc.c#L442
[23:25:09 CEST] <BtbN> this is where the error comes from
[23:25:37 CEST] <BtbN> cuInit is like... by far the most basic CUDA function. I can't think of a good reason for it to ever fail, other than the driver or the system being massively messed up
[23:25:52 CEST] <Cracki> can I assume this code will always grab a device? there are env vars that direct which devices/classes to use...
[23:26:09 CEST] <BtbN> cuInit is way before it even cares about devices.
[23:26:13 CEST] <Cracki> ah
[23:26:52 CEST] <Peste_Bubonica> I need to install some specific to use hevc_nvenc?
[23:26:56 CEST] <dorvan> Peste_Bubonica: you need 3 components, nvidia video codec and encoder, and cuda
[23:27:04 CEST] <BtbN> Usually just the nvidia-driver is enough
[23:27:13 CEST] <BtbN> But who knows what your specific distro did to it
[23:27:19 CEST] <BtbN> maybe they split it up and something is missing
[23:27:19 CEST] <JEEB> because nvidia distributes it all in a single package
[23:27:23 CEST] <Peste_Bubonica> I see. I'm using Gentoo.
[23:27:26 CEST] <JEEB> BtbN: or mismatching versions
[23:27:32 CEST] <BtbN> you're safe on Gentoo.
[23:27:39 CEST] <BtbN> There only is one monolithing nvidia-driver
[23:28:09 CEST] <BtbN> can I see lsmod | grep nvidia please
[23:28:12 CEST] <dorvan> Peste_Bubonica: me also, if you need i've the pkg list you need..
[23:28:22 CEST] <Peste_Bubonica> course, one moment
[23:28:53 CEST] <Peste_Bubonica> BtbN, https://pastebin.mozilla.org/9029569
[23:29:09 CEST] <BtbN> modprobe nvidia_uvm
[23:29:15 CEST] <BtbN> and then try again with ffmpeg/nvenc
[23:29:42 CEST] <Peste_Bubonica> BtbN, this module don't exists here
[23:29:52 CEST] <Peste_Bubonica> modprobe: FATAL: Module nvidia_uvm not found in directory /lib/modules/4.9.34-gentoo
[23:30:07 CEST] <dorvan> Peste_Bubonica: .... you have compiled ffmpeg with nvenc? and followed by emerge -uDN --with-bdeps=y @world ?
[23:30:10 CEST] <BtbN> you're missing the uvm useflag on the nvidia-drivers then
[23:30:14 CEST] <BtbN> Pretty sure CUDA needs that
[23:30:23 CEST] <Peste_Bubonica> nice, let me check
[23:30:47 CEST] <dorvan> Peste_Bubonica: dev-util/nvidia-cuda-sdk dev-util/nvidia-cuda-toolkit media-gfx/nvidia-cg-toolkit media-video/nvidia-video-codec sys-firmware/nvidia-firmware x11-drivers/nvidia-drivers
[23:31:00 CEST] <dorvan> Peste_Bubonica: check USE with eix....
[23:31:11 CEST] <BtbN> you don't need any of those, except the nvidia-drivers
[23:31:32 CEST] <Peste_Bubonica> BtbN, you are right, the i'm not using uvm useflag
[23:31:33 CEST] <dorvan> Peste_Bubonica: check the requirements on rebuilds
[23:31:56 CEST] <Peste_Bubonica> dorvan, thanks, I will try it right after rebuild the driver
[23:31:57 CEST] <dorvan> BtbN not true with cuda, but probably it's like the question of mp4
[23:32:28 CEST] <BtbN> yes, it is true. ffmpeg bundles all the needed dependencies. It only needs the runtime libraries from the driver.
[23:32:50 CEST] <dorvan> Peste_Bubonica: check the version alignement are changing something on that side..
[23:33:51 CEST] <Peste_Bubonica> not it works BtbN
[23:33:59 CEST] <Peste_Bubonica> its the uvm module
[23:34:01 CEST] <BtbN> *now?
[23:34:09 CEST] <dorvan> Peste_Bubonica: media-video/nvidia-video-codec it's aligne with latest CUDA toolkit, check it with nvenc use flags, if you have problem, check it through opencv.
[23:34:27 CEST] <Peste_Bubonica> i've recompiled the driver and loaded nvidia_uvm and its running
[23:34:34 CEST] <Peste_Bubonica> 154 fps
[23:34:42 CEST] <BtbN> yeah, CUDA needs the unified memory thing
[23:34:48 CEST] <ZeroWalker> been trying to compile ffmpeg with libx264 for ages now, why is it so hard:(
[23:35:16 CEST] <Peste_Bubonica> dorvan, this codec is from 6.0.1 version on gentoo
[23:35:32 CEST] <Peste_Bubonica> i've read that ffmpeg bundles the 8x version in 3.3 version, its right?
[23:35:55 CEST] <BtbN> yes, ffmpeg has no external build time dependencies for cuvid and nvenc
[23:36:17 CEST] <BtbN> It only loads cuda/nvenc at runtime. And those libs come from the driver
[23:36:21 CEST] <BtbN> no need for all those SDKs at all
[23:36:47 CEST] <Peste_Bubonica> very good
[23:36:50 CEST] <BtbN> Except if you need/want scale_npp or scale_cuda
[23:36:57 CEST] <BtbN> those actually require the CUDA SDK
[23:37:22 CEST] <Peste_Bubonica> I will check now, if mpv can use hardware decoding for x265
[23:37:26 CEST] <Peste_Bubonica> after loading this module
[23:37:37 CEST] <BtbN> probably not on its own
[23:37:57 CEST] <Peste_Bubonica> BtbN, I think that they use ffmpeg to decode
[23:38:14 CEST] <BtbN> Yeah, but hw accelerate decoding is not automatic
[23:38:35 CEST] <dorvan> Peste_Bubonica: what's you use case for this?
[23:39:09 CEST] <Peste_Bubonica> dorvan, right now, only encode my files fast
[23:39:19 CEST] <dorvan> ah ok...
[23:39:20 CEST] <Peste_Bubonica> with my CPU, it was producing 25fps
[23:39:37 CEST] <Peste_Bubonica> x264 to x265
[23:39:52 CEST] <BtbN> it's still h264/5
[23:39:57 CEST] <Peste_Bubonica> its not a professional usage. Only noob things :)
[23:39:57 CEST] <BtbN> x264 and x265 are applications
[23:39:59 CEST] <dorvan> ffmpeg with nvenc call as dependency  media-video/nvidia_video_sdk-6.0.1::gentoo so it's not "including alll it self"
[23:40:12 CEST] <dorvan> i've compiled it now...
[23:40:14 CEST] <BtbN> That's a wrong dependency then.
[23:40:18 CEST] <JEEB> ^this
[23:40:23 CEST] <dorvan> BtbN
[23:40:23 CEST] <kerio> hardware encoding has worse results tho
[23:40:26 CEST] <dorvan> ahahah
[23:40:30 CEST] <JEEB> it requires it if you want to use it
[23:40:33 CEST] <JEEB> not a build-time dep
[23:41:28 CEST] <BtbN> it would also be way too old to build a current ffmpeg
[23:41:50 CEST] <dorvan> ffmpeg +nvenc, needs  media-video/nvidia_video_sdk-6.0.1::gentoo, without -nvenc,  media-video/nvidia_video_sdk-6.0.1::gentoo it's not called in...
[23:42:00 CEST] <BtbN> That dependency is plain wrong.
[23:42:09 CEST] <BtbN> Or you're looking at a really old ffmpeg version.
[23:42:19 CEST] <dorvan> 3.3.3
[23:42:25 CEST] <BtbN> wrong dependency then
[23:42:31 CEST] <BtbN> ffmpeg 3.3 has the video sdk bundled.
[23:42:31 CEST] <Peste_Bubonica> dorvan, are you sure? I've installed 3.3.3 version here, without install nvidia video sdk
[23:42:54 CEST] <dorvan> Peste_Bubonica: done it now, let's try...
[23:42:57 CEST] <BtbN> well, "video sdk". It's a single header file.
[23:43:55 CEST] <dorvan> BtbN: yes probably but it need it... not "no, it's wrong" ....
[23:44:03 CEST] <BtbN> the ebuilds needs it
[23:44:06 CEST] <BtbN> ffmpeg does not need it
[23:44:13 CEST] <dorvan> :-)
[23:44:17 CEST] <BtbN> It can plain be removed from the ebuild, it's wrong.
[23:44:45 CEST] <dorvan> BtbN say to ebuild maintainer, but, pls, check it before...
[23:45:03 CEST] <BtbN> You realize I'm the author and maintainer of all the nvenc/cuda stuff in ffmpeg?
[23:45:13 CEST] <BtbN> I'm very well aware of what I'm saying.
[23:45:27 CEST] <BtbN> well, most of it by now.
[23:46:44 CEST] <dorvan> BtbN not I don't now... but I know your approach as irc user... and this suggest me to verify everything you wrote.. like you don't.
[23:47:20 CEST] <Cracki> thanks for all that nvenc/cuda stuff :D
[23:47:37 CEST] <dorvan> BtbN so if you are the maintainer, and as I reported to you, say to ebuild maintainer
[23:47:39 CEST] <Peste_Bubonica> Before come here to ask you guys, i've installed manually the video sdk, thinking that their lack could be the cause, so I've checked that this old version, does not support hvec (6.0.1). I will need de 8x. Then I found this page: https://developer.nvidia.com/ffmpeg
[23:47:49 CEST] <Peste_Bubonica> that says that ffmpeg 3.3 bundles it
[23:47:59 CEST] <BtbN> dorvan, I'm not responsible for the ebuilds in Gentoo. Not a Gentoo dev.
[23:48:00 CEST] <Cracki> btw, tips for making decode really fast on a gtx 560? I'm not getting beyong 160 fps (24 Mbps H.264 video from a camera)
[23:48:37 CEST] <Peste_Bubonica> so I removed this codec package. Its not a dependency for ffmpeg, even with the nvenc useflag
[23:48:56 CEST] <dorvan> BtbN: ok, you can simply reduce you "no" here, can be good enough :-)
[23:49:11 CEST] <BtbN> 160 fps might be the best that card can do
[23:49:18 CEST] <BtbN> Those older cards weren't overly fast.
[23:49:35 CEST] <BtbN> But not sure, never used anything older than a 760 with nvenc/cuvid
[23:50:07 CEST] <shayr> hello all, I have a small question (at least I hope it's small)
[23:50:41 CEST] <Peste_Bubonica> nice. MPV is working with cuda for decoding now
[23:50:42 CEST] <Peste_Bubonica> Using hardware decoding (cuda).
[23:50:42 CEST] <Peste_Bubonica> VO: [opengl] 1920x816 cuda[p010]
[23:50:46 CEST] <Peste_Bubonica> goood :)
[23:50:46 CEST] <dorvan> BtbN: so nvenc in ffmpeg 3.3.3 does not have deps also on cuda or X ?
[23:50:52 CEST] <shayr> if i want to lower the bitrate of a file to get a smaller size, what can I do to guarantee as best quality as possible? the original file is 15Mb and I really need it smaller
[23:51:14 CEST] <BtbN> It has no deps on the CUDA sdk. It does depend on the cuda runtime libraries, but those come with the nvidia-driver.
[23:51:43 CEST] <Peste_Bubonica> it takes 20 minutes to "recode" a 4.7GB h264 file
[23:54:18 CEST] <dorvan> BtbN: dev-util/nvidia-cuda-sdk dev-util/nvidia-cuda-toolkit x11-drivers/nvidia-drivers  ....seems toolkit 8.0.61, it's not included in nvidia-drivers, only libs?
[23:54:50 CEST] <dorvan> Peste_Bubonica: what about I/O, hdd speed?
[23:54:51 CEST] <BtbN> libcuda.so.1 and libnvidia-encode.so.1 are the only things ffmpeg needs for nvenc
[23:55:21 CEST] <BtbN> Everything else it needs to compile is bundled, so no need for external dependencies at build time.
[23:55:43 CEST] <Peste_Bubonica> dorvan, i've reading the file from a magetic disk, Sata 3, with 180MB/s of read speedy, and writing to an SSD that can handle about 450MB/s at max
[23:57:02 CEST] <dorvan> Peste_Bubonica: how many RAM? your bottleneck it's the source..
[23:57:25 CEST] <BtbN> you can use nvidia-smi to see the video engine load
[23:57:31 CEST] <BtbN> if it's at 100%, you have your bottleneck right there
[23:57:52 CEST] <dorvan> BtbN: true, but on a 1070....
[23:58:02 CEST] <BtbN> he's encoding h265
[23:58:12 CEST] <BtbN> that's not that crazy fast after all
[23:58:57 CEST] <Peste_Bubonica> dorvan, 32GB DDR3 1866
[23:59:34 CEST] <dorvan> BtbN: h265 does not more cpu friendly respect h265?
[23:59:57 CEST] <BtbN> h265/hevc uses _a lot_ more resources to encode than h264
[00:00:00 CEST] --- Sun Aug 13 2017