[Ffmpeg-devel-irc] ffmpeg.log.20160621

burek burek021 at gmail.com
Wed Jun 22 02:05:01 CEST 2016


[01:36:03 CEST] <sk41> Is it possible to make ffmpeg preserve the PID numbers of the streams when streaming live streaming a mpegts file?
[01:41:21 CEST] <hyponic> i am trying to transcode using nvenc on a machine with two graphic cards. can i spesify in ffmpeg what device i want it to use for the transcoding? it's just using the first one now.
[01:42:36 CEST] <c_14> hyponic: -gpu <number>
[01:43:07 CEST] <hyponic> c_14 thanks.. will give that a try :)
[01:44:03 CEST] <hyponic> I am willing to get something developed or fixed in ffmpeg. if anyone maybe a dev can help please pm me.
[01:44:36 CEST] <hyponic> *willing to pay :)
[03:01:27 CEST] <hyponic> does nvenc not decode the stream on the gpu, it seems like its done in software ?
[03:01:27 CEST] <hyponic> [03:00:33] Lars Larsen: and what would the switches be for decoding if it can?
[04:44:31 CEST] <Prelude2004cXXX> hey guys.. question
[04:45:04 CEST] <Prelude2004cXXX> running ubuntu ... compile is complaining about not having libfdk_aac not found and freetype is missing.. they are installed on the server .. tis is a clone of another server
[04:45:14 CEST] <Prelude2004cXXX> what's the problem.. ldconfig doesn't do anything to update the links
[04:45:18 CEST] <Prelude2004cXXX> somehow it doesn't see them
[04:49:26 CEST] <vladashram> Prelude2004cXXX I haven't used Ubuntu much for a couple years so i might not be remembering correctly, but do you have the devel packages of them installed?
[04:50:08 CEST] <Prelude2004cXXX> yes they are installed too
[04:50:22 CEST] <vladashram> What version of Ubuntu?
[04:54:31 CEST] <Prelude2004cXXX> 14
[05:53:40 CEST] <Bray90820> Is this 5.1 or 2.0 or what?
[05:53:40 CEST] <Bray90820> a52 audio (aka AC3) (ac-3)
[06:00:34 CEST] <kepstin> Bray90820: need more information. ac3/a52 audio can have multiple channel layouts, including both stereo and 5.1
[06:00:47 CEST] <kepstin> (ffmpeg -i and ffprobe will say...)
[06:02:00 CEST] <Bray90820> 3F2R/LFE looks like 2.1 right?
[06:02:55 CEST] <Bray90820> kepstin:
[06:03:21 CEST] <kepstin> 3F2R/LFE is 5.1
[06:03:39 CEST] <kepstin> but where are you seeing that? that doesn't sound like ffmpeg/ffprobe output, I think?
[06:04:32 CEST] <Bray90820> It's not
[06:04:39 CEST] <Bray90820> It was in VLC
[06:05:34 CEST] <Bray90820> But now I know my conversions are taking the correct audio track
[06:08:28 CEST] <pZombie> a quite interesting build using cheap dual xeon e5-2670 cpus https://www.youtube.com/watch?v=6b-01q09bLU
[06:24:15 CEST] <Shiny_> Hey
[06:24:29 CEST] <Shiny_> How do I install FFMPEG
[06:24:36 CEST] <Shiny_> I googled a bit but none of it works
[09:51:54 CEST] <nug700> does -r and -framerate do the same thing?
[09:55:31 CEST] <hero100> As an input option, ignore any timestamps stored in the file and instead generate timestamps assuming constant frame rate fps.  This is not the same as the
[09:55:31 CEST] <hero100>            -framerate option used for some input formats like image2 or v4l2 (it used to be the same in older versions of FFmpeg).  If in doubt use -framerate instead
[09:55:31 CEST] <hero100>            of the input option -r.
[09:55:31 CEST] <hero100>            As an output option, duplicate or drop input frames to achieve constant output frame rate fps.
[09:55:43 CEST] <hero100> sorry
[09:56:28 CEST] <nug700> from what I read, I use -framerate before -i to specify input rate and -r after to specify output rate
[09:58:01 CEST] <hero100> According to the man, yes.  I don't know the difference exactly. And there is an example in man: ffmpeg -r 1 -i input.m2v -r 24 output.avi
[10:22:28 CEST] <Bray90820> Can I convert an entire folder of video files at a time
[10:22:45 CEST] <Bray90820> Basically this but with an entire folder "ffmpeg -i Video.mkv -c:a copy Video.mp4
[10:29:13 CEST] <termos> https://www.ffmpeg.org/ffprobe-all.html#realtime_002c-arealtime I can't figure out how this realtime filter works, how can I specify it to ffprobe?
[10:36:09 CEST] <yagiza> how can I set options for format context?
[11:04:19 CEST] <pZombie> crf is the quality value (0-63 for VP9). To trigger this mode, you must use a combination of crf <q-value> and b:v 0. bv MUST be 0.
[11:04:48 CEST] <pZombie> that last part is quite important when encoding vp9. I was wondering why using different crf did not make a difference
[11:06:14 CEST] <pZombie> the standard settings for vp9 are pretty bad. Might be best if ffmpeg would override the standard and use something like -crf 33 -b:v 0 as standard
[11:06:28 CEST] <pZombie> otherwise the videos encoded look like there is a bug
[11:18:13 CEST] <pZombie> i am getting really nice results with those settings:
[11:18:15 CEST] <pZombie> ffmpeg -i 4k_test.mkv -vf scale=1280:720 -c:v libvpx-vp9 -crf 36 -b:v 0 -threads 22 -speed 1 -pix_fmt yuv420p output.webm
[11:18:29 CEST] <pZombie> not sure if -pix_fmt yuv420p  matters
[11:20:21 CEST] <pZombie> yes, it does, setting yuv444p almost doubled the bitrate
[11:21:44 CEST] <pZombie> it's strange however, the original material being 4:2:0 according to mediainfo, yet when pixel format is not set for the output, the bitrate goes about 20% higher
[11:22:18 CEST] <furq> you're rescaling it
[11:22:40 CEST] <pZombie> yes
[11:22:59 CEST] <pZombie> but which pixel format should i use for the output when the input is 420?
[11:23:05 CEST] <pZombie> not touch it at all?
[11:23:21 CEST] <furq> always use 420 unless you have a good reason not to
[11:24:02 CEST] <pZombie> is 420 the default if i do not set it?
[11:24:15 CEST] <furq> i have no idea for vp9
[11:24:34 CEST] <furq> ffmpeg will tell you what output format it's using, and you can check with ffprobe
[11:25:43 CEST] <pZombie> anyway, here is the output i got using crf 36 and b:v 0 https://drive.google.com/open?id=0BxMwFJfbOvWvcWJHaGFydnRnZ1E   , input video was this http://4ksamples.com/ses-astra-uhd-test-2-2160p-uhdtv/
[11:26:06 CEST] <pZombie> the quality is pretty damn good i think, and the encoding speed was also quite good
[11:26:25 CEST] <pZombie> at 0.45x
[11:27:08 CEST] <furq> i don't think -threads 22 is actually doing anything there
[11:27:38 CEST] <pZombie> i can test for that
[11:27:39 CEST] <furq> for 720p you'll only be able to use four columns
[11:27:42 CEST] <pZombie> by setting 1
[11:27:49 CEST] <furq> but you need to explicitly set -tile-columns 4
[11:28:03 CEST] <furq> unless it changed recently
[11:28:30 CEST] <pZombie> threads 22 definitely does something
[11:28:44 CEST] <furq> it definitely won't be using 22 threads though
[11:29:04 CEST] <pZombie> that i do not know, but it uses more than before
[11:29:04 CEST] <furq> i think you'd need to be encoding 8k to be able to use that many
[11:29:20 CEST] <pZombie> but this made me discover something else
[11:29:34 CEST] <pZombie> it does not use 100% of my cpu, but only 25-30%
[11:29:40 CEST] <furq> yeah libvpx's multithreading is bad
[11:30:06 CEST] <furq> with 4 columns that'll be using eight threads
[11:31:01 CEST] <pZombie> does 4 columns improve the quality or lower the size of the output?
[11:31:12 CEST] <furq> it does the opposite of those things
[11:31:26 CEST] <pZombie> why would i use it then?
[11:31:37 CEST] <furq> i assume it's using it automatically if it's using 8 threads
[11:31:54 CEST] <furq> without using tile-columns you won't get any multithreading at all
[11:32:33 CEST] <pZombie> the quality and size cost is negligible i assume?
[11:32:41 CEST] <furq> i should have thought so
[11:35:48 CEST] <pZombie> the only thing i did not like about the encoded video are the balloons at the end, having some distortion of the sky area around them if you look closely
[11:36:18 CEST] <pZombie> but x264 has this too at this compression rate
[11:41:42 CEST] <pZombie> the utilization of vp9 is so low, it does not even cause the CPU to go to it's maximum frequency
[11:41:52 CEST] <pZombie> when your power options in the OS are set to balanced
[11:41:55 CEST] <pZombie> at least for win10
[11:42:14 CEST] <pZombie> had to set minimum processor state to 100% to get it to encode at max speed
[11:42:29 CEST] <BtbN> libvpx is pretty bad at using your CPU.
[11:46:01 CEST] <pZombie> frame= 3552 fps= 13 q=0.0 Lsize=   29482kB time=00:02:22.04 bitrate=1700.3kbits/s speed=0.515x    now with processor state at 100%
[11:46:22 CEST] <pZombie> ffmpeg -i 4k_test.mkv -vf scale=1280:720 -c:v libvpx-vp9 -crf 33 -b:v 0  -threads 22 -speed 1 -pix_fmt yuv420p output.webm  settings
[11:47:05 CEST] <pZombie> a skylake 6700k might outperform me heavily however, if it's true that they have vp9 hardware support
[11:47:12 CEST] <pZombie> i am not sure they do
[11:47:17 CEST] <furq> skylake doesn't
[11:47:36 CEST] <furq> kaby lake will have vp9 decode
[11:47:42 CEST] <furq> i don't think anything has been announced yet with vp9 encode
[11:47:48 CEST] <pZombie> "Hardware encoding/decoding support[edit]
[11:47:48 CEST] <pZombie> The following chips, architectures, CPUs, GPUs and SoCs[33] provide hardware acceleration of VP9"
[11:47:53 CEST] <pZombie> https://en.wikipedia.org/wiki/VP9
[11:47:56 CEST] <pZombie> skylake is checked
[11:51:19 CEST] <pZombie> looking through this code https://cgit.freedesktop.org/vaapi/intel-driver/tree/src/i965_device_info.c    .is_haswell = 1 part of it. Maybe it works for haswell too?
[11:52:05 CEST] <furq> haswell and up have some gpu acceleration for vp9 iirc
[11:52:13 CEST] <furq> for decode, that is
[11:53:29 CEST] <pZombie> "Intel has added VP9 encoding support to the Video Acceleration API (VA-API).
[11:53:29 CEST] <pZombie> VA-API has already supported VP9 for accelerated video decoding, but now there's VP9 encode support added to its implementation. "
[11:53:34 CEST] <pZombie> http://www.phoronix.com/scan.php?page=news_item&px=VP9-Hardware-VAAPI-Encode
[11:54:16 CEST] <jkqxz> For Kaby Lake and later.
[11:54:25 CEST] <furq> it being added to vaapi doesn't mean there are any extant products which use it
[11:55:11 CEST] <pZombie> jkqxz not if the wikipedia page is correct, having a checkmark for skylake as supporting vp9 hardware encoding
[11:55:31 CEST] <pZombie> but that's wikipedia, you never know
[11:55:57 CEST] <furq> some articles i found mention something about "partial gpu encode" on skylake
[11:55:57 CEST] <jkqxz> There is weird GPU stuff implemented in the Windows media SDK stuff, I believe.  There is no fixed-function hardware.
[11:56:08 CEST] <jkqxz> None of that is implemented on Linux.
[11:56:12 CEST] <furq> i assume that's nothing special anyway
[11:57:14 CEST] <jkqxz> (It's really only meant for decode/playback on weak client devices.)
[11:57:52 CEST] <furq> i'm still using X58 so i have no idea about any of this
[11:58:21 CEST] <furq> it doesn't sound like i'm missing out on much though
[11:58:50 CEST] <pZombie> x58 was the best platform they ever released
[11:59:19 CEST] <furq> you'll be pleased to hear i'm planning on staying for a while longer then
[11:59:24 CEST] <furq> if i can find a cheap x5675 on ebay
[11:59:32 CEST] <pZombie> the only reason i left my i7-930 oced to 4.4ghz and went 5960x is because my sister needed a PC and i found an excuse to upgrade
[11:59:52 CEST] <furq> i assume you were also sick of paying six figure power bills
[12:00:21 CEST] <pZombie> power consumption wasn't that high with HT turned off
[12:00:32 CEST] <pZombie> and in games HT made zero difference
[12:01:51 CEST] <pZombie> you are also doing it wrong if you do not set the bios to adjust the voltage depending on your cpu frequency
[12:02:51 CEST] <furq> i'm running stock volts anyway
[12:03:10 CEST] <pZombie> yes, but that is a different matter
[12:03:23 CEST] <pZombie> does your CPU clock down on low load and adjust the voltage accordingly?
[12:04:05 CEST] <pZombie> if yes, the i7-9xx cpu will draw almost nothing on simple tasks like browsing/office etc
[12:04:11 CEST] <furq> it's been years since i looked at the bios of this thing
[12:04:17 CEST] <furq> speedstep or whatever it's called is enabled
[12:04:30 CEST] <pZombie> speedstep alone does not suffice
[12:04:43 CEST] <pZombie> you would need to check using cpu-z if you are using windows
[12:04:55 CEST] <pZombie> see if the voltage is static or dynamic depending on the frequency
[12:05:06 CEST] <furq> i seem to remember the board picking incredibly generous voltages if i left it on auto
[12:05:11 CEST] <furq> but like i say, it's been years
[12:06:22 CEST] <furq> also i'm in the middle of an encode so i can't really check for another couple of hours
[12:07:43 CEST] <pZombie> i thought i would see a huge difference upgrading from the i7-9xx, but not at all
[12:09:22 CEST] <furq> the 5675 is a higher multi and lower tdp so i figure i can at least clock it further
[12:09:28 CEST] <furq> plus two extra cores, obviously
[12:10:06 CEST] <furq> for about 15% of the price of the cheapest new intel 6-core
[12:10:57 CEST] <pZombie> those cpus have a very powerful APU, iris pro hd 6200
[12:11:24 CEST] <pZombie> not as powerful as the iris pro 580 thought
[12:11:26 CEST] <pZombie> though*
[12:11:43 CEST] <pZombie> but that one is available only on notebook cpus
[12:12:07 CEST] <pZombie> and no notebook out with it yet, except intel's own creation of a low power NUC
[12:12:32 CEST] <furq> i mean the xeon x5675
[12:12:35 CEST] <furq> http://ark.intel.com/products/52577/Intel-Xeon-Processor-X5675-12M-Cache-3_06-GHz-6_40-GTs-Intel-QPI
[12:12:41 CEST] <pZombie> i see
[12:13:17 CEST] <furq> i object on principle to paying a premium for a board and cpu with overclocking enabled
[12:13:42 CEST] <pZombie> you are getting this from ebay?
[12:13:47 CEST] <furq> probably
[12:13:58 CEST] <brontosaurusrex> overclocking is mostly gaming related, or ... ?
[12:14:06 CEST] <furq> it's nice for encoding too
[12:14:20 CEST] <viric> xeons use to have lower clock... single-thread things will work slower
[12:14:21 CEST] <brontosaurusrex> ok, but why not just have another machine or two
[12:14:30 CEST] <pZombie> overclocking is nice for everything
[12:14:40 CEST] <pZombie> as long as you do not kill your cpu
[12:15:15 CEST] <furq> yeah i've never been one of those people who spend a week constantly getting prime95 bluescreens in an attempt to get another 7mhz
[12:15:27 CEST] <viric> here an i7 gets over 90° without overclock, with the Intel cooler.
[12:15:34 CEST] <furq> i usually won't even overvolt unless it's one step for >200mhz
[12:16:00 CEST] <pZombie> 200mhz is what mine was running at
[12:16:21 CEST] <pZombie> but you have to be lucky, both with the cpu and motherboard to achieve that
[12:16:34 CEST] <furq> i meant one voltage step for a big cpu clock gain
[12:16:48 CEST] <furq> i don't really touch the bus speeds or anything like that
[12:17:09 CEST] <viric> what kind of cooling do you use in those situations, if you aim at something silent?
[12:17:20 CEST] <furq> i just get used to fan noise
[12:17:23 CEST] <viric> oh.
[12:17:25 CEST] <viric> I hate it.
[12:17:32 CEST] <furq> i've had a watercooling setup in the past but it was a pain in the arse to maintain
[12:17:41 CEST] <viric> furq: was it more silent than a fan?
[12:17:48 CEST] <viric> I have never listened to watercooling
[12:17:50 CEST] <furq> well it still has four fans on the radiator
[12:17:56 CEST] <furq> or mine did
[12:18:03 CEST] <viric> so it was not more silent?
[12:18:11 CEST] <furq> not really
[12:18:19 CEST] <pZombie> furq but you should. Those i7-9xx CPUs love 200mhz, and 200mhz extra clock speed for an i7-9xx is a joke. They almost all can go 3.8ghz with HT on and some proper cooling, and up to 4.4ghz with HT off
[12:18:25 CEST] <viric> ok. that's great to know. I never had a minimal idea about what noise to expect from water cooling
[12:18:46 CEST] <furq> pZombie: this thing will do 4ghz if i don't mind it melting the hsf
[12:18:56 CEST] <furq> viric: i assume you can get quieter systems
[12:19:07 CEST] <viric> Isn't there any part of, let's say, x264, that could be done collaborating with a gpu?
[12:19:13 CEST] <furq> you can use the gpu for lookahead
[12:19:19 CEST] <furq> it's not that useful though
[12:19:48 CEST] <viric> I meant more some opencl specfic thing
[12:19:57 CEST] <viric> or cuda, or whatever general calculation
[12:23:03 CEST] <pZombie> furq - i would recommend a netop 140xl watercooler, but spending 80¬ on a watercooler for an older system might not make so much sense
[12:23:21 CEST] <furq> yeah i'm not watercooling again, i don't plan on pushing the clocks that far
[12:23:34 CEST] <pZombie> that cooler cools my 5960x without making any noise up to 200w power draw. Above 200w it can get loud
[12:23:41 CEST] <furq> i should hopefully be able to hit 4ghz easily with an x5675
[12:23:59 CEST] <viric> pZombie: nice.
[12:24:04 CEST] <pZombie> but you need a motherboard with fan control for that one. Both for fans and the pump, or it will sound like a jet taking off
[12:24:24 CEST] <viric> don't all motherboards have fan control?
[12:24:35 CEST] <pZombie> the new ones do most of the time
[12:24:35 CEST] <viric> or you mean tuneable fan control?
[12:24:49 CEST] <pZombie> the old ones, just had for the cpu only at times
[12:24:59 CEST] <pZombie> you need to control the pump and fans separately
[12:25:04 CEST] <viric> auhm
[12:25:14 CEST] <viric> ah, then you want a motherboard with pump control, not fan control
[12:25:25 CEST] <pZombie> no, it's the same
[12:25:31 CEST] <brontosaurusrex> I'd like to hit a hertz or two with that blonde I saw today ...
[12:25:34 CEST] <pZombie> same connector
[12:25:47 CEST] <viric> pZombie: but the controller needs to know that one is a pump
[12:26:02 CEST] <viric> or not?
[12:26:06 CEST] <pZombie> viric - well, then i am surprised my system did not explode yet
[12:26:20 CEST] <furq> you can control pwm fans in software
[12:26:24 CEST] <viric> I'm just asking :) I have never seen water cooling
[12:26:31 CEST] <pZombie> my pump goes down to around 3000 rpm according to hwcontrol and up to 6000 rpm when more cooling is needed
[12:26:51 CEST] <viric> that's quite fast
[12:26:54 CEST] <pZombie> of course, that rpm value might just mean something else
[12:26:58 CEST] <viric> I didn't expect those speeds
[12:27:06 CEST] <pZombie> but you can definitely hear the pump when it's at max
[12:27:16 CEST] <viric> 6000rpm is 100Hz
[12:27:24 CEST] <viric> should be noticeable
[12:27:27 CEST] <furq> yeah that'll be the number it thinks a fan would be spinning at
[12:28:32 CEST] <viric> so the motherboard has to know how to control 'two fans' for one cpu?
[12:28:38 CEST] <viric> separately.
[12:28:42 CEST] <furq> you can do it in software
[12:29:01 CEST] <viric> well, everything can be done in software. I wonder what kind of support is that pZombie said in motherboards
[12:30:00 CEST] <pZombie> just normal pwm fan connectors the bios allows you to control and set their rpm depending on cpu temp or motherboard temp
[12:30:11 CEST] <pZombie> my motherboard allows me to control 4
[12:30:17 CEST] <pZombie> the rest cannot be controlled
[12:32:02 CEST] <pZombie> i believe that some corsair watercoolers allow you to control their behavior via software. The heatsink itself having a sensor and them being connected to USB
[12:32:18 CEST] <pZombie> if you do not want to do it over the motherboard fancontrol for whatever reason
[12:53:09 CEST] <pZombie> viric i made a little video to show you how this 140xl works in my 5960x system
[12:53:29 CEST] <pZombie> https://drive.google.com/open?id=0BxMwFJfbOvWvYUl3akdkeHdsNW8
[12:54:25 CEST] <pZombie> cpufanin and auxfanin3 are the two cpu fans on the nepton heatsink
[12:54:39 CEST] <pZombie> auxfanin0 is the pump
[12:55:02 CEST] <pZombie> as long as the fans are below 1300rpm, they are almost inaudible
[12:55:13 CEST] <pZombie> 1500 is ok still. Anything above is too loud
[12:55:31 CEST] <pZombie> the pump at 3500-4000 rpm is almost inaudible
[12:55:41 CEST] <pZombie> but when at max speed, it can be heard
[12:56:20 CEST] <pZombie> for getting max clockspeeds on a 5960x this is not the right heatsink, if you want it to be silent
[12:56:34 CEST] <viric> pZombie: thank you!
[12:56:47 CEST] <pZombie> any power draw above 220w and it will sound like a jet engine if you want to keep your cores quiet
[12:57:08 CEST] <pZombie> for a 6700k however, this heatsink would be ideal. No way you get more than 200w draw, no matter how high you overclock it
[12:57:57 CEST] <viric> pZombie: what guides auxfanin? based on what?
[12:57:57 CEST] <pZombie> my 5960x was a lemon anyway. It cannot go higher than 4.2ghz stable, and requires almost 1.3v to get 4.2ghz stable
[12:58:33 CEST] <pZombie> they are all guided by the motherboard, set in the bios. auxfanin0 which is the pump listens to the cpu temperature to set itself
[12:58:47 CEST] <pZombie> i can set to either listen to the cputemp or the motherboard temp
[12:58:51 CEST] <viric> so both fans are run by the cpu temperature
[12:59:06 CEST] <viric> ok, that's a bios option. right?
[12:59:11 CEST] <pZombie> both fans and the pump all listen to the cpu temp
[12:59:14 CEST] <viric> ok
[12:59:36 CEST] <viric> how does it decide what to run faster, one or the other?
[12:59:39 CEST] <pZombie> the implementation of fan control can vary from motherboard to motherboard
[12:59:46 CEST] <viric> otherwise it is like running them as a single fan
[13:00:06 CEST] <pZombie> i guess you are referring to the difference in rpm between the two fans
[13:00:10 CEST] <viric> yes
[13:00:14 CEST] <pZombie> i am not sure why they are different
[13:00:18 CEST] <pZombie> they should be about the same
[13:00:28 CEST] <viric> well
[13:00:32 CEST] <viric> they may get equal volt
[13:00:37 CEST] <viric> they just run at different speed
[13:01:03 CEST] <pZombie> there is a "silent mode" as setting on my board, which allows to completely spin one of the cpu fans down, having only one working, and only up it when it get hotter
[13:01:15 CEST] <viric> ok
[13:01:27 CEST] <pZombie> this applies only to the whatever is connected to the two cpu1 and cpu2 fan connecters
[13:01:30 CEST] <pZombie> connectors*
[13:01:38 CEST] <pZombie> not all motherboards have this either
[13:01:40 CEST] <viric> yes, I think that at most the MB may tune the volt to reach a minimum speed. Other than that, I don't think it checks the speed for anything
[13:02:01 CEST] <viric> just not to go below a minimum.
[13:02:09 CEST] <viric> The rest is just "more volt"
[13:02:13 CEST] <pZombie> yep, might be
[13:02:48 CEST] <viric> volts may be equal for the two fans, as they are based on a single temperature
[13:02:50 CEST] <pZombie> i had it at silent, but i wanted to draw a little bit more heat out of the system for the GPUs
[13:03:22 CEST] <pZombie> even at 500rpm with just 1 fan running, the system stays plenty cool unless you are maxing all cores
[13:03:32 CEST] <pZombie> the pump and the single fan at 500 rpm do a fine job still
[13:04:13 CEST] <pZombie> yes, and the fans might not be exactly equal
[13:04:19 CEST] <pZombie> spinning at different rates
[13:04:47 CEST] <pZombie> maybe dust got the best part of the one i had spinning all along when i was using silent mode, while the other was idle
[13:05:02 CEST] <pZombie> dust is a real killer of fans
[13:05:24 CEST] <pZombie> all my 7970 gpu fans died within a year when i was using them for bitcoin mining 24/7
[13:06:23 CEST] <pZombie> i am looking for a case now with some serious dust filters that can be cleaned easy without having to unscrew anything
[14:55:06 CEST] <flux> did you have at least non-serious ones before? even flimsy ones that are easy to remove do wonders.
[14:56:02 CEST] <furq> easy to remove being the key
[14:56:16 CEST] <furq> don't do what i did and buy a 4in3 kit where you have to remove all four disks to remove the filter
[14:56:36 CEST] <furq> thanks for that, coolermaster
[14:56:46 CEST] <pZombie> :D
[14:57:38 CEST] <pZombie> i had a silverstone on my x58 system. A really pretty a nice case, except if you wanted to clean the power supply filter, you had to unscrew the power supply first...
[14:58:37 CEST] <pZombie> most cases do not offer any reasonable dust filters. The meshes are rarely fine enough to really hold back any dust
[14:59:32 CEST] <pZombie> the one i am using right now is a haf coolermaster, and i got this one because it allowed you to mount the psu both on top and bottom
[14:59:58 CEST] <pZombie> i needed the extra space at the bottom for the gpus in my btc mining rig
[15:00:04 CEST] <pZombie> back then
[15:03:59 CEST] <HoloPed> Hello gents. I wish to use ffmpeg to capture H.264 data coming from a webcam that has on-board encoding (Logitech C920). Is this possible?
[15:05:09 CEST] <pZombie> did you check this https://trac.ffmpeg.org/wiki/Capture/Webcam ?
[15:07:04 CEST] <HoloPed> pZombie, I did, yes. I tried to list my devices and I get an error: Unknown input format: 'v4l2'
[15:07:12 CEST] <HoloPed> is 'v4l2' something I need to install ?
[15:07:16 CEST] <HoloPed> Running Win10
[15:08:01 CEST] <HoloPed> oh, that seems to be a linux thing ?
[15:09:02 CEST] <pZombie> when you do ffmpeg -y -f vfwcap -i list   do you get anything in the list at all?
[15:09:48 CEST] <HoloPed> pZombie, this is what I get http://pastebin.com/YxnYBQGU
[15:09:55 CEST] <HoloPed> note the IO error at the end
[15:09:56 CEST] <BtbN> From my experience, the h264 these webcams produce is horrible. Bad quality, a lot of errors.
[15:10:09 CEST] <HoloPed> BtbN, intersting
[15:10:39 CEST] <HoloPed> I need to capture 1080p video from the webcam , and then upload it
[15:10:45 CEST] <HoloPed> so trying to keep the size down,
[15:10:45 CEST] <pZombie> HoloPed - this is what i get too :D welcome to the club
[15:10:57 CEST] <HoloPed> what does that mean though ?
[15:11:00 CEST] <HoloPed> do you know ?
[15:11:04 CEST] <pZombie> nope, no idea
[15:11:17 CEST] <BtbN> With the very same Logitech WebCam, i resorted to using mjpeg output, and transcoding to h264 in software.
[15:11:36 CEST] <HoloPed> BtbN, how long does it take to transcode ? can you do it in realtime ?
[15:11:51 CEST] <thebombzen> it's generally a better idea to read mjpeg from webcams than it is to read rawvideo, unless it's over usb3
[15:11:57 CEST] <thebombzen> there's too much data otherwise
[15:12:06 CEST] <BtbN> Even libx264 on ultrafast way outmatches the quality the constrained baseline hw encoder in that camera is able to produce.
[15:12:16 CEST] <BtbN> thebombzen, h264, not raw video.
[15:12:21 CEST] <thebombzen> ah
[15:12:24 CEST] <BtbN> usb2 is too slow for h264 in raw video anyway.
[15:12:29 CEST] <BtbN> *for 1080p
[15:12:57 CEST] <HoloPed> You mean that even at h264 compression the data is more than the USB2 bandwidth ?
[15:12:59 CEST] <thebombzen> my webcam drops to 10fps to send rawvideo. which is why I use mjpeg from mine.
[15:13:27 CEST] <BtbN> The hardware encoders for h264 they put into those webcams are just bad.
[15:13:46 CEST] <BtbN> It's basically mpeg2, if not worse.
[15:13:57 CEST] <HoloPed> Is it possible to encode h264 in realtime ?
[15:14:01 CEST] <HoloPed> or does that require hardware
[15:14:06 CEST] <pZombie> holoped yes, it is
[15:14:18 CEST] <BtbN> if you have a somewhat decent CPU.
[15:14:21 CEST] <pZombie> holoped did you try to capture with virtualdub?
[15:14:26 CEST] <HoloPed> not yet
[15:14:30 CEST] <HoloPed> good idea
[15:14:32 CEST] <furq> you would need a pretty weak cpu to not be able to encode in realtime
[15:14:54 CEST] <HoloPed> I need to do it in code, eventually
[15:15:01 CEST] <HoloPed> I was looking at AForge,
[15:15:03 CEST] <BtbN> 1080p in real time does require quite a bit of CPU power.
[15:15:09 CEST] <HoloPed> it has an FFMPEG wrapper
[15:15:18 CEST] <pZombie> you have to disable "remove unchanged frames" under screen capture settings in virtualdub, or it will kill all performance for some magical reason
[15:15:38 CEST] <furq> screen capture?
[15:15:43 CEST] <pZombie> then use 264vfw to capture in realtime
[15:16:02 CEST] <pZombie> i am referring to virtualdub if he was to try that for his webcam
[15:16:19 CEST] <furq> it sounds like you're telling him to screen capture a webcam
[15:16:23 CEST] <furq> i hope that's not right
[15:16:41 CEST] <HoloPed> its ok, I got his gist :)
[15:16:48 CEST] <HoloPed> what is 264vfw ?
[15:17:30 CEST] <furq> the vfw version of x264
[15:17:40 CEST] <furq> if you're unfortunate enough to be stuck using vfw
[15:17:47 CEST] <HoloPed> yes yes
[15:17:52 CEST] <HoloPed> windows sucks etc etc
[15:17:57 CEST] <HoloPed> I am
[15:17:59 CEST] <HoloPed> stuck with it
[15:18:07 CEST] <furq> you're not stuck with vfw if you can get ffmpeg to work
[15:18:22 CEST] <HoloPed> oh, I thought ffmpeg used vfw
[15:18:31 CEST] <furq> thankfully not
[15:18:32 CEST] <pZombie> furq well, in case he also wanted to capture the screen i guess. That setting was reducing my max possible realtime x264 capture framerate in 2560x1440 when using ultrafast to below 30 vs 160 without
[15:18:38 CEST] <HoloPed> it uses v4l ,  no?
[15:19:06 CEST] <furq> well i guess it uses vfw for getting the feed off the webcam
[15:19:24 CEST] <furq> but then it sensibly doesn't use it for the rest of the process
[15:19:44 CEST] <HoloPed> very sensibly
[15:20:36 CEST] <furq> oh
[15:20:40 CEST] <furq> https://trac.ffmpeg.org/wiki/Capture/Webcam
[15:20:45 CEST] <furq> i guess you can use directshow directly instead
[15:21:23 CEST] <furq> https://trac.ffmpeg.org/wiki/DirectShow
[15:21:39 CEST] <HoloPed> so how do I get virtualbug to capture the h264 stream
[15:21:43 CEST] <HoloPed> I want to see how bad is it
[15:22:16 CEST] <furq> do people still use virtualdub
[15:22:28 CEST] <pZombie> i do :D
[15:22:38 CEST] <furq> can it output to a container which isn't garbage yet
[15:22:49 CEST] <furq> or is it still hacking everything into avi
[15:22:59 CEST] <pZombie> avi :/
[15:23:22 CEST] <HoloPed> ok guys I need you to focus
[15:23:28 CEST] <furq> HoloPed: read that directshow article
[15:23:38 CEST] <pZombie> but contrary to others, it gives you a lot of information about how many frames were dropped, inserted, lagtimes etc
[15:23:41 CEST] <HoloPed> dis: https://trac.ffmpeg.org/wiki/DirectShow
[15:23:42 CEST] <HoloPed> ?
[15:23:44 CEST] <furq> yes
[15:25:41 CEST] <pZombie> if he does not get any devices listed when doing ffmpeg -list_devices true -f dshow -i dummy , can he capture at all?
[15:26:00 CEST] <furq> not as far as i can tell
[15:26:07 CEST] <HoloPed> I do get devices listed
[15:26:34 CEST] <pZombie> what's the output of that ?
[15:26:43 CEST] <HoloPed> it shows the camera
[15:26:52 CEST] <HoloPed> I can do this:
[15:26:53 CEST] <HoloPed> ffmpeg -f dshow -list_options true -i video="Logitech HD Pro Webcam C920"
[15:26:56 CEST] <HoloPed> and I get a list of formats
[15:27:07 CEST] <furq> sounds like you're good then
[15:27:09 CEST] <HoloPed> but no h624 ub tgere
[15:27:15 CEST] <HoloPed> but no h624 in there *
[15:27:30 CEST] <furq> what options do you get
[15:28:00 CEST] <HoloPed> furq, http://pastebin.com/nVfuKcuz
[15:28:32 CEST] <BtbN> you don't want to use the h264 it produces anyway, so mjpeg is fine.
[15:28:49 CEST] <furq> it claims it can do bgr24 at 1080p30
[15:28:54 CEST] <furq> that doesn't seem likely to work over usb2 though
[15:28:56 CEST] <pZombie> why not bgr24 and encode it into h264, wouldn't that be the best quality?
[15:29:01 CEST] <pZombie> oh, usb2
[15:29:05 CEST] <furq> assuming it is usb2
[15:29:16 CEST] <BtbN> it claims to support uncompressed 1080p with 30 fps though
[15:29:21 CEST] <BtbN> but that seems way too much for usb2
[15:29:29 CEST] <BtbN> Or the driver decodes mjpeg for you...
[15:30:16 CEST] <HoloPed> so I guess it doesn't expose the h264 ?
[15:30:21 CEST] <HoloPed> its not on that list
[15:30:25 CEST] <HoloPed> is that a right conclusion ?
[15:30:30 CEST] <HoloPed> would it be on the list if it did ?
[15:30:38 CEST] <BtbN> I think dshow just doesn't support h264
[15:30:46 CEST] <BtbN> Via v4l it supports h264
[15:30:52 CEST] <BtbN> but seriously, you don't want to use that for anything.
[15:30:56 CEST] <HoloPed> ok
[15:31:21 CEST] <HoloPed> I need to test if I can encode in realtime and still have enough cycles to actually drive the application around it
[15:31:26 CEST] <furq> if the yuv420p stream actually works at 30fps then use that
[15:31:29 CEST] <HoloPed> can I display the video on screen while I encode it ?
[15:31:29 CEST] <furq> otherwise use the mjpeg stream
[15:31:53 CEST] <pZombie> Holoped which CPU you use?
[15:32:03 CEST] <HoloPed> I will buy a computer for this project
[15:32:23 CEST] <HoloPed> it's needs to be small though, it will be installed into a fixture
[15:32:29 CEST] <BtbN> Get a NUC.
[15:32:29 CEST] <HoloPed> think video booth thingy
[15:32:35 CEST] <HoloPed> what is a NUC ?
[15:32:38 CEST] <BtbN> fast
[15:32:44 CEST] <BtbN> If you buy the Skylake ones
[15:32:53 CEST] <BtbN> it can easily encode and play stuff
[15:32:59 CEST] <furq> you're mostly paying for the form factor with those aren't you
[15:33:08 CEST] <furq> if mini-ITX is small enough then that'll probably be cheaper
[15:33:11 CEST] <BtbN> The non-skylake ones are quite ok
[15:33:17 CEST] <furq> as long as you get a decent cpu ofc
[15:33:28 CEST] <BtbN> But the Skylake ones have a realy fast CPU, the price seems fair.
[15:33:29 CEST] <haasn> I have a video that's 30 fps but encoded as 60 fps (every frame is duplicated), is there a video filter that will help me undo this?
[15:33:50 CEST] <BtbN> -r 30?
[15:33:57 CEST] <furq> if every frame is duplicated then -vf fps=30 should work
[15:34:04 CEST] <furq> or -r 30 which is the same thing iirc
[15:34:07 CEST] <haasn> maybe dejudder or something
[15:34:10 CEST] <haasn> ah
[15:34:19 CEST] <HoloPed> how much are those NUC Skylakes ?
[15:34:24 CEST] <haasn> well I'm looking for something that I can use with mpv's vf_lavfi
[15:34:29 CEST] <BtbN> 300-400¬
[15:34:35 CEST] <HoloPed> ok, that would work
[15:34:37 CEST] <HoloPed> thanks guys
[15:34:39 CEST] <BtbN> A Braswell NUC might also be fast enough, they are way cheaper.
[15:34:44 CEST] <BtbN> Like, NUC5PPYH
[15:35:00 CEST] <haasn> furq: ah great
[15:35:03 CEST] <pZombie> 300-400 but without memory and ssd, you got to add those extra
[15:35:05 CEST] <haasn> --vf=lavfi=fps=30 # this works!
[15:35:19 CEST] <BtbN> They also have a hardware encoder, but you have to deal with vaapi to use them.
[15:35:22 CEST] <haasn> now I can watch this without getting a headache :)
[15:35:38 CEST] <BtbN> you should not see any visual difference if your monitor is running at 60Hz
[15:35:47 CEST] <BtbN> It will just re-duplicate the frames on display.
[15:38:44 CEST] <haasn> BtbN: No, I'm using mpv's interpolation option which blends the in-between frames
[15:38:58 CEST] <haasn> But to do that, the source needs to be correct
[15:44:09 CEST] <pZombie> the intel skull canyon NUC is an amazing little machine drawing a total of only 75 watts at maximal load
[15:44:19 CEST] <pZombie> i just read this review http://www.pcworld.com/article/3074962/computers/intel-nuc6i7kyk-review-this-skull-canyon-nuc-smashes-all-mini-pc-preconceptions.html
[15:45:09 CEST] <pZombie> the iris pro 580 is the fastest APU currently. Good enough for some casual gaming i guess if you are no hardcore gamer
[15:45:36 CEST] <jnorthrup> i bought one for my boy, hes happy with it for games
[15:45:48 CEST] <pZombie> but i consider it as a lost chance for a completely fanless design
[15:46:11 CEST] <pZombie> jnorthrup - does the fan of it make a lot of noise when under max load?
[15:46:21 CEST] <jnorthrup> haha 75 watts fanless would be a trick
[15:46:28 CEST] <jnorthrup> afaic no noise
[15:47:13 CEST] <pZombie> no trick at all. Just a nice big aluminum heatsink on top of it
[15:47:17 CEST] <jnorthrup> dont take my word ive never seen it under load
[15:47:44 CEST] <pZombie> the 75 watts is for the complete system. The cpu is max 45 watts
[15:48:50 CEST] <pZombie> i guess they wanted to keep the form factor small and pretty. But a second fanless option would have been nice indeed
[15:49:54 CEST] <furq> 45W is quite a lot in that form factor
[15:50:28 CEST] <pZombie> yes, but it's a monster skylake cpu inside
[15:50:48 CEST] <pZombie> which has the currently fastest apu built in on top
[15:52:30 CEST] <jnorthrup>         my 10 year old son doesn't know enough about it to understand that it's only running an apu
[15:52:36 CEST] <pZombie> you pay more money for this than if you would invest into a gaming computer, but the form factor is a plus, as well as your power bill will make up for it in the long run
[15:53:52 CEST] <pZombie> and you can later add a thunderbolt/mini pcie docking station if you ever wanted to connect some high end GPU to it
[15:53:59 CEST] <pZombie> but those are currently way overpriced
[15:54:17 CEST] <pZombie> they will drop to around $200 in the near future is my guess
[15:58:54 CEST] <Illya> The 1070 looks nice for a cheap, but solid GPU
[15:59:16 CEST] <pZombie> not cheap at all, but very nice, yes
[15:59:26 CEST] <pZombie> for gaming
[15:59:35 CEST] <furq> are they out yet
[15:59:40 CEST] <pZombie> yes
[15:59:44 CEST] <furq> oh so they are
[15:59:55 CEST] <pZombie> their double precision performance is castrated however
[15:59:56 CEST] <furq> yeah these are not cheap
[16:00:18 CEST] <pZombie> if you ever need to use it for compute which requires double precision, this is not the card to get
[16:00:53 CEST] <furq> i wonder if the 9 series prices have dropped
[16:01:06 CEST] <furq> it'd make a nice change
[16:01:06 CEST] <yagiza> Any iLBC specialis around?
[16:04:02 CEST] <pZombie> if you wait a few weeks, the 1070 prices will drop too, after the initial hype is over
[16:04:13 CEST] <furq> i assume they'll drop after the 480 comes out
[16:09:51 CEST] <BtbN> They'll drop once the non-founders editions start appearing.
[16:10:00 CEST] <BtbN> Nvidia set a fixed price for those.
[16:10:16 CEST] <pZombie> they say the rx 480 might be sold at around $200, so prices will have to drop a lot if true
[16:10:27 CEST] <BtbN> The rc480 isn't nearly as fast though
[16:10:36 CEST] <BtbN> it's more like a 970?
[16:10:38 CEST] <pZombie> as the 480 is between the gtx 970 and the 980 which sell for much more
[16:10:54 CEST] <pZombie> yes, but prices will have to drop nevertheless if they sell the 480 for that little
[16:11:00 CEST] <BtbN> There will probably be a 1050 and 1060 to answer those cards.
[16:11:14 CEST] <furq> if $200 actually gets converted properly then that's the same price as a 960
[16:11:49 CEST] <furq> but it's been ages since i've seen a retailer actually drop the price of the previous gen cards
[16:12:13 CEST] <furq> i think they just count on people who don't follow the news buying all the remaining stock
[16:16:36 CEST] <pZombie> it even draws less power than the gtx 970 which makes sense, given it is 14nm
[16:16:38 CEST] <BtbN> https://geizhals.de/?phist=1299709
[16:16:47 CEST] <BtbN> they are dropping. A lot.
[16:17:03 CEST] <pZombie> so why would one go for a gtx 970 instead of the rx 480, unless he is using a g-sync monitor
[16:17:19 CEST] <BtbN> One would go for the 1050 instead of the rx480
[16:17:27 CEST] <BtbN> They should appear soon
[16:18:19 CEST] <haasn> I can't wait to see if the new AMD cards will be usable with Linux
[16:18:28 CEST] <haasn> Using the GTX 970 right now since nothing else works well
[16:18:55 CEST] <pZombie> so amd drivers on linux aren't that good still?
[16:19:41 CEST] <haasn> the proprietary drivers are a joke
[16:19:48 CEST] <haasn> The free drivers seem to continuously be getting better though
[16:20:28 CEST] <haasn> Last time I tried them they were still too slow for realtime playback in mpv :(
[16:20:54 CEST] <pZombie> doesn't sound right
[16:22:02 CEST] <pZombie> i remember long long ago using a laptop with an amd card built in, playing WoW inside wine, and it was actually faster than on windows, mainly because i was using a very lightweight linux and this laptop had only 512mb RAM
[16:22:20 CEST] <pZombie> no problems at all
[16:22:21 CEST] <BtbN> WoW used to run on OpenGL though.
[16:22:30 CEST] <haasn> WoW + OpenGL is a joke last time I checked
[16:22:42 CEST] <haasn> WoW + DirectX via WINE was unplayably slow last time I checked
[16:22:43 CEST] <BtbN> Since they came up with the DX11 renderer, they abandoned it.
[16:22:50 CEST] <BtbN> Same for DX9
[16:23:14 CEST] <haasn> (has WINE+DX11 ever worked?)
[16:23:15 CEST] <BtbN> The strange thing is, on OSX, they have an up to date OpenGL renderer with all features the Windows-Client only has via DX11.
[16:23:29 CEST] <haasn> yeah
[16:23:40 CEST] <haasn> but Blizzard has never really cared about Linux, nor about writing good software
[16:23:45 CEST] <haasn> And now they don't even care about making good games anymore :)
[16:24:06 CEST] <pZombie> but starcraft 2 supposedly works perfectly fine with wine
[16:24:12 CEST] <pZombie> how do you explain that one?
[16:24:15 CEST] <haasn> Diablo III worked in WINE too
[16:24:19 CEST] <haasn> what do you mean
[16:24:33 CEST] <pZombie> isn't sc2 dx11 too by now?
[16:25:08 CEST] <haasn> Usually people tag stuff as platinum on WINE appdb if they can play the game
[16:25:34 CEST] <haasn> To me, platinum should mean every single graphical option works, looks the same as on windows, and FPS is at least as good as windows
[16:25:41 CEST] <haasn> otherwise it's a subpar experience
[16:26:01 CEST] <pZombie> WoW is tagged as platinum currently
[16:26:07 CEST] <pZombie> so it has to be working well
[16:26:14 CEST] <haasn> Like I said
[16:26:19 CEST] <haasn> people usually tag it as platinum if they can play the game
[16:28:20 CEST] <pZombie> https://www.youtube.com/watch?v=0YcqGvHuqrA  looks fine to me.
[16:29:23 CEST] <haasn> that doesn't look like max settings to me, does he show the graphics options any where in the video?
[16:29:29 CEST] <haasn> ah, at the beginning
[16:29:40 CEST] <haasn> yeah do you see those big yellow exclamation marks at the right
[16:29:44 CEST] <haasn> that means the higher settings were disabled
[16:30:00 CEST] <haasn> probably because he's using OpenGL instead of D3D
[16:30:09 CEST] <pZombie> i see
[16:30:16 CEST] <haasn> and this is what I mean
[16:30:26 CEST] <pZombie> now i get it
[16:30:38 CEST] <pZombie> so you can pick opengl still
[16:30:39 CEST] <haasn> people never understand what I mean by on par with windows
[16:30:55 CEST] <haasn> hard for me to tell, I don't read russian
[16:31:00 CEST] <haasn> but there is a setting somewhere for the renderer
[16:31:05 CEST] <haasn> you could also e.g. pick a lower directx version on windows
[16:31:09 CEST] <pZombie> if blizzard did not care about linux gaming at all, they would have disabled the opengl option altogether i suppose
[16:31:33 CEST] <haasn> they care about OS X
[16:31:49 CEST] <haasn> I don't see D3D running on OS X any time soon
[16:36:23 CEST] <pZombie> all it would take to get gamers to linux would be 1 title like WoW that is exclusive for linux only
[16:37:06 CEST] <furq> is super tux kart not good enough for you
[16:37:24 CEST] <pZombie> it's too super good. Overkill
[16:40:57 CEST] <haasn> Like 80% of the games I care about already run on Linux
[16:41:00 CEST] <haasn> I just want the other 20% to catch up :)
[16:43:01 CEST] <pZombie> some older good games like "the settlers IV" might not run on newest windows version anymore, but run in wine
[16:43:04 CEST] <pZombie> on linux
[16:44:55 CEST] <kepstin> ut99 can run better in wine on linux (with directx renderer) than either natively on modern windows, or natively on linux :/
[16:47:19 CEST] <pZombie> yep, same experience i had many years ago running wow inside wine on a laptop
[16:47:27 CEST] <pZombie> more fps, less loading
[16:47:36 CEST] <pZombie> wine is quite a mystery to me
[16:49:12 CEST] <pZombie> you can compile the linux kernel to a higher 1000hz frequency as well. While this means more context switching and more stress on the CPU, it might also mean smoother animations if your CPU can handle the load
[16:49:21 CEST] <pZombie> also less input lag
[16:49:58 CEST] <haasn> kepstin: weird. It ran fine natively on linux for me either way
[16:50:40 CEST] <pZombie> linux in theory could be far better for gaming than windows
[16:50:56 CEST] <kepstin> haasn: last time I tried, the opengl renderer (used in native linux version) had issues with game speed when running on systems with either dynamic cpu speed or multiple cores (forget which). Particularly bad on network games.
[16:51:13 CEST] <kepstin> dunno if that might have been patched at some point :/
[16:52:32 CEST] <__jack__> pZombie: in practice too; bad software will always be junk
[16:54:15 CEST] <pZombie> gaming keyboards and gaming mice with extra buttons are no problem either, if you buy smart, as in buying keyboards/mice which store their settings on their own permanent memory
[16:54:37 CEST] <pZombie> even if there are no drivers, you can once set them up in windows and keep the settings for linux
[16:56:32 CEST] <haasn> I just bought a Zowie FK
[16:56:37 CEST] <haasn> works without proprietary drivers
[16:57:03 CEST] <pZombie> of course it does. It does not have any extra keys like some mmo mice
[16:57:53 CEST] <pZombie> setting up an mmo mouse in linux without drivers however, is doable, but not for the average joe
[16:58:29 CEST] <haasn> meh, I use twice as many keybinds as the average MMO player and I've never have a problem with keybindings :p
[16:58:58 CEST] <haasn> especially now that Blizzard basically took all of the complexity out of WoW
[17:01:11 CEST] <pZombie> some people are keyboard acrobats. I keep hitting the wrong keys if there is fast action when only using the keyboard. Plus, you have only 2 hands. You cannot move around with aswd at the same time while casting 1232190 different spells in an mmo
[17:01:29 CEST] <haasn> I can :)
[17:01:51 CEST] <pZombie> no, it is physically impossible, unless your wiener is that long it can reach the keyboard
[17:02:34 CEST] <furq> i never had an issue with that playing wow
[17:02:41 CEST] <furq> unless it's become much more complex since 2006
[17:03:18 CEST] <furq> i had 20+ easily-accessible binds though
[17:04:13 CEST] <furq> it used to drive me insane watching my flatmate play with no binds at all
[17:06:30 CEST] <haasn> I must have used over 50
[17:07:39 CEST] <furq> i probably had that many buttons, but most of them were binds i didn't need to use in combat
[17:07:50 CEST] <furq> i usually wouldn't bother binding those
[17:09:07 CEST] <haasn> I was running out of space in Bartender4 despite having every single bar enabled; I also ran out of character-specific macros so I had to steal some of my account-wide macro slots for my main
[17:09:40 CEST] <haasn> I used every even remotely accessible key, generally including modifiers (e.g. I would have E, shift+E, ctrl+E and alt+E bound separately)
[17:10:12 CEST] <furq> i had 1-5, f1-5, alt+1-5, plus the keys around wasd
[17:10:18 CEST] <haasn> I made a list once
[17:10:28 CEST] <haasn> as a baseline I used 1 2 3 4 5, Q W E T, F G, \ Z X C V
[17:10:37 CEST] <haasn> with shift: 1 4 V Z X F E R 5 space
[17:10:40 CEST] <haasn> with ctrl: R E
[17:10:43 CEST] <haasn> with alt: X Z C 1
[17:10:48 CEST] <haasn> also some misc utility stuff on F1-F9
[17:11:05 CEST] <haasn> rest were bindings for interface elements
[17:11:18 CEST] <haasn> (e.g. shift+C for character sheet, b for bags, n for talents and so on)
[17:11:18 CEST] <kepstin> I just play games that don't require me to program a new game user interface :/
[17:11:24 CEST] <haasn> meh
[17:11:29 CEST] <haasn> to me programming the interface was the only fun thing in wow
[17:11:33 CEST] <haasn> the actual raiding was like 5% of it
[17:11:36 CEST] <furq> i think i had tabs for inventory and everything else i didn't care
[17:11:40 CEST] <haasn> the other 95% were preparing myself for raids
[17:11:58 CEST] <furq> s/tabs/tab/
[17:12:31 CEST] <furq> it might be worrying that i can still remember this much about my ui setup from a game i've not played for 10 years
[17:12:58 CEST] <haasn> In a bout of extreme stupidity I managed to forget to back up my Screenshots folder before nuking my windows partition so I don't have anything to show, sadly :(
[17:13:09 CEST] <haasn> (I did back up the WTF folder, I also checked it into git :p)
[17:13:55 CEST] <haasn> installing addons was the most fun in the game
[17:13:58 CEST] <furq> speaking of lua scripting, i should check out what mpv uses it for
[17:14:07 CEST] <haasn> mpv uses it for pretty much any user-driven behavior
[17:14:26 CEST] <furq> can you modify the look of the osd with it
[17:14:31 CEST] <haasn> the entire OSD is written in lua
[17:14:36 CEST] <furq> neat
[17:14:52 CEST] <haasn> pretty much the only thing you can't do in lua is extend the renderer
[17:15:00 CEST] <haasn> (although I was thinking about possible ways to implement that)
[17:15:00 CEST] <furq> hopefully it's not horrible i-wish-i-was-writing-javascript lua like a lot of lua scripted apps
[17:15:11 CEST] <haasn> oh man
[17:15:15 CEST] <haasn> never look at the lua source code of wow addons
[17:15:18 CEST] <haasn> you'll go insane
[17:15:23 CEST] <furq> i've seen plenty of it from idling in #lua
[17:15:24 CEST] <pZombie> watching koreans with high APM playing sc2 cracks me up. How do they even reach that kind of speed and precision with the keyboard is beyond me. Like in this video https://www.youtube.com/watch?v=-ZLKBO8y6CY
[17:15:26 CEST] <furq> it looks dreadful
[17:15:38 CEST] <furq> functions taking 47 callbacks and shit
[17:16:27 CEST] <furq> and that's just the actual API provided by the professional programmers at blizzard
[17:16:34 CEST] <haasn> pZombie: from what I've heard, it's pretty important to make sure important functions are not near each other, since you can and will accidentally press the key next to the one you intended to press
[17:17:13 CEST] <haasn> furq: the best part about the lua API is the hoops they go through to make it so you can't write a bot in it
[17:17:18 CEST] <pZombie> the guy in the video has alien hands, like they were made for playing sc2
[17:18:19 CEST] <haasn> furq: it's actually kind of fascinating, they basically slapped a mandatory access control system onto lua to make it so that data and code paths can live in different levels of execution rights
[17:19:30 CEST] <furq> pZombie: https://www.youtube.com/watch?v=zHI4Ycx5b5g&t=3m20s
[17:19:32 CEST] <furq> this is more like it
[17:20:09 CEST] <pZombie> not even close
[17:21:05 CEST] <pZombie> that's just a very high skill quick thinker tetris player
[17:21:08 CEST] <haasn> the downside of the protected/tainted model that blizzard used is that if an addon running in an unprivileged code path ends up writing to something that a privileged code path would need to access, that privileged code path will itself become tainted. So as a result, it's extremely easy for an addon to break some component of the official interface simply by accidentally writing to it
[17:22:39 CEST] <haasn> pZombie: going back to what you said about moving around and casting spells at the same time: one of the tricks I abused to get around it was simply jumping. If you move in a direction and jump, you will continue to move in that direction, allowing you to take your finger off the corresponding direction key for long enough to queue up the next spell. (I played with a spell queue, which is basically
[17:22:42 CEST] <haasn> blizzard's built-in latency compensation feature except artificially turned up to the max)
[17:23:14 CEST] <haasn> (so I could press a spell up to 400ms in advance of when I'd actually be able to cast it)
[17:24:00 CEST] <haasn> Another technique was rapid camera shifts. If you turn the camera, you can change which key you need to press to go into the direction you want, freeing up a finger to press a spell
[17:24:05 CEST] <pZombie> haasn yes, but maybe you did not want to jump, but move around randomly or in another direction, which you cannot do effectively when two of your fingers are busy hitting some shortcuts
[17:24:59 CEST] <pZombie> an mmo mouse however adds other disadvantages, so not sure if that is right path either
[17:25:22 CEST] <furq> you mean disadvantages like getting the shit kicked out of you
[17:25:31 CEST] <pZombie> lol
[17:25:49 CEST] <furq> http://assets.razerzone.com/eeimages/products/13785/razer-naga-2014-right-03.png
[17:25:53 CEST] <furq> even i would beat someone up if i saw them using that
[17:25:55 CEST] <haasn> right, as a PvE player though most of my movement was get from where you are to point X as fast as possible so the direction was pretty much constant. I imagine for PvP it's different
[17:26:12 CEST] <furq> did you really need 50 binds for pve
[17:26:18 CEST] <haasn> I think the moving around randomly is mostly a melee thing
[17:26:34 CEST] <furq> my experience of raiding as a priest was "press 1 until you wipe, then press f1, repeat"
[17:26:37 CEST] <haasn> Melee players always seem to love to do that in order to make it hard for spellcasters to turn in their direction
[17:26:55 CEST] <furq> you're trying to get behind then so that targeting fails
[17:26:57 CEST] <furq> them
[17:27:01 CEST] <pZombie> haasn yep, which brings us to darkfall, a pvp mmo with tons of spells.. back then
[17:27:03 CEST] <haasn> furq: well, define need
[17:27:13 CEST] <haasn> furq: do you need 50 binds to go into LFR and get free loot from the loot mill? no
[17:27:37 CEST] <furq> i don't think that acronym existed when i was playing
[17:27:47 CEST] <pZombie> haasn in darkfall you had to move like in an fps game, with real targeting, and on top of that, you had a gazillion spells to cast and abilities to use in combat
[17:27:51 CEST] <haasn> for me the only point in playing was worldoflogs
[17:27:55 CEST] <haasn> if I wasn't #1 dps I was losing
[17:28:53 CEST] <haasn> pZombie: heh, sounds interesting
[17:29:17 CEST] <pZombie> it was the best game i ever played but the developer team was retarded and killed it
[17:29:29 CEST] <haasn> I guess personally, if I want to try and hone my aimbot skills, I'd play ut2004 instagib
[17:29:39 CEST] <haasn> WoW for me was mostly about micro-management
[17:29:53 CEST] <pZombie> now they sold the license to someone else and they are in alpha, soon to relaunch it. Not sure if i want to go through all it again however
[17:29:56 CEST] <haasn> my interface is filled to the brim with information that I had to be processing at all points in time to make the right decision
[17:30:22 CEST] <pZombie> it's also a very stressing game, since there is full loot and no safe zones
[17:30:42 CEST] <haasn> it's not really a game about aiming or timing or speed, it's a game about processing lots and lots of information at once
[17:30:50 CEST] <haasn> WoW that is
[17:31:31 CEST] <haasn> furq: did you play during vanilla? :p
[17:31:39 CEST] <haasn> that sounds a lot like some classes in the past..
[17:31:49 CEST] <furq> yes
[17:31:54 CEST] <furq> i only played during vanilla
[17:32:00 CEST] <haasn> ah, makes sense
[17:32:23 CEST] <haasn> WoW has come a long way. Unfortunately, it's currently regressing past to its vanilla state
[17:32:31 CEST] <haasn> s/past/back/
[17:32:32 CEST] <furq> priest during raids was "target someone, flash heal until the boss/group is dead"
[17:32:54 CEST] <furq> warrior was slightly more interesting unless you were main tanking
[17:33:26 CEST] <haasn> Nah, WoW these days is basically a completely different game
[17:33:37 CEST] <pZombie> the actual fun part was searching and finding the right strategy
[17:33:39 CEST] <furq> i gathered that when you mentioned builtin macros
[17:34:00 CEST] <furq> vanilla was great fun in pvp though
[17:34:10 CEST] <furq> unless you did the organised pvp arenas which were dumb
[17:37:05 CEST] <haasn> furq: nowadays there isn't a spec you can play properly without consulting spreadsheets
[17:37:08 CEST] <haasn> world of spreadsheets :p
[17:39:13 CEST] <furq> sounds like great fun
[17:39:15 CEST] <haasn> furq: https://youtu.be/Vx6ipbVOWvY !
[17:40:01 CEST] <haasn> raiding in 2016^
[17:41:21 CEST] <pZombie> it was much much more fun back in the old days when i launched and this type of game was relatively new to most
[17:42:18 CEST] <pZombie> nowadays when a new instance is opened, people go through it in just a few days
[17:49:06 CEST] <pgorley> how does avcodec_decode_video2 decide which pixel format the decoded frame should be?
[17:49:39 CEST] <haasn> pZombie: the problem is that people have gotten too good. They all know the bosses by heart from doing them on the PTR forever and ever. They stay up for 20 hours days for days on end, racing to become world first
[17:49:47 CEST] <haasn> The best of the best of the best guilds are picking the best of the best of the best players
[17:49:55 CEST] <haasn> If you're not good enough, they have 20 other players waiting to replace you
[17:50:00 CEST] <haasn> How do you design content that will challenge those players?
[17:50:22 CEST] <haasn> They kill the end boss with items that are like 20 levels below what the boss was tuned to be beatable at
[17:50:39 CEST] <pZombie> haasn exactly, they have gotten too good and new content has become routine. The early days when everything had to be explored and figured out were the best
[17:51:56 CEST] <pZombie> haasn you cannot, which is a shame. Unless you had a machine to wipe out everyone's memories so they can experience the early days again
[17:53:14 CEST] <bencoh> ut2k4 instagib was fun indeed
[17:53:44 CEST] <bencoh> (I wonder how you ended up discussing this on #ffmpeg though :])
[17:53:54 CEST] <furq> quakeworld for life
[17:54:20 CEST] <bencoh> furq: xonotic? :)
[17:54:28 CEST] <furq> is that what warsow turned into
[17:54:56 CEST] <furq> oh no, it's nexuiz
[17:55:20 CEST] <furq> is this the one where they added the bad caricature of bunnyhopping where you just hold down jump and magically go faster
[17:55:21 CEST] <bencoh> no, that's one of the opensource quake derivatives (prolly the most achieved one I'd say)
[17:55:28 CEST] <bencoh> yeah nexuiz
[17:55:53 CEST] <furq> instead of going faster by letting go of forward, strafing left and turning left (just like in real life)
[17:59:29 CEST] <haasn> pZombie: the thing that kept me entertained was having to continuously figure out how to min-max every single boss
[17:59:56 CEST] <haasn> one of the things that I absolutely loved about the new WoW was, for example, the new talent system. For the first time, you could realistically change your talents and glyphs for every single encounter
[18:00:10 CEST] <haasn> Most players didn't really bother, but for me it was one of the things that was the most fun
[18:00:23 CEST] <haasn> Being able to tailor my spec, build and playstyle to every single boss
[18:00:51 CEST] <haasn> for top guilds they even tailor their classes and roster for every boss
[18:01:40 CEST] <haasn> furq: xonotic is the one with the hold-space-and-move-forwards-to-bunnyhop thing, yeah
[18:01:55 CEST] <haasn> but I thought that was a great decision. Makes it accessible; it's something I can play at LAN parties
[18:01:58 CEST] <pZombie> i didn't get to that point. Left much earlier when everything became routine :D
[18:01:59 CEST] <haasn> and everybody finds the bunny hopping fun
[18:02:19 CEST] <bencoh> especially the ones aiming for bunnies? :D
[18:02:23 CEST] <pZombie> i was hopping for the starcraft mmo to jump back in, but they ditched this
[18:02:29 CEST] <haasn> quake and unreal are more of the person who plays it 24/7 will 1v10 everybody else
[18:02:41 CEST] <haasn> which is also fun, but only fun for the one player that has mastered it :p
[18:02:46 CEST] <bencoh> not so true about unreal
[18:02:56 CEST] <bencoh> quake ... prolly yeah
[18:03:21 CEST] <furq> the movement is about 90% of what i like about quake
[18:03:40 CEST] <haasn> what I liked most about unreal was vehicle CTF on custom maps
[18:03:46 CEST] <haasn> (rembo's server anybody?)
[18:04:14 CEST] <pZombie> i remember reaching a level in quakeworld CTF where i would actually dodge peoples' rockets like neo in the matrix and was moving so unpredictable, i would take down people with the shotgun while they were failing to hit me
[18:04:20 CEST] <haasn> I miss the days when you could get an 8v8 going 24/7 with 3 spectators waiting to play..
[18:04:30 CEST] <pZombie> it was like i became the character itself
[18:04:36 CEST] <haasn> I logged in on that server a few weeks ago, it was just me and a few bots. Then eventually one person joined, and left after I killed him a few times :/
[18:05:16 CEST] <pZombie> i totally loved that grappling hook in CTF games
[18:05:33 CEST] <haasn> pZombie: my favorite bit about ut2004 was the movement as well; the custom maps were filled to the brim with secret passageways and tricks that you can only reach by mastering the movement system
[18:05:34 CEST] <pZombie> not sure how the mod was called, axes4all i think
[18:06:05 CEST] <bencoh> haasn: haven't checked ut2k4 for years ... so nobody plays it anymore?
[18:06:08 CEST] <haasn> when I got bored of the constant action I used to load up the popular maps in custom games and spend hours practicing the jumps
[18:06:11 CEST] <haasn> nope :/
[18:06:14 CEST] <bencoh> :(
[18:25:40 CEST] <P4Titan> Hello all
[18:26:16 CEST] <P4Titan> I wish to create an ouptut codec context of pcm_f32be
[18:26:21 CEST] <P4Titan> but its frame size is 0
[18:26:45 CEST] <P4Titan> after initializing it with the avcodec_open2
[18:56:14 CEST] <sangy> Hello, I"m trying to compile ffmpeg with --enable-libiec61883. It complains stating that I don't have it installed, but I do. Any ideas on how to debug this?
[18:57:52 CEST] <sangy> It says you people might use this https://ptpb.pw/CVLt
[18:59:29 CEST] <sangy> Nvm, I studied it and it seems I need lavc1394 and lrom1394?
[19:00:56 CEST] <sangy> That was it. It might be obvious for the weathered AV/programmer, but would a patch to improve this error message help?
[19:01:42 CEST] <P4Titan> I mean, submitting that patch for reviewal wouldn't hurt right
[19:02:58 CEST] <sangy> P4Titan: sure, I'll give the source an eye later :) Thanks!
[19:03:13 CEST] <P4Titan> yep
[19:03:15 CEST] <P4Titan> np
[19:14:13 CEST] <furq> sangy: https://github.com/FFmpeg/FFmpeg/blob/master/configure#L1291
[19:15:42 CEST] <sangy> furq: I see, so would it be parsing the output for checklib a little bit?
[19:20:20 CEST] <pZombie> so i tried screen capturing with ffmpeg and am getting a large amount of duplicated frames. What could cause this?
[19:21:15 CEST] <pZombie> ffmpeg -f gdigrab -i desktop -c:v libx264 -threads 22  -r 60 -preset ultrafast -crf 20 -pix_fmt yuv444p out.mp4   , this is what i tried
[19:21:22 CEST] <pZombie> anything wrong with those settings?
[19:21:50 CEST] <P4Titan> Does anyone know what the frame_size for pcm's encode_init is 0
[19:23:18 CEST] <furq> pZombie: add -framerate 60 before -i
[19:24:01 CEST] <pZombie>  i did that, but it does not make a difference. Had this on before
[19:24:40 CEST] <pZombie> even if i set both to 30, there are duplicates still
[19:24:55 CEST] <pZombie> a lot of duplicates
[19:25:36 CEST] <P4Titan> anyone?
[19:25:45 CEST] <pZombie> it s not the CPU, even if i set laughable crf 30, there are duplicates
[19:26:10 CEST] <pZombie> is gdigrab known to be low performance?
[19:26:30 CEST] <furq> i don't think it's as efficient as dshow
[19:26:43 CEST] <furq> iirc kepstin wrote it so maybe you could ask him
[19:27:43 CEST] <kepstin> yeah, gdigrab uses some really old api on windows that has varying qualities of accelleration in different drivers
[19:27:54 CEST] <kepstin> I wouldn't expect to get full motion video out of it
[19:28:31 CEST] <kepstin> which version of windows, btw? I found that it was /really/ slow on 7, but a bit faster on 8.
[19:28:35 CEST] <pZombie> win10
[19:29:03 CEST] <kepstin> amusingly, when I was testing it, I found it was way faster in wine on linux than on actual windows :)
[19:29:06 CEST] <pZombie> gdigrab does not even manage 10 frames without giving me a few duplicates here and there
[19:29:29 CEST] <kepstin> what graphics card?
[19:29:35 CEST] <pZombie> but worst of all, when dragging windows, it's not smooth anymore. The motion looks jerky
[19:30:03 CEST] <pZombie> in vdub i can record 60fps 2560x1440 without noticing any jerkiness
[19:30:14 CEST] <pZombie> two 7970 in CF
[19:30:29 CEST] <kepstin> Ah, ATI/AMD's drivers :/
[19:30:35 CEST] <pZombie> 5960x cpu, i don't think it's the hardware
[19:30:54 CEST] <kepstin> software that captures at full framerate is probably doing a directx frontbuffer grab, which is a lot less overhead
[19:31:11 CEST] <kepstin> you're probably hitting software fallback paths on the old GDI api, so it's  just gonna suck.
[19:31:49 CEST] <pZombie> nothing wrong with amd drivers. I can record up to 160 fps x264 2560x1440p in realtime using vdub, without any duplicates or drops when going below around 100
[19:32:02 CEST] <pZombie> yes, it has to be the GDI
[19:32:30 CEST] <kepstin> so, if you want faster capture in ffmpeg on windows, you'll have to write a new capture driver using different apis :)
[19:32:48 CEST] <P4Titan> Does anyone know why the pcm codec inits the codec context with a frame size of 0
[19:32:56 CEST] <furq> pZombie: https://github.com/rdp/screen-capture-recorder-to-video-windows-free
[19:33:01 CEST] <furq> use that for capturing with dshow
[19:33:01 CEST] <pZombie> kepstin i am not sure this is true. There are probably alternatives
[19:33:40 CEST] <kepstin> my goal with the 'gdigrab' one was to get a basic screencapture driver that works on any windows system no matter what, and is fast enough for capture of simple desktop applications (educational purposes mostly)
[19:33:46 CEST] <pZombie> but now that you say it, i have actually written a library once for a java capture program, which can grab both gdi and direct2d screenshots
[19:33:51 CEST] <pZombie> not sure if that would help
[19:35:36 CEST] <pZombie> i never finished the capture program, but the code for capturing frames in direct2d (if i remember right) is there still and should be much faster
[19:36:16 CEST] <pZombie> i was sending the frames to java via a direct bytebuffer, because java's robot screenshot capture was WAAY too slow for 30 fps
[19:36:50 CEST] <pZombie> let me see if i can find the code
[19:42:39 CEST] <pZombie> damn, it's so long ago, i do not even remember how i did it
[19:42:48 CEST] <pZombie> and i cannot find the full file, searching still
[20:16:28 CEST] <pZombie> i must have deleted it at some point accidentally. Lol, i worked several days to get it working and now it's poof gone.
[20:18:38 CEST] <nick0> For some reason calling av_opt_set_sample_fmt() throws an error: "The value set by option 'sample_fmts' is not a sample format", although it compiles and the arguments are correct?
[20:25:23 CEST] <BluesKaj> can ffmpeg convert a video iso file to .mov or avi etc?
[20:26:41 CEST] <llogan> i doubt it. but you could mount the ISO file then convert the VOB files.
[20:28:28 CEST] <BluesKaj> llogan, right, didn't think of that , thanks !
[20:29:02 CEST] <llogan> you may want to look into using tools/dvd2concat
[20:29:50 CEST] <furq> oh neat
[20:29:53 CEST] <furq> how did i not know about this
[20:39:51 CEST] <kepstin> hmm, does that script handle multiple angles correctly?
[20:40:47 CEST] <kepstin> I ended up using tccat from the transcode project last time I needed dvd input to ffmpeg, since it does do angles :/
[20:44:10 CEST] <Bray90820> Mpeg 4 really takes a lot less space than mpeg 2 that's for sure
[20:46:08 CEST] <Bray90820> Just converted a 5Gb movie down to down to 900mb and I can't notice a quality difference
[20:47:21 CEST] <kepstin> Bray90820: mpeg-4? you mean mpeg-4 asp like xvid? or are you talking about mpeg-4 avc aka h.264?
[20:47:33 CEST] <kepstin> if the latter, most people don't call it mpeg-4, because that's confusing :)
[20:48:06 CEST] <Bray90820> I believe I am talking about h.264 I am talking about "-c:a copy"
[20:48:35 CEST] <Bray90820> To be more specific ffmpeg -i "Video.mkv" -c:a copy Video.mp4
[20:48:48 CEST] <furq> so x264 then
[20:49:06 CEST] <Bray90820> What's the difference?
[20:49:08 CEST] <kepstin> Bray90820: "-c:a copy" means "copy audio without re-encoding".
[20:49:15 CEST] <furq> x264 is an h.264 encoder
[20:49:31 CEST] <kepstin> with that command line, you're getting the default video codec for mp4
[20:49:45 CEST] <Bray90820> which is x.264?
[20:49:49 CEST] <furq> yes
[20:49:55 CEST] <pomaranc> x.264 does not exist
[20:50:14 CEST] <kepstin> which is h.264, and the default (and only supported in ffmpeg) h.264 encoder is x264.
[20:50:32 CEST] <kepstin> well, i guess not only supported if you include hardware encoders :)
[20:50:34 CEST] <pomaranc> that's not true
[20:50:43 CEST] <pomaranc> ok, now it's true
[20:51:07 CEST] <furq> what about openh264
[20:51:08 CEST] <Bray90820> So my converted video files are h.264?
[20:51:12 CEST] <furq> Bray90820: yes
[20:51:42 CEST] <Bray90820> That's exactly what I thought
[20:52:01 CEST] <Bray90820> And it takes a hell of a lot less space
[20:52:16 CEST] <Bray90820> Is there a way I can choose to copy something than audio track 1 tho?
[20:52:24 CEST] <Bray90820> Like audio tract 2 or 3 etc...
[20:52:30 CEST] <llogan> -map option
[20:52:41 CEST] <kepstin> Bray90820: sure, you can even copy multiple tracks if you prefer
[20:52:59 CEST] <kepstin> huh, when did openh264 support get merged? I'm not keeping on top of things ;)
[20:53:16 CEST] <llogan> -map 0:a:1 for second audio stream. 0 is input file index. a is stream specifier. 1 is stream index.
[20:54:03 CEST] <Bray90820> Now my next question if you happen to know can Plex down mix 5.1 to 2.0?
[20:55:35 CEST] <soulshock> any way to specify the size of data to read from input file when using -t and -ss ? meaning each time it reads from file, it will read for example 10MB
[20:56:34 CEST] <kepstin> soulshock: that sounds like you're asking for a solution to the wrong problem. what are you trying to do?
[20:57:07 CEST] <soulshock> my idea is to have many clients encoding chunks of the source using -t and -ss and I want to prevent the clients from preventing the disk read-ahead on the disks of the source material from working correctly
[20:57:30 CEST] <soulshock> ie. prevent having to seek excessively on the disk
[20:59:24 CEST] <kepstin> soulshock: well, the answer with the ffmpeg command-line tool is 'no', it relies on the operating system to schedule reads appropriately. If you don't trust your OS, you could write such a thing either using the api, or by having a separate program handle the reads and passing the data to ffmpeg e.g. via a pipe.
[21:00:02 CEST] <soulshock> ahh pipes yes. good idea
[21:00:03 CEST] <kepstin> (in which case, you wouldn't use -ss or -t at all, and you'd have to rely on having a stream format like mpeg-ts)
[21:00:32 CEST] <kepstin> it would be complicated :)
[21:00:58 CEST] <soulshock> hm
[21:01:21 CEST] <kepstin> probably only worth it if your source media is *really slow* to seek, e.g. optical media (blu-ray, dvd)
[21:01:28 CEST] <soulshock> my concern is that running 50 clients reading from the same file will confuse the storage
[21:01:30 CEST] <kepstin> on a hard drive, i just wouldn't bother at all
[21:01:58 CEST] <kepstin> make sure you have lots of ram so the disk caching works, done.
[21:02:08 CEST] <soulshock> it's a 50GB source file though
[21:02:26 CEST] <soulshock> xdcam mpeg hd 422 would be the source
[21:03:12 CEST] <kepstin> what container?
[21:03:15 CEST] <kepstin> mxf?
[21:03:21 CEST] <soulshock> yes
[21:03:56 CEST] <kepstin> hmm, I don't really know enough about that container format to know what you could do with it.
[21:04:15 CEST] <kepstin> does it have seek indexing? if so, the multiple ffmpegs with -t and -ss would probably work ok
[21:04:35 CEST] <kepstin> if it doesn't, you'd be better off having a single process split it into chunks, then encode the chunks
[21:04:51 CEST] <soulshock> I'm not sure
[21:05:15 CEST] <soulshock> true. that could be
[21:06:19 CEST] <soulshock> I suppose the general answer should be: don't optimize prematurely. I should do some tests and find out if it's an issue at all
[21:06:41 CEST] <kepstin> if mxf has an index table, and ffmpeg supports it, each new ffmpeg process would simply read the index table from disk cache, seek to a new spot in the file, and read from there. I'd expect the OS to handle that reasonably.
[21:07:19 CEST] <kepstin> it should be doing normal OS buffered reads, that e.g. will get linux to automatically do some readahead
[21:08:27 CEST] <soulshock> I should mention it's currently using samba shares to access the source files, though that's not set in stone
[21:08:37 CEST] <soulshock> so I suppose samba could complicate thingns
[21:09:12 CEST] <soulshock> again leading to I need to test and benchmark
[21:09:37 CEST] <kepstin> that'll be all sorts of fun. You're dependent on the io scheduling and disk caching of the server side at least as much as the client side then :/
[21:11:01 CEST] <soulshock> yeah. I'm starting to think I should let the server guys handle that side of things. they'll know tons more about it than me
[21:13:02 CEST] <kepstin> soulshock: if it ends up that the seeking doesn't really work, something like using ffmpeg, stream copy, with the 'segment' muxer to split the file into chunks as separate files might be a good pre-processing step.
[21:13:24 CEST] <soulshock> interesting idea
[21:44:57 CEST] <soulshock> any way to do aac encoding in chunks like -ss and -t and then mux it back together? with ffmpeg -f concat I get a clicking noise on every join in the muxed file
[21:45:08 CEST] <soulshock> or does aac simply not support doing that
[21:50:04 CEST] <kepstin> yeah, lossy codecs like aac, mp3, vorbis, opus, etc. all use overlapped blocks, so concatenating them isn't gonna behave that well.
[21:50:31 CEST] <soulshock> ok
[21:50:41 CEST] <soulshock> worth a shot :)
[21:50:43 CEST] <soulshock> cheers
[21:51:07 CEST] <kepstin> usually the audio encoding is sufficiently fast that I'd just do it all at once, perhaps at the same time as you're encoding the video chunks?
[21:56:27 CEST] <soulshock> yeah I do it at the same time. it's just that the aac encoder does not seem to support multi core and the machines are quad core
[21:56:46 CEST] <soulshock> whereas the x264 codec does support multi core
[21:58:16 CEST] <kepstin> huh, you're distributing each chunk to a different machine for encoding?
[21:58:23 CEST] <soulshock> yes
[21:58:26 CEST] <kepstin> I was wondering why the network fs, that explains it :)
[21:58:29 CEST] <soulshock> encoding all the chunks in parallel
[23:59:56 CEST] <Mooniac> I'm recording my screen with ffmpeg, and I always get the last 3-4 seconds cut off. Same with OBS.
[00:00:00 CEST] --- Wed Jun 22 2016


More information about the Ffmpeg-devel-irc mailing list