[Ffmpeg-devel-irc] ffmpeg.log.20190507

[burek]

[00:28:14 CEST] <another> kepstin: well newish. opus came out in 2012
[02:10:26 CEST] <DaniloPereira> hello boys
[02:10:30 CEST] <DaniloPereira> and grills
[02:13:14 CEST] <DaniloPereira> ok so
[02:13:14 CEST] <DaniloPereira> i tried to use h264_vaapi
[02:13:14 CEST] <DaniloPereira> but i don't know how to make the file smaller
[02:13:14 CEST] <DaniloPereira> and the documentatio about this is very confuse
[02:30:40 CEST] <DaniloPereira> can someone help me with vaapi?
[03:13:22 CEST] <ekiro> I'm using ffmpeg on ubuntu to convert videos to mp4, but I cannot get it to utilize 100% of my CPU (according to top). Seems to use 25-30% of my CPU. I have threads set to 0 and I tried other numbers as well but nothing gives. Any idea what's going on?
[03:27:21 CEST] <kepstin> ekiro: probably means you have a single-threaded bottleneck. what's your command line look like? ideally, paste the ffmpeg output too.
[03:41:29 CEST] <ekiro> kepstin, https://pastebin.com/1syCEwAn
[03:42:29 CEST] <ekiro> oh let me get you the ffmpeg output too
[03:46:04 CEST] <ekiro> kepstin, https://pastebin.com/bJfX6sNa
[03:48:03 CEST] <kepstin> ekiro: huh - do you even need to re-encode that video at all? you could just copy it instead, that'll be much faster: `ffmpeg -i video.mkv -c copy -movflags faststart video.mp4`
[03:49:03 CEST] <kepstin> ekiro: that said, all i can think of there is that with the selected encoding options, the encoder is doing each frame sufficiently fast that it simply never builds up much parallelism. using a slower preset or a higher res video might increase cpu usage.
[03:50:37 CEST] <ekiro> there are subtitles which i am encoding into the video, but i am doing this which i left out, which is why it would require encoding
[03:51:04 CEST] <furq> that's probably the bottleneck then
[03:51:21 CEST] <kepstin> options and filters affect this, i wanted to see the command line you were actually running
[03:51:30 CEST] <ekiro> ffmpeg -threads 0 -i video.mkv -c:v libx264 -tune animation -preset fast -profile:v high -crf: 22 -vf ass=subs.ass -c:a copy -movflags +faststart video.mp4
[03:51:36 CEST] <ekiro> thats the full command
[03:52:28 CEST] <kepstin> but yeah, the subtitle rendering is single-threaded iirc and depending on the complexity (animations, etc.) and amount of text it may or may not add a noticable slowdown.
[03:54:34 CEST] <ekiro> i see
[03:54:34 CEST] <kepstin> ekiro: if it's not using 100% cpu with those options/filter, you might as well use a slower x264 preset (e.g. medium) - it'll give a bit better quality and probably will be almost as fast.
[03:54:57 CEST] <ekiro> ill give it a try
[03:55:39 CEST] <kepstin> (well - either better quality or smaller filesize)
[03:59:28 CEST] <ekiro> getting around 20% cpu utilization with preset medium
[03:59:39 CEST] <ekiro> 15-20
[04:01:35 CEST] <ekiro> https://i.imgur.com/d34NrbC.png
[04:03:03 CEST] <ekiro> i experimented with a bunch of different settings before and it never utilized 100%
[04:03:14 CEST] <furq> are you running it with nice or something
[04:03:41 CEST] <ekiro> sorry but what is nice?
[04:03:57 CEST] <furq> nvm i forgot it says that in top
[04:04:57 CEST] <ekiro> also i installed ffmpeg from the ubuntu package, not sure if that's important or not
[04:05:23 CEST] <furq> anyway that screenshot shows 90% cpu usage, not 20%
[04:05:58 CEST] <furq> the total usage is us + sy + ni
[04:06:29 CEST] <ekiro> ahh i see. im an idiot :)
[04:06:41 CEST] <furq> idk why it would be showing 68.5% niced though
[04:07:45 CEST] <ekiro> i am using a python script that executes the ffmpeg commands
[04:07:54 CEST] <ekiro> maybe that has something to do with it?
[04:12:41 CEST] <ekiro> here is the htop output which seems to show ~90% usage https://i.imgur.com/FyCXJJF.png
[04:16:47 CEST] <furq> oh right
[04:16:56 CEST] <furq> yeah that shows a bunch of processes with nice 10
[04:17:06 CEST] <furq> so i guess that adds up
[04:17:30 CEST] <furq> fwiw if you're running a bunch of processes concurrently you probably want to set libx264 to use fewer threads
[04:17:38 CEST] <furq> or just run one at a time
[04:27:31 CEST] <kepstin> htop is showing multiple threads from the one ffmpeg run separately there
[04:28:14 CEST] <kepstin> no idea what's up with the nice stuff :/
[04:28:26 CEST] <ekiro> :(
[04:28:28 CEST] <kepstin> maybe x264 is internally doing some per-thread prioritization?
[04:29:19 CEST] <ekiro> i haven't changed any configs, everything is default as installed from the ubuntu package
[04:32:21 CEST] <ekiro> here is the htop output when i run 4 processes concurrently each set with 2 threads https://i.imgur.com/m3yEAyN.png
[04:35:01 CEST] <ekiro> this is a 4core cpu. i also tested this on a 24-core cpu and had similar results. lots of cores were underutilized
[04:46:11 CEST] <alex88> hello everyone, I'm trying to overlay a series of transparent png images over a video, with `ffmpeg -i video-background.mp4 -framerate 25 -pattern_type glob -i 'screenshots/*.png' -vcodec libx264 -pix_fmt yuv420p video.mov` however I get a black image with the images overlayed, any suggestion?
[04:50:03 CEST] <alex88> sorry the command is `ffmpeg -i video-background.mp4 -framerate 25 -pattern_type glob -i 'screenshots/*.png' -filter_complex overlay -vcodec libx264 -pix_fmt yuv420p video.mov`
[04:52:55 CEST] <alex88> oh nvm, it was because of the image size vs the video size
[11:32:00 CEST] <hans_> i want to seek to the nearest keyframe of second #100
[11:32:21 CEST] <hans_> -ss 100 may be a keyframe, but it may also not be, how can i get the nearest keyframe?
[11:43:37 CEST] <hans_> seems i can get keyframe info here with this: ffprobe -select_streams v -skip_frame nokey -show_frames -show_entries frame=pkt_pts_time,pict_type -print_format json
[11:43:56 CEST] <hans_> i can make due with that if nobody has a better idea
[11:47:42 CEST] <hans_> also seems like i just encountered a bug.. https://i.imgur.com/BA9FQc6.png
[11:49:21 CEST] <hans_> my ffmpeg installation is 5 months old tho, anyone with a recent version of ffmpeg want to try to replicate?
[11:52:42 CEST] <bencoh> hans_: I don't really know what -show_frames does, but assuming it just 'prints' frame in input stream order (without decoding) then your issue is just that pts != dts
[11:53:43 CEST] <bencoh> (encoded frames aren't ordered according to pts)
[11:55:41 CEST] <hans_> btw same thing from a today's git build
[11:56:03 CEST] <hans_> (just updated my ffmpeg)
[11:56:52 CEST] <bencoh> (oh and, as such, this is proper behavior and notabug(tm))
[11:56:59 CEST] <hans_> what is pts?
[11:57:19 CEST] <bencoh> presentation timestamp ... ie when a frame is supposed to be displayed
[11:57:25 CEST] <bencoh> (vs decoding timestamp)
[12:00:18 CEST] <bencoh> I'd try and display every frame (remove -skip_frame) just to make sure
[12:00:31 CEST] <hans_> so.. it's decoded in a different order than presented, eg in this example the frame that is supposed to be displayed 2 seconds into playback is the first frame to be decoded, and the frame that is supposed to be displayed 0 seconds into playback is the second frame to be decoded?
[12:00:42 CEST] <bencoh> yeah
[12:00:48 CEST] <hans_> weird
[12:01:00 CEST] <hans_> well.. unless it's a buffer thing maybe?
[12:01:07 CEST] <bencoh> there is a buffer, yeah
[12:01:31 CEST] <hans_> why would the frames for second #2 be encoded before the frames for second #1 ?
[12:02:14 CEST] <bencoh> although the goal isn't buffering, but effective compression. decoder buffers decoded frames and compute current frame based on current encoded (input) frame and previously decoded frames
[12:02:36 CEST] <bencoh> https://en.wikipedia.org/wiki/Group_of_pictures
[12:02:57 CEST] <hans_> i see
[12:03:06 CEST] <bencoh> what bothers me though is that you're listing keyframes only ...
[12:03:21 CEST] <hans_> yeah i'm looking for keyframes
[12:03:51 CEST] <bencoh> and reordering between keyframes sounds ... odd
[12:03:53 CEST] <hans_> because i'm going to cut up a movie into little pieces using -c:v copy, and now i'm doing it without caring about keyframes and the result is that several of the clips don't have video for the first 3 seconds or so
[12:07:46 CEST] <hans_> like i'm doing "ffmpeg -i foo.mkv -c:v copy -ss 100 -t 100 clip2.mkv"  then "ffmpeg -i foo.mkv -c:v copy -ss 200 -t 100 clip3.mkv" and so on, and if there wasn't a keyframe on second #100 then there's no video until the next keyframe, and if the next keyframe is on second #103 then there's no video for the first 3 seconds of that clip~
[12:17:54 CEST] <DaniloPereira> hello
[12:51:52 CEST] <JEEB> Krishnakumar: your mpeg-2 video is broken (most likely)
[12:52:14 CEST] <JEEB> concealing errors is something some of the decoders in FFmpeg do in case the input is not good
[12:52:27 CEST] <JEEB> sometimes it just means that you have started decoding before a random access point
[12:52:44 CEST] <JEEB> in which case the input is not broken, but you cannot decode it properly until you hit a random access point :P
[14:48:09 CEST] <thnee> Can anyone please help me understand what the problem is with RTMP, in this debug log? https://gist.github.com/thnee/35c4dc06652d239d11d509d7e8a880a4
[14:48:56 CEST] <thnee> The RTMP server is nginx with RTMP module, and it is not showing any errors
[14:50:09 CEST] <JEEB> raw H.264 over RTMP probably isn't going to fly
[14:50:11 CEST] <furq> thnee: rtmp expects -f flv
[14:50:15 CEST] <JEEB> yes
[14:50:22 CEST] <JEEB> FLV over RTMP
[14:50:35 CEST] <furq> also you'll need to set -c:v libx264 -c:a aac or else it'll default to whatever the defaults for flv are
[14:50:47 CEST] <furq> which are probably vp6 and some awful audio codec
[14:51:40 CEST] <thnee> Hmm really? In our iOS app we seem to be using h264 with the HaishinKit client.
[14:51:52 CEST] <thnee> But I tried with -f flv and it fails with the same cryptic error.
[14:52:56 CEST] <thnee> I tried adding those -c flags, and it still fails with the same cryptic error..
[14:53:10 CEST] <JEEB> make sure the URL etc are correct I guess :P
[14:53:17 CEST] <furq> yeah you're not setting an endpoint in the url
[14:53:21 CEST] <thnee> You can see in the log that there is no proble on the TCP layer, it connects fine
[14:53:28 CEST] <JEEB> if you are pushing H.264+AAC over FLV over RTMP - that part is a-OK
[14:53:32 CEST] <furq> [rtmp @ 0x56408ec16ec0] Proto = rtmp, path = , app = , fname =
[14:53:42 CEST] <furq> i'm pretty sure rtmp requires at least an app in the url
[15:24:28 CEST] <thnee> Ah ok, that explains what the application in the server config is for. Thank you kindly guys!
[18:09:13 CEST] <DaniloPereira> can someone help me with vaapi?
[18:14:08 CEST] <TikityTik> DaniloPereira: what's the problem?
[18:14:26 CEST] <TikityTik> best to just ask
[18:14:35 CEST] <TikityTik> on what your question or problem is
[18:16:48 CEST] <DaniloPereira> its impossible to use normal options in vp9_vaapi like "-deadline best -cpu-used 0" ?
[18:19:02 CEST] <TikityTik> DaniloPereira: try -quality best
[18:19:28 CEST] <TikityTik> also post your command line that you are running
[18:21:40 CEST] <DaniloPereira> ffmpeg -vaapi_device /dev/dri/renderD128 -i video_test.mp4 -vf 'format=nv12,hwupload' -c:v vp9_vaapi -deadline best -cpu-used 0 output2.webm
[18:21:53 CEST] <DaniloPereira> is this correct?
[18:25:35 CEST] <BtbN> I don't see any deadline or cpu-used option in vaapi_encode_vp9
[18:25:57 CEST] <BtbN> Those look like libvpx options
[18:26:20 CEST] <DaniloPereira> i know
[18:26:20 CEST] <DaniloPereira> :T but it is used in normal vp9 encoder
[18:28:02 CEST] <BtbN> They don't apply to vaapi
[18:28:15 CEST] <TikityTik> DaniloPereira: i only see 3 options for vp9_vaapi. ffmpeg --help encoder=vp9_vaapi
[18:28:22 CEST] <SixEcho> "crop=x=0:y=ih*0.05:out_h=ih-y" errors with "Error when evaluating the expression 'ih-y'", but docs say out_h can contain y as a computed value. tried h alias as well.
[18:30:49 CEST] <DaniloPereira> ;n; i know...
[18:30:50 CEST] <DaniloPereira> but what this options can be used for?
[18:32:36 CEST] <TikityTik> DaniloPereira: have you checked this yet? https://www.ffmpeg.org/ffmpeg-codecs.html#VAAPI-encoders
[18:33:42 CEST] <TikityTik> i assume -low_power just means using less voltage & performance
[18:34:15 CEST] <DaniloPereira> so low_power is a bad thing?
[18:34:15 CEST] <TikityTik> -loop_filter_level has similar values to crf, so i want to guess it's works like crf
[18:35:46 CEST] <TikityTik> DaniloPereira: it will be lower performance
[18:35:56 CEST] <TikityTik> it's generally used if you want to save battery on laptops i think
[18:36:05 CEST] <DaniloPereira> :T
[18:36:06 CEST] <TikityTik> DaniloPereira: https://wiki.archlinux.org/index.php/ATI#Dynamic_power_management
[18:36:08 CEST] <DaniloPereira> well
[18:37:11 CEST] <DaniloPereira> i encoded a 1.2 gb video in vp9 and got a 306 mb video
[18:37:11 CEST] <DaniloPereira> using vaapi in default setthings
[18:37:11 CEST] <DaniloPereira> you think this is good?
[18:38:19 CEST] <TikityTik> DaniloPereira: in vp9_vaapi?
[18:38:28 CEST] <DaniloPereira> yes
[18:39:03 CEST] <DaniloPereira> i can't use h264 or h265 to decrease the video size
[18:39:03 CEST] <DaniloPereira> some times the size increases
[18:39:07 CEST] <TikityTik> don't know, either the compression is extremely good or a lot of data was discard
[18:39:37 CEST] <TikityTik> i want to guess that vp9_vaapi does no compression
[18:40:24 CEST] <DaniloPereira> i gave a look in the video frame by frame
[18:40:24 CEST] <DaniloPereira> is possible to see a very little quality loss
[18:40:24 CEST] <DaniloPereira> but if its possible to change the crf it can be changed
[18:40:31 CEST] <DaniloPereira> to lossless
[18:40:59 CEST] <TikityTik> you could try -b:v 0 if you don't want quality loss
[18:41:16 CEST] <TikityTik> unlimited availabe bitrate
[18:41:43 CEST] <DaniloPereira> you think i could reach the same level of file size?
[18:42:02 CEST] <TikityTik> probably, it might even be larger
[18:43:47 CEST] <DaniloPereira> well
[18:43:47 CEST] <DaniloPereira> i think i can reduce the crf a little
[18:43:47 CEST] <DaniloPereira> and maintain the size level
[18:43:47 CEST] <DaniloPereira> the video didn't lost much quality in the end
[18:45:53 CEST] <TikityTik> i don't know if loop_filter_level is crf really, i just merely stated it shares the same value range as crf
[18:48:33 CEST] <DaniloPereira> what about hevc_vaapi?
[18:48:33 CEST] <DaniloPereira> what is this -profile and -level
[18:49:05 CEST] <another> https://en.wikipedia.org/wiki/HEVC#Profiles https://en.wikipedia.org/wiki/HEVC#Tiers_and_levels
[18:49:33 CEST] <TikityTik> can't even use vp9_vaapi lol
[18:49:35 CEST] <TikityTik> ffmpeg -vaapi_device /dev/dri/renderD128 -i FUKKIRETA.webm -c:v vp9_vaapi out.webm -y
[18:50:10 CEST] <TikityTik> "Impossible to convert between the formats supported by the filter 'Parsed_null_0' and the filter 'auto_scaler_0' Error reinitializing filters!"
[18:51:48 CEST] <DaniloPereira> yeah some times even if the coman is correct it says something is wrong
[18:55:21 CEST] <DaniloPereira> you need to run in  -vf 'format=nv12,hwupload'
[18:58:28 CEST] <SixEcho> bug?  "crop=x=0:y=ih*0.05:out_h=ih-y" errors with "Error when evaluating the expression 'ih-y'", but docs say out_h can contain y as a computed value. tried h alias as well.
[19:00:52 CEST] <SixEcho> ^ nvm& docs clarify later that x/y are calculated after out_h/w
[19:02:07 CEST] <bowbles> I am getting out of sync audio in trying to combine a bunch of mp4 videos. Tried a few things including re-encoding everything and still no juice. See https://pastebin.com/R8NnrJd0 for my last attempt. Resulting video is a bit over 2 hours long, and audio cuts out around 1 hour 45 mins.
[19:04:40 CEST] <bowbles> I am guessing it may have something to do with the tbc variance?
[19:07:20 CEST] <bowbles> Also, I first started down this path with ffmpeg -f concat -safe 0 -i mylist.txt -c copy output_ff.mp4, but that threw a lot of messages that said Non-monotonous DTS in output stream 0:0; previous: 17157593, current: 14156382; changing to 17157594. This may result in incorrect timestamps in the output file.
[20:50:27 CEST] <ekiro> how do i tell ffmpeg to use the fonts i dumped out of the video for its subtitles?  is there a option to pass the fontdir for it to look through?
[21:09:02 CEST] <leif> I'm capturing screen data with the command `ffmpeg -f v4l2 -i '/dev/video1' out.mp4`, and its coming out like this: https://www.youtube.com/watch?v=X8SBRoTpO0g&feature=youtu.be
[21:09:31 CEST] <leif> (All of the colors and screen positions are messed up.)
[21:09:49 CEST] <BtbN> Whatever software you use to make your desktop appear as video1 is most likely to blame.
[21:10:08 CEST] <leif> When I capture it with OBS it appears fine, so I doubt its v4l2 proper.
[21:10:32 CEST] <leif> BtbN: I'm not using software, I"m using a usb capture card.
[21:10:45 CEST] <BtbN> You're not passing the correct parameters to it then
[21:10:57 CEST] <BtbN> But why use a capture card to capture the PC itself?
[21:11:30 CEST] <TheAMM> To test it and find the proper configuration for other things
[21:11:54 CEST] <TheAMM> You've got a PC right there to test it on, so why not test it on that
[21:12:00 CEST] <BtbN> Capture Cards are finnicky, specially on Linux.
[21:12:20 CEST] <BtbN> OBS has a whole separate library just to deal with various hardware specific quirks
[21:13:39 CEST] <leif> TheAMM: Yup, I'm eventually goin to be use it to capture data from a video camera.
[21:13:56 CEST] <leif> BtbN: Ya...you're talking about the v4l2 quirks thing, right?
[21:14:07 CEST] <BtbN> Most of them are actually for Windows DirectShow stuff
[21:14:18 CEST] <leif> Anyway, the specific card I'm using is the ezcap 261 https://www.amazon.com/gp/product/B07JLY6RS3/ref=ppx_yo_dt_b_asin_title_o02_s00?ie=UTF8&psc=1
[21:14:33 CEST] <leif> Ah, hmm...okay.
[21:14:49 CEST] <BtbN> Weird off-brand card there
[21:14:51 CEST] <leif> Anyway, so OBS uses its own thing and doesn't even touch v4l2?
[21:14:57 CEST] <BtbN> wonder what's in there
[21:15:17 CEST] <BtbN> Of course OBS uses v4l2, it has to, if that's how the card is driven.
[21:15:24 CEST] <TheAMM> ezcaps are shit mate
[21:15:36 CEST] <TheAMM> Luckily I've finished my USB capture card quest
[21:15:40 CEST] <BtbN> But it probably has special defaults
[21:15:51 CEST] <leif> BtbN: ah, okay, that makes sense.
[21:16:03 CEST] <BtbN> so basically... try different settings until it works
[21:16:11 CEST] <TheAMM> I bought an ezcap, it was utter shit on 60fps https://mygi.ga/a4x/aedHI.png and not much better on 30 https://mygi.ga/LdD/aedHJ.png
[21:16:16 CEST] <leif> TheAMM: Ya, the quality isn't great. but honestly, I only need it to capture preseter information.
[21:16:20 CEST] <leif> (For conference talks)
[21:16:27 CEST] <leif> So cheap is better in this instance.
[21:16:30 CEST] <TheAMM> I bought a rebranded ezcap locally, identically shit
[21:16:31 CEST] <BtbN> You're much better of with an Elgato CamLink or something for that
[21:16:50 CEST] <leif> BtbN: LOL. I'm sitting nextg to a camlink as we speak.
[21:16:56 CEST] <TheAMM> I bought some unbranded different capture device from ebay, it was shit but not utter
[21:17:06 CEST] <TheAMM> Then I bought the camlink and that's the one
[21:17:18 CEST] <BtbN> Magewell stuff is also nice, but pretty expensive
[21:17:20 CEST] <leif> TheAMM: Makes sense.
[21:17:24 CEST] <TheAMM> It's a bit finicky because of course it is
[21:17:30 CEST] <TheAMM> why would anything *work*
[21:17:36 CEST] <leif> LOL...sigh.
[21:17:45 CEST] <TheAMM> but it's more than acceptable
[21:17:47 CEST] <BtbN> The nice thing about the Camlink is that it takes pretty much anything you throw at it
[21:17:58 CEST] <TheAMM> Really?
[21:18:06 CEST] <BtbN> Weird off-spec resolution/refresh rates, no problem
[21:18:12 CEST] <TheAMM> mmh
[21:18:21 CEST] <TheAMM> I've heard bit of the opposite, since it has no scaler
[21:18:31 CEST] <BtbN> Yeah, it doesn't
[21:18:33 CEST] <TheAMM> but I haven't found out all that much
[21:18:39 CEST] <BtbN> Why would it need one? It just captures
[21:18:46 CEST] <TheAMM> because nobody savvy wants to talk about capture devices online
[21:18:58 CEST] <leif> Yes, as far as I can tell, it works as widely as it does because its so....logically stupid...
[21:19:05 CEST] <leif> (for lack of a better phrase.)
[21:19:21 CEST] <TheAMM> I can't fucking comprehend why the chinese shits are so fucking analog
[21:19:31 CEST] <TheAMM> It just leaves me confused
[21:19:38 CEST] <BtbN> It's great for capturing weird digitizes 62.5Hz retro consoles, which most equipment will refuse to sync
[21:19:53 CEST] <leif> mmm...okay. So the problem I had with the camlinks are that v4l2 kept looking them.
[21:20:02 CEST] <BtbN> ?
[21:20:06 CEST] <TheAMM> locking?
[21:20:09 CEST] <leif> (As in they ould dissapear from /dev/video)
[21:20:12 CEST] <leif> Possibly.
[21:20:15 CEST] <leif> Any way I can check?
[21:20:21 CEST] <TheAMM> Haven't had that
[21:20:29 CEST] <BtbN> sounds like wonky USB connection
[21:20:41 CEST] <leif> I see. Are you using the new cam link 4ks or the older camlinks?
[21:20:41 CEST] <BtbN> You need a non-ass USB3 controller for them
[21:20:48 CEST] <TheAMM> 4K
[21:21:16 CEST] <TheAMM> I have to reset the card sometimes by replugging it, and had to buy a proper USB3 card since the mobo's wasn't good enough
[21:21:27 CEST] <leif> BtbN: Yes, I've tried it with several USB controllers. Including the intel NUC's usb controllers.
[21:21:40 CEST] <TheAMM> also protip, setpts=PTS-STARTPTS
[21:21:41 CEST] <BtbN> Intel ones are usually fine
[21:21:49 CEST] <TheAMM> for both audio and video if you're doing that
[21:21:51 CEST] <leif> TheAMM: Mmmmm...okay. We were using the older camlinks. Maybe they changed something that made it more stable on linux?
[21:21:57 CEST] <another> I hear good things about Blackmagic devices
[21:22:01 CEST] <TheAMM> :shrug:
[21:22:12 CEST] <BtbN> Blackmagic needs special and half-proprietary drivers
[21:22:13 CEST] <BtbN> not a fan
[21:22:20 CEST] <BtbN> The Camlink presents itself as a UVC webcam, so just works, on any OS
[21:22:33 CEST] <TheAMM> The Camlink 4K is in my limited experience and medium research the best price/quality device that works on linux
[21:22:36 CEST] <leif> Yes. Having something be UVC is soooo much nicer imo.
[21:22:47 CEST] <TheAMM> Paid 130¬ for it
[21:22:49 CEST] <leif> TheAMM: makes sense. thanks.
[21:22:56 CEST] <BtbN> Magewell also has some nice UVC Capture Cards.
[21:22:57 CEST] <TheAMM> I'd love a magewell but those are 400¬+
[21:22:59 CEST] <leif> Ya, its the same price as the old cam links.
[21:23:12 CEST] <leif> I just bought mine right at the tail end of its lifecycle. :(
[21:23:13 CEST] <TheAMM> Sadly I'm not a professional who could just write it off on the business
[21:23:30 CEST] <TheAMM> I just want me PS4 gameplay
[21:24:17 CEST] <TheAMM> https://mygi.ga/XDu/aedRe.webm
[21:24:31 CEST] <leif> Ya....the problem with using a $400 magewell device is that we need 4 of them (minimum), which would make it harder to convince sigplan to get them.
[21:25:12 CEST] <kepstin> looks like there are firmware updates available for some of the older camlink devices. no idea if they'd help with the specific issues you're seeing.
[21:25:23 CEST] <kepstin> looks like you need to use the mac or windows drivers to get it, tho :/
[21:25:51 CEST] <kepstin> or maybe i'm just looking in the wrong spot and there's actually nothing
[21:26:03 CEST] <leif> kepstin: sad that they require it, although I can swing it.
[21:26:30 CEST] <leif> (I do have access to macs and windows machines.)
[21:26:42 CEST] <kepstin> i suppose you can try just running the device on a windows machine once and see if it does anything.
[21:26:44 CEST] <leif> I'll look into it more, thanks y'all for the help. :D
[21:27:47 CEST] <leif> kepstin: oh, the camlinks work beutifuly (for my purposes). They just have been finicky on linux (for me anyway).
[21:41:31 CEST] <BtbN> Magewell does have 4-Way-HDMI cards iirc
[21:41:59 CEST] <BtbN> Which is still expensive, but not as crazy expensive as 4 Cam-Links plus 4 USB3-Extension-Cards you'll need to use them in parallel
[21:49:59 CEST] <SixEcho> question: i'm using ffmpeg to do preprocessing before i load into another utility which only opens mp4 files& i'm curious if there are any better/faster non-x264 mp4 codecs that will also preserve high quality.  i value quality>cpu>disk.   for reference i'm using a hardware x264 encoder and getting ~1.5x encoding speed with 25Mbitrate& but the hardware encoder does not have any quality options yet (early MacOS support?)
[21:50:44 CEST] <BtbN> That sounds like x264 will outperform it in every aspect, including speed
[21:53:33 CEST] <furq> SixEcho: you mean a hardware h264 encoder, x264 is one specific software encoder
[21:54:00 CEST] <furq> and also yeah x264 is as good as you'll get
[21:56:24 CEST] <SixEcho> furq: yes h264.  was hoping maybe there was a fast codec that didn't compress well. (lossless would be going to far though if there is such a thing).
[21:56:38 CEST] <kepstin> x264 can do lossless even, so..
[21:56:53 CEST] <kepstin> x264 can also be configured to be faster with less compression
[22:01:05 CEST] <DHE> x264 in quality selection mode + a faster preset may be what you want
[22:01:24 CEST] <DHE> -preset:v veryfast -crf 10   # should look decently good. lower the crf value for even better quality
[22:02:01 CEST] <DHE> does -crf 0 provide lossless, or do you still need -qp 0 instead?
[22:02:48 CEST] <kepstin> -crf 0 provides lossless, but only when using 8bit video
[22:02:53 CEST] <kepstin> -qp 0 works in all modes
[22:03:33 CEST] <BtbN> You usually DO want lossless, intra-only even, for intermediate files
[22:03:54 CEST] <kepstin> assuming you have the disk io speed and capacity to handle it, yeah
[22:04:15 CEST] <kepstin> and that your decoder supports it (it's a different profile from non-lossless)
[22:06:57 CEST] <kepstin> if you don't *need* intra only (which is useful for making arbitrary video cuts without reencoding), i'd normally suggest a small gop (5-10 frames maybe?), since it still helps seek speed a lot while also making the video smaller than intra.
[22:07:05 CEST] <kepstin> less difference in lossless of course
[22:08:56 CEST] <savolla> guys I'm trying to bind a key to screencast with ffmpeg and starting screencast and video quality is great. But how can I quit/finish screen cast via command? I always do pkill -f ffmpeg but keybinding does not accept pkills
[22:10:47 CEST] <kepstin> sounds like an issue with your keybinding program, no reason it shouldn't be able to run pkill
[22:11:30 CEST] <SixEcho> savolla: what are you using to bind keys?
[22:11:41 CEST] <savolla> let's say I don't want to killing the process. I want to exit this program normally
[22:11:57 CEST] <savolla> I'm using i3wm
[22:12:16 CEST] <savolla> this is not the case. I just want to know how to exit screencast via command
[22:13:02 CEST] <SixEcho> savolla: tried  pkill -SIGQUIT ffmpeg ?   or SIGINT
[22:13:21 CEST] <savolla> guys..
[22:14:10 CEST] <savolla> killing a process for quiting is like destroting a house instead of just leaving it
[22:14:43 CEST] <savolla> isn't there a normal quit command for this software?
[22:15:43 CEST] <SixEcho> savolla: sending different signals does not necessarily kill the software& it depends if ffmpeg listens to those or how it reacts to them.  it's like sending it a ctrl+c in some cases.
[22:16:40 CEST] <savolla> oh okay that makes sense now
[22:17:19 CEST] <kepstin> ffmpeg exits nicely (finalizing files, etc) upon receiving SIGINT.
[22:17:32 CEST] <kepstin> which is what happens when you type ctrl-c in a terminal in linux
[22:17:51 CEST] <kepstin> if you send it a *second* sigint, it will exit immediately
[22:18:37 CEST] <bowbles> Any advice on fixing a "Non-monotonous DTS in output stream" error? I tried https://pastebin.com/R8NnrJd0 and it still produces audio that becomes progressively more offset the longer it goes on.
[00:00:00 CEST] --- Wed May  8 2019