[Ffmpeg-devel-irc] ffmpeg.log.20190201

[burek]

[00:23:31 CET] <mfwitten> The `minterpolate' filter is producing nice results, but at the very beginning of the results, there is some jerky motion as though no interpolation were done, or as though frames are duplicated. Is this an expected artifact of the algorithm?
[00:30:03 CET] <mfwitten> Specifically, I'm interpolating to `fps=180000/1001' and then making the result slow motion via `setpts=6*PTS'.
[00:30:47 CET] <mfwitten> At the very beginning of the resulting stream, there is non-smooth/stuck motion
[00:36:47 CET] <mfwitten> When I drop the `setpts' part, it naturally produces a result with `179.82 fps'. Interestingly, the encoding process shows `frame= 4253 ... dup=3544'. That seems suspicious.
[01:05:44 CET] <raytiley> can I make this filter run at realtime? ffmpeg -f lavfi -i smptehdbars=1920x1080 -f lavfi -i "sine=frequency=1000"
[01:18:52 CET] <mfwitten> The only way I can get rid of the jerky motion at the beginning of the stream is to save the stream, and then `-ss 1' to seek past it and just copy out the rest of the stream. That seems so absurd
[01:42:36 CET] <mfwitten> I've used `trim=start_frame=...' to remove an initial 6x multiple of the ~179.82 FPS stream, and then used `setpts=6*(PTS-STARTPTS)'
[01:42:48 CET] <mfwitten> This at least cuts out the bad part, but I do wonder why there is a bad part at all.
[17:24:45 CET] <yrios> I always get "Native MSYS builds are discouraged, please use the MINGW environment." ... Why does the configure script even think I use MSYS?
[17:44:01 CET] <BtbN> Because the target os matches msys*
[17:44:09 CET] <BtbN> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=configure;h=e1412352fa450b9d0df81d5981dd7fcfb0b30a2d;hb=HEAD#l5205
[18:08:52 CET] <yrios> I had multiple 'uname' installed, so I had to fix the order in which they appeared in %PATH%. seems to work now
[18:13:20 CET] <BtbN> how even
[18:42:22 CET] <yrios> not my machines ..
[18:42:42 CET] <furq> i was baffled by that at first but it's windows so there's probably 17 broken msys installs on there still in the path
[18:43:16 CET] <yrios> yep
[18:50:32 CET] <yrios> the generated Makefile only contains 'include ./Makefile' ... What could be the reason?
[19:05:18 CET] <mfolivas> not sure how I can get my videos to show on quicktime using vaapi
[19:17:22 CET] <mfolivas> using this: ffmpeg -vaapi_device /dev/dri/renderD128 -i some-random-video.mp4 -vf 'hwupload,scale_vaapi=w=1920:h=1080:format=nv12' -c:v h264_vaapi -level 4.1 -preset fast -b:v 5M -profile 578 -bf 0 small-video.mp4
[19:17:37 CET] <mfolivas> I figured that I should use a profile but can't find the right one
[20:05:31 CET] <mfolivas> one of our investors wants me to try to see if we can use Intel's nueral compute stick 2
[20:05:44 CET] <mfolivas> does anyone knows if this can be used for decoding?
[20:05:49 CET] <mfolivas> https://venturebeat.com/2018/11/14/intels-neural-compute-stick-2-is-8-times-faster-than-its-predecessor/
[20:12:01 CET] <durandal_1707> nope
[20:12:33 CET] <pink_mist> do you mean that nobody knows, or do you mean it can't be used :P
[20:14:01 CET] <durandal_1707> if someone build neurons specifically for decoding some codecs that it can be used
[20:32:55 CET] <mfolivas> durandal_1707 The hardware accelerators include things like hardware encode and decode of H.264 and Motion JPEG, as well as a warp engine for handling a fish-eye lens, dense optical flow, and stereo depth perception
[20:33:16 CET] <mfolivas> so, it does has encode and decode
[20:33:46 CET] <durandal_1707> not all codecs ...
[20:35:55 CET] <furq> apparently it's not quicksync so it won't be supported by ffmpeg
[20:36:22 CET] <furq> "won't be" meaning isn't now, someone might add it
[20:37:11 CET] <JEEB> if it's not vaapi or QS, then lul
[20:38:11 CET] <furq> https://www.movidius.com/myriadx
[20:38:13 CET] <furq> it's this thing
[22:11:39 CET] <saml> if key frames are at irregular interval, I need to encode it again to make key frames at regular interval, right?
[22:12:36 CET] <JEEB> only if you actually need that
[22:12:44 CET] <JEEB> and often when you think you need that, it's not actually true
[22:15:13 CET] <saml> cdn vendor, akamai, cannot support HLS packaging of videos with irregular key frame interval
[22:16:07 CET] <saml> JEEB, how would you generate HLS (m3u8) properly given a video with irregular key frame interval?
[22:16:14 CET] <saml> i should try first
[22:16:33 CET] <JEEB> the usual way? I mean, I've had mismatching GOPs between profiles in HLS for ages :P
[22:17:00 CET] <JEEB> there might be some minor devices that don't support it, maybe, but android's exoplayer and iOS's player support that stuff for ages
[22:17:14 CET] <saml> https://ffmpeg.org/ffmpeg-formats.html#hls-2   It says Make sure to require a closed GOP when encoding and to set the GOP size to fit your segment time constraint.
[22:18:03 CET] <mfolivas> need help with VAAPI, can't get it to play with QuickTime
[22:18:17 CET] <mfolivas> I'm sure if has to be some settings that I can use
[22:18:24 CET] <mfolivas> I tried using -profile
[22:20:26 CET] <JEEB> saml: that just means that you probably want closed GOPs (not open), and that the GOP size is at least having a max length set
[22:20:30 CET] <JEEB> to what you need
[22:40:18 CET] <GuiToris> hey, can you help me encode a video with vp9 codec?
[22:40:40 CET] <GuiToris> it must have a different scale than x265
[22:41:22 CET] <GuiToris> I've tried crf 28 but the output file was way too big
[22:41:33 CET] <GuiToris> 1:50 ~ 250Mb
[23:01:59 CET] <mfolivas> ffmpeg -hide_banner -vaapi_device /dev/dri/renderD128 -i input.mp4 -vf 'format=yuv420p' -c:v h264_vaapi output.mp4
[23:02:15 CET] <mfolivas> gives me this error: "Impossible to convert between the formats supported by the filter 'Parsed_format_0' and the filter 'auto-inserted scaler 0'"
[00:00:00 CET] --- Sat Feb  2 2019