[Ffmpeg-devel-irc] ffmpeg.log.20190404

Fri Apr 5 03:05:01 EEST 2019

[05:51:00 CEST] <cardz> Are there any tools out there to compare two videos, or to rate and score a video's quality on any number of factors
[05:51:25 CEST] <cardz> The goal being to conclude that one video is "better" than another, and how much better.
[05:55:24 CEST] <furq> cardz: https://ffmpeg.org/ffmpeg-filters.html#libvmaf
[06:00:46 CEST] <cardz> furq: thanks.  Do you know if this netflix lib is also ported to Windows?
[13:41:55 CEST] <lemourin> Is it expected behavior that the result of conversion from rgb24 to yuv420p is platform dependent?
[13:49:03 CEST] <JEEB> yes, there's some SIMD in there most likely
[13:49:20 CEST] <JEEB> and SSE2 etc might give different results to plain C
[13:54:01 CEST] <lemourin> JEEB: is there a way to make it platform independent? by e.g. disabling optimizations?
[13:59:29 CEST] <JEEB> the tests utilize a bitexact flag
[14:00:51 CEST] <JEEB> -sws_flags bitexact seems to be one
[14:01:18 CEST] <lemourin> thanks, i'll try this out
[14:29:30 CEST] <thebombzen> did kmsgrab ever get fixed on nVidia?
[14:44:04 CEST] <ahmadjamalmughal> Hi. Is anyone here applying for GSoC?
[14:50:13 CEST] <JEEB> ahmadjamalmughal: some people have popped up on the -devel channel, but I'm not sure how many mentors are on IRC
[16:49:43 CEST] <zPhantomGG> hi
[16:51:33 CEST] <zPhantomGG> Can someone answer me if this is a bug? http://phantsc.rf.gd/AAA/Bbb.html exactly in the second 7
[16:53:03 CEST] <zPhantomGG> using: 'ffmpeg.exe -i in.mp4 -vcodec copy -acodec aac -hls_list_size 0 -hls_segment_size 17000000 -f hls out.m3u8'
[16:55:56 CEST] <zPhantomGG> someone?
[16:58:10 CEST] <pink_mist> wait for people with knowledge in the area to respond
[16:58:17 CEST] <pink_mist> (that does not include me)
[17:00:24 CEST] <faLUCE> zPhantomGG: why do you see it as a bug?
[17:00:37 CEST] <faLUCE> is your mp4 different?
[17:05:03 CEST] <zPhantomGG> a video crash occurs.
[17:05:26 CEST] <zPhantomGG> when going from segment0.ts to segment1.ts
[17:06:15 CEST] <zPhantomGG> in the link I sent you can see clearly
[17:07:25 CEST] <zPhantomGG> I'm trying to get m3u8 with bytesranges, and each segment with a maximum size
[17:07:33 CEST] <zPhantomGG> not a size exactly, but approximate
[17:07:55 CEST] <zPhantomGG> so I use: -hls_segment_size
[17:09:32 CEST] <w1kl4s> it does?
[17:10:26 CEST] <faLUCE> zPhantomGG: with which criterium did you set  hls_segment_size to 17000000 ?
[17:11:15 CEST] <zPhantomGG> for each segment to have an approximate size
[17:11:44 CEST] <zPhantomGG> ~17MB
[17:11:45 CEST] <faLUCE> zPhantomGG: ok, but why did you choose the value 17000000 ?
[17:11:57 CEST] <faLUCE> did you try with another value?
[17:11:59 CEST] <zPhantomGG> https://github.com/FFmpeg/FFmpeg/blob/9461e4bc694b8d95d4224226b9781f8166d969b1/libavformat/hlsenc.c#L2895
[17:12:13 CEST] <zPhantomGG> yes
[17:12:36 CEST] <faLUCE> also with very different values?
[17:12:36 CEST] <zPhantomGG> this video in the link that I sent, was with a lower value
[17:12:53 CEST] <faLUCE> for example: 1 Mb
[17:13:00 CEST] <faLUCE> or even less
[17:13:40 CEST] <zPhantomGG> even different values, this always happens in the video, when it goes from one segment to the other
[17:15:55 CEST] <faLUCE> zPhantomGG: I would suggest you to re-encode with the same codec, but not with a copy
[17:16:13 CEST] <faLUCE> and then mux again with the same segment val and see what happens
[17:22:04 CEST] <zPhantomGG> Like this ? -vcodec libx264 -acodec aac -hls_list_size 0 -hls_segment_size 1000000 -f hls
[17:23:04 CEST] <faLUCE> yes
[17:23:19 CEST] <zPhantomGG> the same thing happens
[17:24:18 CEST] <faLUCE> zPhantomGG: try to mute the audio codec
[17:24:26 CEST] <faLUCE> -acodec none
[17:24:31 CEST] <zPhantomGG> ok
[17:24:59 CEST] <zPhantomGG> Unknown encoder 'none'
[17:25:10 CEST] <faLUCE> -an or something like that
[17:28:06 CEST] <zPhantomGG> -vcodec libx264 -an -hls_list_size 0 -hls_segment_size 1000000 -f hls
[17:28:25 CEST] <zPhantomGG> The same thing happens, the video hangs
[17:29:15 CEST] <faLUCE> and you don't hear the audio, right?
[17:29:49 CEST] <zPhantomGG> no audio, I can see the video hang
[17:29:55 CEST] <faLUCE> ok, wait
[17:34:09 CEST] <faLUCE> zPhantomGG: [hls @ 0x5602e0a4b380] Cannot use rename on non file protocol, this may lead to races and temporary partial files
[17:35:54 CEST] <zPhantomGG> I did not understand..
[17:38:11 CEST] <DHE> the traditional method of writing your new files to a .tmp file and then rename() it to the proper name for an atomic content update doesn't work if you're doing, say, direct HTTP PUT uploads
[17:41:25 CEST] <zPhantomGG> I am not uploading via HTTP PUT ...
[17:48:19 CEST] <faLUCE> zPhantomGG: https://docs.peer5.com/guides/production-ready-hls-vod/
[18:00:26 CEST] <zPhantomGG> I already saw this tutorial
[18:00:29 CEST] <zPhantomGG> I'm looking to make an m3u8 file that has byte range
[18:00:56 CEST] <zPhantomGG> without using '-hls_flags single_file'
[18:01:43 CEST] <zPhantomGG> but I also want to set an approximate size for each segment
[18:06:17 CEST] <zPhantomGG> it's probably some problem with '-hls_segment_size'
[18:06:18 CEST] <zPhantomGG> alguém tem uma ideia alternativa?
[18:06:24 CEST] <zPhantomGG> Does anyone have an alternative idea?
[18:11:18 CEST] <kevinnn> does anyone know how to properly handle a 60fps h264 network stream that has sliced NAL units? Previously I would just render after each NAL unit was received with libav but not that they are sliced there are actually much more than 60 NAL units coming in per second on a 60fps stream
[18:48:29 CEST] <poisond> hi :) I'm having problems with nvenc since a recent update: https://pastebin.ubuntu.com/p/S4mMKgKQT3/ ... is this a known problem? Any workaround?
[18:50:39 CEST] <JEEB> that really says nothing :D
[18:51:36 CEST] <JEEB> initialized successfully and then unloaded
[18:55:02 CEST] <poisond> JEEB: sorry, here's with an actual file https://pastebin.ubuntu.com/p/dnZSmYp2px/
[18:55:28 CEST] <JEEB> [hevc_nvenc @ 0x55f55b535500] OpenEncodeSessionEx failed: out of memory (10)
[18:55:36 CEST] <JEEB> so you're trying to do 5 encodes
[18:55:42 CEST] <JEEB> due to teh cover art in there :P
[18:56:16 CEST] <JEEB> I think nvidia tells you to eff off if it's like over 2 or 4
[18:56:20 CEST] <JEEB> in the driver
[18:56:28 CEST] <JEEB> unless you patch the driver to remove that limit
[18:57:36 CEST] <netkam2> can I record a command line only with ffmpeg
[19:00:57 CEST] <poisond> JEEB: ok, never encountered this before O_o so nvenc doesn't like multiple video stream?
[19:01:08 CEST] <poisond> JEEB: or is it the format it's complaining about?
[19:02:34 CEST] <poisond> I can transcode with -map 0:v:0
[19:02:50 CEST] <JEEB> poisond: it's the amount of encodes being done at a time :P
[19:03:02 CEST] <JEEB> the driver has a hard-coded limit for normal customer cards
[19:04:12 CEST] <poisond> JEEB: OK, thanks, that solved my problem! (still confuses me since I tend to do parallel transcodes using nvenc)
[19:04:57 CEST] <JEEB> you can do multiple, but in this case you were doing 5
[19:05:16 CEST] <JEEB> which I think is more than nvidia lets you do at once
[19:06:23 CEST] <JEEB> https://developer.nvidia.com/video-encode-decode-gpu-support-matrix
[19:06:30 CEST] <JEEB> poisond: ^ see this
[19:07:01 CEST] <JEEB> so for GTX 1080 it's one chip and 2 instances per chip
[19:07:20 CEST] <JEEB> and then max concurrent sessions is 2
[19:07:31 CEST] <JEEB> that's of course a 100% driver limitation :P
[19:07:50 CEST] <JEEB> nvidia wants you to pay more moneys for a quadro if you want more encodes at once
[19:08:23 CEST] <poisond> oh, only need to shelf out a few thousand bucks to do 3 parallel, that's quite generous of them to allow 2 on consumer cards then xD
[19:16:31 CEST] <nicolas17> can I make ffmpeg only decode intra frames? I'm speeding up a video like 60x and decoding the input is the bottleneck, it seems to be decoding every frame and then dropping most of them
[19:18:12 CEST] <JEEB> nicolas17: with the API you could do that, not sure if you can drop with the command line app
[19:22:42 CEST] <nicolas17> ok I think I can download the input file to my 4-core PC faster than I can transcode on my 1-shared-core VPS :/
[19:27:55 CEST] <kevinnn> I am getting this warning while using ffmpeg
[19:27:56 CEST] <kevinnn> chroma_format_idc 10 is not implemented
[19:28:11 CEST] <kevinnn> is it necessary to update ffmpeg in this case?
[19:29:22 CEST] <kevinnn> google literally returns this for a search on the internet
[19:29:30 CEST] <kevinnn> "No results found for "chroma_format_idc 10 is not implemented"."
[19:31:26 CEST] <kevinnn> which means I am the first human to have this issue?
[19:31:40 CEST] <nicolas17> what version of ffmpeg?
[19:31:46 CEST] <JEEB> is that HEVC or AVC?
[19:32:23 CEST] <kevinnn> JEEB: AVC
[19:32:38 CEST] <kevinnn> nicolas17: whatever the latest version debian stretch offers
[19:32:49 CEST] <kevinnn> I am using the libraries not ffmpeg directly
[19:33:12 CEST] <JEEB> should quickly check if it's logging it as bits
[19:33:16 CEST] <JEEB> or as a value
[19:33:20 CEST] <JEEB> like, is 10 3
[19:33:22 CEST] <JEEB> or is it 10
[19:33:30 CEST] <JEEB> because 3 makes sense
[19:33:33 CEST] <JEEB> 10 less so
[19:34:00 CEST] <kevinnn> it shows 10
[19:34:08 CEST] <kevinnn> I am not even sure what that means to be honest
[19:34:48 CEST] <JEEB> ok, so it's really 10
[19:34:59 CEST] <JEEB> just looked at the place where it logs where it requests a sample
[19:35:09 CEST] <JEEB> if it's more than 3 it logs that out
[19:35:21 CEST] <JEEB> let's see if that value makes any sense according to the specification :)
[19:36:28 CEST] <kevinnn> JEEB: hmm, okay but what exactly does it mean?
[19:36:39 CEST] <JEEB> "chroma_format_idc shall be in the range of 0 to 3, inclusive."
[19:36:43 CEST] <JEEB> according to the specification
[19:36:53 CEST] <JEEB> "chroma_format_idc specifies the chroma sampling relative to the luma sampling as specified in clause 6.2."
[19:36:54 CEST] <nicolas17> so your input file is corrupted or otherwise not conforming
[19:37:08 CEST] <nicolas17> i guess
[19:37:26 CEST] <kevinnn> oh okay, weird, why is my input file like that? It plays just fine on other decoders
[19:37:38 CEST] <kevinnn> well specifically openmax IL
[19:37:50 CEST] <JEEB> hw decoders anyways expect chroma_format_idc 1
[19:37:58 CEST] <JEEB> (or no value which is same as 1)
[19:38:10 CEST] <kevinnn> this is a software decoder
[19:38:11 CEST] <JEEB> which is 4:2:0 video
[19:38:22 CEST] <JEEB> kevinnn: anyways I guess quickest way to check is to get a sample
[19:38:35 CEST] <JEEB> but the initial guesstimate is that the stream is in some way corrupted
[19:38:40 CEST] <kevinnn> okay let me see...
[19:39:37 CEST] <nicolas17> ok no, looks like my slow VPS will finish encoding before my slow home internet finishes downloading input -.-
[19:48:42 CEST] <kevinnn> JEEB: https://www.mediafire.com/file/kthp42ahfzouy0h/output/file
[19:48:49 CEST] <kevinnn> this is that first packet
[19:48:57 CEST] <kevinnn> does it look corrupt?
[19:49:20 CEST] <JEEB> annex b?
[19:49:26 CEST] <kevinnn> yes
[20:02:15 CEST] <kevinnn> JEEB: is it super corrupt?
[20:03:35 CEST] <JEEB> kevinnn: I'm trying to get some food etc
[20:04:01 CEST] <kevinnn> JEEB: okay sorry, didn't mean to bother you
[20:41:00 CEST] <kevinnn> can anyone else take a look at my first frame?
[23:30:32 CEST] <TheSashmo> anyone know if there is a way to apply biss code to a file after its been recorded?  I have the ecnrypted file and the biss key
[00:00:00 CEST] --- Fri Apr  5 2019