[Ffmpeg-devel-irc] ffmpeg.log.20180206

Wed Feb 7 03:05:01 EET 2018

[00:00:36 CET] <pzich>      
[00:00:38 CET] <pzich> ~
[00:01:00 CET] Last message repeated 2 time(s).
[01:06:41 CET] <intrac> I'm trying to check for corruption or continuity problems in a set of .ts files. if I play the files with ffplay, at various points I get errors like:
[01:06:43 CET] <intrac> [h264 @ 0x7f6e781f9a00] error while decoding MB 0 54, bytestream 45085
[01:06:47 CET] <intrac> [h264 @ 0x7f6e781f9a00] concealing 1729 DC, 1729 AC, 1729 MV errors in I frame
[01:07:22 CET] <intrac> is there a way to get ffmpeg to scan a file and only report back if it encounters these errors?
[01:07:35 CET] <intrac> I've tried this suggestion without success: https://superuser.com/questions/100288/how-can-i-check-the-integrity-of-a-video-file-avi-mpeg-mp4
[01:07:38 CET] <furq> maybe ffmpeg -xerror
[01:07:43 CET] <furq> i forget if that bails on warnings though
[01:08:33 CET] <furq> if you just want it to log the warnings then you'll need -v warn, not -v error
[01:09:48 CET] <intrac> furq: I've tried 'error' and 'warning', but they don't show the same errors that ffplay does.
[01:10:36 CET] <intrac> the 'concealing X errors in I frame'
[01:10:55 CET] <intrac> also, it'd be helpful if the timecode was output as well
[01:12:18 CET] <intrac> it doesn't look like it's actually scanning through the whole file as it completes within a second or two
[01:12:35 CET] <intrac> and it's a 1 hour HD recording
[01:14:12 CET] <intrac> fairly sure that superuser posting is incorrect. it's not reading the frames, just the header/meta/basic file formats
[01:14:39 CET] <furq> that command will decode the streams
[01:15:04 CET] <furq> maybe add -enable_er 0 before -i
[01:15:43 CET] <intrac> furq: but at what level? it might extract the elementary stream, but I can't see that it's actually decoding every frame from the h264 stream in a couple of seconds
[01:16:01 CET] <intrac> and it might need to do a full decode to spot these errors
[01:38:03 CET] <intrac> hm, maybe ffprobe can extract the packet data
[02:58:34 CET] <FishPencil> Does FFmpeg support x264 and x265's multi bitdepth?
[03:28:10 CET] <kepstin> FishPencil: not yet (you need to use a pre-multi-bit-depth x264 or it'll fail to compile)
[03:28:28 CET] <kepstin> at least in the 3.4 release and last time i checked git master
[03:43:34 CET] <kepstin> FishPencil: x265 multi bit depth already works with ffmpeg
[03:45:54 CET] <DHE> x265 has had it for years now. x264 is, what a month or two?
[03:46:41 CET] <kepstin> committed in december 2017, yeah
[03:46:54 CET] <kepstin> (iirc)
[03:47:24 CET] <DHE> yep, committed december 24th
[03:47:39 CET] <DHE> oddly, authored january 6th 2017... that sat for a while
[03:47:54 CET] <kepstin> it's been a wip on their mailing list for a long time
[03:49:20 CET] <FishPencil> How is it done?
[03:50:50 CET] <kepstin> I haven't looked closely, but I expect it basically works by compiling the library multiple times (with symbol prefixes) and then linking them together with some wrapper code that calls the correct functions based on runtime config
[03:51:02 CET] <DHE> that's what the patch implies
[03:51:16 CET] <FishPencil> oh sorry, how do I use it in FFmpeg
[03:51:21 CET] <FishPencil> is it done with -pix_fmt?
[03:51:41 CET] <DHE> yeah, give it a 10bit format and it'll use the 10bit x264 code
[03:51:44 CET] <kepstin> FishPencil: the x265 wrapper in ffmpeg just selects based on the pix_fmt used, yeah
[03:52:16 CET] <kepstin> FishPencil: you can see this with "ffmpeg -h encoder=libx265", the list of pixel formats at the top of the output shows what the library can do
[03:52:32 CET] <kepstin> I expect the x264 wrapper will work similarly when it's added
[03:53:01 CET] <DHE> it's supported now in the Git version. support was added pretty quick. like 2 days later
[03:53:29 CET] <kepstin> oh, it is supported in ffmpeg? I took a look through the commits but didn't see anything obvious
[03:53:35 CET] <kepstin> maybe I just didn't look back far enough
[03:54:44 CET] <kepstin> https://git.videolan.org/?p=ffmpeg.git;a=commit;h=c6558e8840fbb2386bf8742e4d68dd6e067d262e well there it is
[04:00:34 CET] <FishPencil> What is a common 10-bit pix_fmt FFmpeg supports?
[04:03:43 CET] <kepstin> FishPencil: you'll almost always be using yuv420p10le
[05:21:14 CET] <prelude2004c> hey guys.. can i get some help.. ${ffmpeg} -hwaccel cuvid -c:v mpeg2_cuvid -deint 2 -resize 1280x720 -hwaccel_device $gpuenc -max_alloc 100000000 -threads 1 -vsync 0 -fflags +discardcorrupt -i "$stream" .. it wont deinterlace for some reason
[05:21:18 CET] <prelude2004c> i still get only 29.97fps
[05:21:30 CET] <prelude2004c> when i do it the other way with -vf yadif it works but then uses up all my CPU
[05:21:36 CET] <prelude2004c> i'm trying to deinterlace on decoder
[05:28:20 CET] <prelude2004c> tried deint bob which is same but no lucky.. still at 29.97fps
[05:48:36 CET] <prelude2004c> anyone around?
[06:30:40 CET] <Diag> prelude2004c: yadif is frame interpolation eh?
[06:36:09 CET] <prelude2004c> i was trying yadif but that is usually on the encoder side right?
[06:36:15 CET] <prelude2004c> i'm trying to deinterlace on the decoder side
[06:36:34 CET] <Diag> uhhh
[06:36:41 CET] <prelude2004c> reading things i'm told like
[06:36:53 CET] <prelude2004c> -c:v mpeg2_cuvid -deint bob
[06:37:01 CET] <Diag> the video should be deinterlaced when its decoded really
[06:37:11 CET] <Diag> im not familiar with how ffmpeg does it
[06:37:11 CET] <prelude2004c> and output shows : Stream #0:0 -> #0:0 (mpeg2video (mpeg2_cuvid) -> hevc (hevc_nvenc))
[06:37:20 CET] <Diag> >nvenc
[06:37:22 CET] <Diag> theres your problem
[06:37:31 CET] <Diag> dont ever expect anything nvidia makes to work
[06:37:35 CET] <Diag> however
[06:37:40 CET] <prelude2004c> well transcoding normally works
[06:37:41 CET] <Diag> How fast is the video being encoded?
[06:37:46 CET] <Diag> in terms of fps
[06:37:47 CET] <prelude2004c> realtime
[06:37:55 CET] <Diag> so its locked to 29.97?
[06:37:57 CET] <prelude2004c> comes in at 29.95 interlaced
[06:38:05 CET] <prelude2004c> yes sorry 97
[06:38:08 CET] <Diag> 29.95 interlaced? grody
[06:38:11 CET] <Diag> oh
[06:38:12 CET] <Diag> lol
[06:38:13 CET] <prelude2004c> and i want to deinterlace to 59.94
[06:38:27 CET] <Diag> you sure you dont got that backwards?
[06:38:40 CET] <Diag> or do you want the interpolation
[06:38:41 CET] <prelude2004c> i tried to use a -vf yadif but cpu was through the roof
[06:38:52 CET] <Diag> yeah its probably pretty intensive
[06:38:55 CET] <Diag> what res video are you using
[06:38:58 CET] <prelude2004c> i dont care how it works :) i just want to convert to progressive
[06:39:14 CET] <prelude2004c> its coming in at 1080i , i am converting to 720p
[06:39:25 CET] <Diag> whats it coming from?
[06:39:31 CET] <prelude2004c> fiber link
[06:39:48 CET] <Diag> what sort of cpu is this running o
[06:39:49 CET] <Diag> on*
[06:39:57 CET] <prelude2004c> xeon's
[06:40:05 CET] <prelude2004c> but.. i run like 15 channels on the system
[06:40:06 CET] <Diag> and what do you consider "through the roof"
[06:40:07 CET] <Diag> for usage
[06:40:15 CET] <prelude2004c> so.. i have to use decoder / encoder on the GPU
[06:40:26 CET] <Diag> gpu cant yadif
[06:40:38 CET] <prelude2004c> ya but its supposed to -deint 2
[06:40:46 CET] <prelude2004c> on cuvid isn't it?
[06:41:21 CET] <Diag> yeah
[06:41:21 CET] <prelude2004c> the idea is.. source is mpeg2video .. i use cuvid to decode and deinterlace, then pass it to nvenc to transcode back to 720p nvenc
[06:41:26 CET] <prelude2004c> hevc
[06:41:51 CET] <prelude2004c> even on the decoder i can -resize so.. i dont' even have to do anything there.
[06:42:01 CET] <Diag> What do you consider through the roof though
[06:42:03 CET] <Diag> for cpu usage
[06:42:09 CET] <prelude2004c> like 200% per channel
[06:42:15 CET] <Diag> 200% being...?
[06:42:16 CET] <Diag> 2 cores?
[06:42:20 CET] <prelude2004c> ya pretty much
[06:42:25 CET] <Diag> what xeons?
[06:42:37 CET] <prelude2004c> considering with GPU work only 30% right now
[06:42:47 CET] <prelude2004c> 5660 i think or something like that
[06:43:02 CET] <prelude2004c>  X5650
[06:43:21 CET] <prelude2004c> they are multhreaded 6's
[06:43:32 CET] <Diag> so not a pos
[06:44:24 CET] <Diag> dunno though
[06:44:29 CET] <prelude2004c> it's weird.. i just want to deinterlace source.. and then do whatever.
[06:44:32 CET] <Diag> interpolation is pretty cpu heavy
[06:44:39 CET] <Diag> why yadif, why not just blend or whatever
[06:44:50 CET] <prelude2004c> blend ? how
[06:44:56 CET] <Diag> idk is that not a thing
[06:44:59 CET] <prelude2004c> not familiar with that
[06:45:55 CET] <Diag> tblend?
[06:45:59 CET] <Diag> it appears? I dk
[06:46:10 CET] <Diag> >The tblend (time blend) filter takes two consecutive frames from one single stream, and outputs the result obtained by blending the new frame on top of the old frame.
[06:47:06 CET] <Diag> idunno, kinda talkin outta my ass here as im not familiar with ffmpeg
[06:47:19 CET] <Diag> I just know blend is a popular shit because its low overhead
[06:48:01 CET] <prelude2004c> sounds like fancy filters
[06:49:45 CET] <Diag> dunnno
[06:49:57 CET] <Diag> youre probably doin some crazy ass shit
[06:50:04 CET] <Diag> Since i know someones gonna ask
[06:50:11 CET] <Diag> you should post the full command youre running
[06:51:04 CET] <prelude2004c> minterpolate < i tried that too
[06:51:17 CET] <prelude2004c> crazy?? not too crazy no
[06:51:58 CET] <Diag> No i mean like
[06:52:23 CET] <Diag> you probably got 29.97 video thats 15 full frames a second
[06:52:35 CET] <Diag> and youre tryna boost it up to 60full frames a second
[06:52:44 CET] <Diag> thatd drive your cpu usage through the roof if it was interpolating it all
[07:03:03 CET] <prelude2004c> well its 29.95 interlaced
[07:03:09 CET] <prelude2004c> and i want 59.94 deinterlaced
[07:03:21 CET] <prelude2004c> just not working.. trying different things and trying to be as efficient as possible.
[07:03:34 CET] <prelude2004c> nobody seems to know :P
[07:03:36 CET] <Diag> you can only be so efficient...
[08:10:57 CET] <kazuma_> you just bob it prelude2004c
[08:46:52 CET] <prelude2004c> kazuma_, I did bob it
[08:46:56 CET] <prelude2004c> i did -deint bob
[08:47:00 CET] <prelude2004c> still nothing .. doesn't do it
[09:15:42 CET] <kazuma_> prelude2004c
[09:15:51 CET] <kazuma_> -vf yadif=1:0
[10:00:45 CET] <slasktrat> Hi. I've updated an one year old, unclosed ticket (http://trac.ffmpeg.org/ticket/6165) do I need to send a notification somewhere?
[10:14:22 CET] <blop> what ./configure option enables --vo=opengl:backend=x11egles  ?
[10:24:44 CET] <blop> ah i already have egl-drm egl-helpers egl-x11
[10:25:03 CET] <blop> in mpv
[12:14:17 CET] <Kapricornus> Hello
[12:29:19 CET] <Projec[T]MayheM> Hello guys
[12:29:28 CET] <Projec[T]MayheM> anyone can please help me with a ffmpeg segment error?
[12:31:41 CET] <Projec[T]MayheM> I'm recording an rtsp stream in 1 minute chunks but some segments show a 1 hour plus 1 minute duration
[12:31:46 CET] <Projec[T]MayheM> with ffprobe I get this: [mpegts @ 0x2e86960] DTS discontinuity in stream 0: packet 31 with DTS 575999, packet 32 with DTS 337043505
[12:32:11 CET] <Projec[T]MayheM> the DTS jumps from 500k to 337M
[12:32:15 CET] <Projec[T]MayheM> anyone?
[12:35:00 CET] <JEEB> mpegts can have fun timestamp stuff
[12:35:43 CET] <Projec[T]MayheM>  any way to reset this?
[12:35:53 CET] <JEEB> unless that mpeg-ts is written by you, in which case you should check the input timestamps from RTSP
[12:36:39 CET] <Projec[T]MayheM> I do this "ffmpeg -y -i rtsp://10.28.2.195/proxyStream-1 -c copy -f segment -segment_time 60 -reset_timestamps 1 -segment_atclocktime 1 -strftime 1 test%Y-%m-%d_%H-%M-%Stest.ts
[12:36:43 CET] <Projec[T]MayheM> ops
[12:36:44 CET] <Projec[T]MayheM> ffmpeg -y -i rtsp://10.28.2.195/proxyStream-1 -c copy -f segment -segment_time 60 -reset_timestamps 1 -segment_atclocktime 1 -strftime 1 test%Y-%m-%d_%H-%M-%Stest.ts
[12:36:55 CET] <Projec[T]MayheM> is there a flag I can use to reset the timestamps?
[12:38:31 CET] <Projec[T]MayheM> maybe a setpts filter?
[13:00:11 CET] <zerodefect> Using the C-API to change the properties of the overlay filter dynamically. I'm using the 'avfilter_graph_send_command' commands to update the overlay properties of the filter before pushing in a new frame.
[13:00:18 CET] <zerodefect> I'm getting an error: 'ENOSYS' or "Function not implemented"
[13:01:04 CET] <zerodefect> Now I'm not to sure on the difference between the 'cmd' and 'arg' params to 'avfilter_graph_send_command'
[13:01:49 CET] <zerodefect> So far, I'm specifying "x=10:y=10:W=720:H=576:w=253:h=253" for cmd and NULL for arg :D
[13:16:26 CET] <furq> pretty sure that should be arg and cmd should be "overlay"
[13:24:34 CET] <styler2go> Is it possible to use ffmepg as a rstp stream relay? (I receive a rtsp stream and provide it from there to other devices) It's to lower the data bandwith from my security camera to a lan-connected device
[13:24:35 CET] <zerodefect> @furq Does the "overlay" not part of the 'target' ?
[13:25:22 CET] <zerodefect> @furq Does the "overlay" string not belong to the 'target' parameter?
[13:25:46 CET] <furq> i could only find one example of someone using it and it had the filter name in target and cmd
[13:27:40 CET] <zerodefect> Are you able to share that example with me ? :)
[13:27:59 CET] <furq> https://github.com/PromyLOPh/pianobar/blob/master/src/player.c#L99
[13:28:05 CET] <furq> it's not that enlightening really
[13:29:08 CET] <jesopo> Hello guys
[13:33:23 CET] <zerodefect> @furq Interesting that the filter's name is 'volume' and it looks like the property being adjusted is 'volume'. I think that is why they match
[13:33:39 CET] <zerodefect> ...but its interesting that the property value goes in arg.
[13:35:59 CET] <Projec[T]MayheM> I'm recording an rtsp stream in 1 minute chunks but some segments show a 1 hour plus 1 minute duration. with ffprobe I get this: [mpegts @ 0x2e86960] DTS discontinuity in stream 0: packet 31 with DTS 575999, packet 32 with DTS 337043505, the DTS jumps from 500k to 337M resulting in 1 hour jump. Anyone?
[13:55:32 CET] <JEEB> Projec[T]MayheM: as I noted check the timestamps from the input, in the worst case use tcdump to dump the traffic to be able to replay it for testing purposes
[13:55:53 CET] <JEEB> and yes, I have no idea how well the segmenting works
[13:57:17 CET] <Projec[T]MayheM> thanks dude
[14:17:47 CET] <styler2go> I am trying to receive a rtsp stream buzt ffmpeg keep sspamming: Past duration 6.499992 too large
[14:17:50 CET] <styler2go> what does that mean?
[14:34:12 CET] <TheoTheo> Is anybody familiar with the message: "Input pad "default" with type audio of the filter instance "in" of afade not connected to any source"?. Trying to get a fade filter working in C code using the libavfilter library
[14:35:22 CET] <TheoTheo> Am using version 2.8.13 of FFmpeg btw
[15:04:19 CET] <tommy``> guys how can i change framerate from 23.976 to 25 and stretch audio 0.95904 ?
[15:07:47 CET] <Nacht> Is 'master_pl_name' something new of HLS ?
[15:08:16 CET] <JEEB> the hls muxer has recently gotten a lot of new stuff by akamai and the chinaffmpeg folk
[15:08:33 CET] <JEEB> https://www.ffmpeg.org/ffmpeg-all.html#hls-2
[15:08:39 CET] <JEEB> see the options here
[15:09:09 CET] <JEEB> TheoTheo: you haven't connected the filters yet?
[15:09:10 CET] <Nacht> Yeah I was looking at it, and suddenly saw the master function. Couldn't recall seeing it before. I'm sure I was looking for it in the past
[15:09:26 CET] <JEEB> yea, the hls muxer until very recently only did the per-profile playlists
[15:11:38 CET] <TheoTheo> JEEB: Still can't figure out the order of things to make a filter graph that works for me
[15:12:21 CET] <TheoTheo> Does a fade need to be preceded by a diffreent filter?
[15:55:05 CET] <waldner> I'm using the following command to segment a stream coming from a camera: ffmpeg -i 'rtsp://.....' -c copy -f segment -segment_time 10 /tmp/capture-%03d.mp4
[15:55:15 CET] <tommy``> is t possible capture stream like amazon prime?
[15:55:19 CET] <tommy``> with ffmpeg
[15:55:33 CET] <waldner> however after writing 3 files, ffmpeg exits by itself without apparent error messages
[15:55:38 CET] <waldner> am I doing something wrong?
[15:57:04 CET] <JEEB> -v verbose or -v debug
[16:00:28 CET] <waldner> this is how it ends with -v debug: https://pastebin.com/raw/ThySknxZ
[16:00:41 CET] <waldner> the camera is a cisco IPC 6400
[16:02:28 CET] <waldner> in fact it stops after 30 seconds even without segmentation
[16:02:32 CET] <waldner> just dumping to a file
[16:02:40 CET] <waldner> (just tried it now)
[16:07:40 CET] <waldner> ffmpeg version 3.4.1 Copyright (c) 2000-2017 the FFmpeg developers
[16:10:18 CET] <waldner> if I select another camera channel it doesn't stop...I guess this can be a workaround for now
[16:17:04 CET] <waldner> another thing, for old segments to be automatically deleted from disk is hls the only muxer? I can't find a similar option for -f segment
[16:33:31 CET] <waldner> (segment_wrap is useless with -strftime 1)
[00:00:00 CET] --- Wed Feb  7 2018