[Ffmpeg-devel-irc] ffmpeg.log.20160712

[burek]

[04:25:16 CEST] <Bray90820> "pts has no value" what does that warning mean
[11:51:37 CEST] <pskosinski> Guys, I'm trying to record desktop with recordmydesktop, it's using ffmpeg. I have a problem with parts of my wallpaper getting visible sometimes, like that: https://imgur.com/0XT7jnf
[11:52:41 CEST] <pskosinski> I'm getting many artefacts if anything on my screen changes position, when I scroll webpage or even move mouse pointer
[12:20:44 CEST] <lilis> Hi guys, I want to combine two webcam streams and output it as another webcam stream. Is it possible to do?
[12:21:07 CEST] <lilis> I want to do something like 360 video using ffmpeg
[12:23:22 CEST] <Fyr> guys, I have a very blurry text in a movie. how do I make it sharpened?
[12:24:15 CEST] <Fyr> I have a screenshot of a video that I'm trying to magnify and recognize label written on the wall.
[13:02:49 CEST] <lilis> I figured out how to place two streams onto a single output file, but I can't seem to get it to output to /dev/video2 (through v4l2loopback)
[13:02:55 CEST] <lilis> Does anybody know why?
[13:05:31 CEST] <lilis> This is one of the errors: [v4l2 @ 0x562187227920] Unknown V4L2 pixel format equivalent for yuva420p
[13:05:58 CEST] <lilis> Actual command is
[13:06:00 CEST] <lilis> ffmpeg -re -i /dev/video0 -i /dev/video1 -filter_complex "nullsrc=size=640x480 [base]; [0:v] setpts=PTS-STARTPTS, scale=320x240 [upperleft]; [1:v] setpts=PTS-STARTPTS, scale=320x240 [upperright]; [base][upperleft] overlay=shortest=1 [tmp1]; [tmp1][upperright] overlay=shortest=1:x=320" -f v4l2 /dev/video2
[13:19:14 CEST] <whald> lilis, it seems you pipe is generating frames w/ an alpha channel, which is v4l sink cannot accept. the acceptable pix fmts are listed here: https://www.ffmpeg.org/doxygen/3.0/v4l2-common_8c_source.html#l00064
[13:20:16 CEST] <whald> lilis, i'm not too familiar with the ffmpeg command line so i'm afraid i don't know where, but probably you'll need to throw in a -pix_fmt=<...> somewhere.
[13:21:41 CEST] <lilis> whald, thanks! I'll try that out
[13:22:27 CEST] <furq> add ;format=yuv420 to the end of filter_complex
[13:22:32 CEST] <furq> yuv420p
[13:23:09 CEST] <furq> ,format=yuv420p
[13:23:11 CEST] <furq> i got there in the end
[13:23:54 CEST] <lilis> D:
[13:24:00 CEST] <lilis> It worked!
[13:24:05 CEST] <lilis> Thanks
[13:27:03 CEST] <lilis> Is there a way to reduce lag between the streams?
[13:27:31 CEST] <lilis> It _is_ different webcams, so I might be asking too much
[13:28:10 CEST] <whald> lilis, are you talking about some offset between the different inputs or total input->output latency?
[13:28:55 CEST] <lilis> Different inputs I believe
[13:29:14 CEST] <lilis> It shows in the combined output
[13:29:39 CEST] <lilis> brb
[14:14:21 CEST] <hdare> is there any web player supporting the stream  from ffmeg trough icecast ?
[14:14:46 CEST] <hdare> i mean video stream
[14:17:19 CEST] <BtbN> Nothing that's not flash, and even flash is quite restricted.
[14:17:42 CEST] <BtbN> Use HLS or DASH for streaming to a Browser.
[14:18:57 CEST] <furq> hdare: you should just be able to point an html5 video tag at it
[14:19:09 CEST] <furq> whether or not it'll work well is another question
[14:27:04 CEST] <__jack__> hls on browser, hum
[14:27:21 CEST] <__jack__> won't work, unless you're using flash, or unless "browser" = "safari"
[14:27:31 CEST] <BtbN> hls.js works fine in every browser.
[14:27:41 CEST] <BtbN> And HLS is less insane than DASH
[14:29:01 CEST] <ritsuka> hls works on edge too, and some version of the android browser
[14:29:15 CEST] <__jack__> consumes a lot a CPU, no ?
[14:29:21 CEST] <BtbN> It works in every Browser that supports MSE, so, all of them.
[14:29:35 CEST] <__jack__> I just tested http://dailymotion.github.io/hls.js/demo/, doh my poor PC
[14:29:56 CEST] <BtbN> So your PC doesn't support hardware decoding?
[14:30:27 CEST] <BtbN> Firefox plays this while using 6% CPU
[14:30:35 CEST] <BtbN> while it's using 5% when doing nothing...
[14:30:52 CEST] <__jack__> dunno, you may be right
[14:31:06 CEST] <__jack__> it worth looking at that code, thanks for the tip
[14:35:08 CEST] <hdare> with video.js and medialement the playback of hls works well and with clappr too
[14:35:52 CEST] <hdare> but how to stream in hLS mode ?
[14:36:11 CEST] <hdare> trough icecast
[14:36:28 CEST] <BtbN> I don't think icecast supports HLS.
[14:36:49 CEST] <hdare> it s what i think too
[14:37:54 CEST] <hdare> html5 accept only audio stream but no video
[14:38:23 CEST] <BtbN> Browsers support only mp4 for video
[14:38:28 CEST] <BtbN> And mp4 can't be streamed.
[14:38:37 CEST] <BtbN> So it's cut into short segments.
[14:40:05 CEST] <hdare> it s must be streamed to an rtmp server to redirect stream to hls format
[14:42:52 CEST] <viric> does firefox have hardware decoding?
[14:43:06 CEST] <BtbN> for h264 and stuff, yes
[14:43:11 CEST] <BtbN> Not for VP9 i think
[14:43:12 CEST] <furq> hdare: ffmpeg has an hls muxer
[14:43:14 CEST] <viric> I thought it did not
[14:43:26 CEST] <viric> (on GNU/Linux)
[14:43:32 CEST] <furq> if you don't need to push to a remote server then you can just serve the fragments with any httpd
[14:43:43 CEST] <BtbN> viric, check about:support for hardware decoder status.
[14:43:58 CEST] <BtbN> for me it says "Hardware H264 Decoding:	Yes; Using D3D11 API"
[14:44:10 CEST] <furq> and yeah hls.js doesn't use much cpu in my experience
[14:44:17 CEST] <viric> Supports Hardware H264 Decoding No
[14:44:24 CEST] <viric> BtbN: that's Windows
[14:44:37 CEST] <BtbN> I'm quite sure it supports hw accell on Linux as well
[14:44:43 CEST] <BtbN> they recently switched to ffmpeg for that iirc
[14:44:48 CEST] <furq> No; D3D9 blacklisted with DLL atiumd64.dll (9.14.10.1128)
[14:44:49 CEST] <furq> huh
[14:44:54 CEST] <furq> you'd think i'd have noticed
[14:45:17 CEST] <BtbN> And before that it was done via gstreamer
[14:45:33 CEST] <BtbN> So if you install the neccesary gstreamer plugins, it can use whatever hwaccel gst supports.
[14:46:34 CEST] <viric> BtbN: hm. I remember a bugzilla ticket about not being available at all.
[14:49:11 CEST] <viric> I see it builds something from firefox-47.0.1/media/ffvpx/libavcodec/x86/...
[14:53:36 CEST] <viric> h264 being inside 'ffvpx' sounds strange
[14:55:48 CEST] <BtbN> they only use lavc, and only for its VP8/9 decoder
[14:56:15 CEST] <viric> BtbN: https://hg.mozilla.org/mozilla-central/file/tip/media/ffvpx/README_MOZILLA
[14:56:25 CEST] <viric> --disable-vaapi --disable-vdpau
[14:57:59 CEST] <viric> vp8 and vp9, ok
[14:58:09 CEST] <viric> BtbN: and where is the h264 part?
[14:58:25 CEST] <DHE> aside, isn't --disable-everything redundant with all the other --disable-* commands?
[14:58:35 CEST] <BtbN> it should be
[14:59:22 CEST] <furq> other way round, surely
[15:02:43 CEST] <viric> BtbN: what about that Cisco openh264?
[15:02:56 CEST] <DHE> isn't that baseline only?
[15:03:02 CEST] <BtbN> It's an encoder, for baseline only.
[15:03:05 CEST] <viric> ahh.
[15:03:10 CEST] <DHE> oh, I thought it was a decoder as well
[15:03:19 CEST] <furq> it does decoding as well
[15:03:21 CEST] <furq> but presumably only baseline
[15:03:24 CEST] <viric> then where is the decoder source of h264 in the firefox tree?I Can't find it
[15:03:35 CEST] <BtbN> viric, firefox does not contain any h264
[15:03:39 CEST] <viric> mh
[15:03:42 CEST] <viric> how does it decode it?
[15:03:46 CEST] <BtbN> that would require them to pay patent feees
[15:03:47 CEST] <BtbN> it doesn't
[15:03:54 CEST] <viric> how does it play h264?
[15:03:54 CEST] <BtbN> it relies on external stuff doing it
[15:04:01 CEST] <viric> like what? You said that it uses ffmpeg
[15:04:06 CEST] <BtbN> It does
[15:04:12 CEST] <BtbN> Installed via your package manager.
[15:04:21 CEST] <viric> auhm
[15:04:27 CEST] <BtbN> And on windows it uses the Microsoft Media APIs
[15:04:45 CEST] <BtbN> Might also still use gst, no idea how far their migration got.
[15:05:01 CEST] <viric> I fail to find it in the recipe
[15:05:59 CEST] <BtbN> I remember some Firefox developer asking in ffmpeg-devel. They seem to have come up with a complex runtime-dynamic-loader for libav*, that detect the version and uses the correct ABI at runtime.
[15:06:14 CEST] <viric> ahhh
[15:06:22 CEST] <viric> and how does it check for hardware acceleration?
[15:06:29 CEST] <BtbN> No idea
[15:14:57 CEST] <viric> I can't find it
[15:18:38 CEST] <viric> BtbN: how would my system ffmpeg decide to decode accelerated or not accelerated?
[15:19:05 CEST] <BtbN> The library user has to decide that. ffmpeg hwaccels don't just work.
[15:19:13 CEST] <viric> Aha
[15:19:15 CEST] <BtbN> It needs quite a bit of work from the user.
[15:19:29 CEST] <viric> then it does not do anything I'd say
[15:20:03 CEST] <viric> https://hg.mozilla.org/mozilla-central/file/tip/dom/media/platforms/ffmpeg
[15:23:43 CEST] <viric> it sets up decoder threads... https://hg.mozilla.org/mozilla-central/file/tip/dom/media/platforms/ffmpeg/FFmpegVideoDecoder.cpp
[15:23:46 CEST] <viric> no accel
[15:24:26 CEST] <viric> BtbN: how is it accelerated on windows?Doesn't it use ffmpeg also?
[15:24:45 CEST] <BtbN> No, it uses the Windows API for decoding and playing stuff
[15:24:55 CEST] <viric> the WMF thing
[15:25:09 CEST] <viric> that *yes* has things about hw acceleration.
[15:25:48 CEST] <viric> also android, also omx (rpi), .. not ffmpeg, there.
[15:28:03 CEST] <BtbN> It will use whatever is available on the platform it's running on.
[15:28:13 CEST] <BtbN> Linux doesn't have a native API, so it uses ffmpeg or gst there.
[15:34:02 CEST] <viric> BtbN: but it could set up hwaccel in ffmpeg
[15:34:26 CEST] <viric> in any case, it matches what I knew about it for more than one year: no support at all for hw acceleration in GNU/Linux
[15:42:33 CEST] <DHE> in theory you can use vdpau under linux, sure. but there's usually a trick to playback, like requiring the matching vdpau video-output driver for rendering. it may not be easy, or possible, to get the raw decoded video frames out
[15:42:51 CEST] <DHE> similar for other hwaccel methods. which is why it doesn't "just work" like the software decoders
[15:45:44 CEST] <BtbN> well, you can always just pull the frames back to system memory and continue as usual
[15:45:54 CEST] <BtbN> Just like ffmpeg.c does it
[15:46:11 CEST] <pgorley> doesn't that hunder performance though?
[15:46:35 CEST] <pgorley> *hinder, sorry
[15:46:47 CEST] <BtbN> sure, but still faster than decoding in software.
[15:47:02 CEST] <pgorley> true
[17:18:50 CEST] Action: Fyr slaps durandal_1707 around a bit with a large trout
[17:59:48 CEST] <durandal_1707> Fyr: ?
[18:00:28 CEST] <Fyr> you wrote about ffmpeg-mt a few months ago. have it been merged with the master branch?
[18:01:15 CEST] <durandal_1707> you mean mt audio encoding?
[18:01:27 CEST] <Fyr> yes
[18:04:19 CEST] <k_sze> Is there a streaming format that can readily embed a real wall clock timestamp with each frame, as metadata?
[18:06:28 CEST] <Fyr> I would dig into TS.
[18:07:47 CEST] <k_sze> I see that ffmpeg has a use_wallclock_as_timestamps option, and the nut container format also has a -syncpoints flag.
[18:07:55 CEST] <k_sze> I wonder how usable those are.
[18:08:32 CEST] <k_sze> I would be somewhat surprised if all container formats allow wallclock timestamps.
[18:08:32 CEST] <Fyr> durandal_1707, still, the 16-core Intel Xeon cluster converts FLAC as fast as my Banana Pi.
[18:08:48 CEST] <Fyr> that thing makes me sad.
[18:09:17 CEST] <k_sze> FLAC encoding doesn't multithread readily, IIRC.
[18:10:04 CEST] <Fyr> k_sze, wrong, a file can be divided into pieces and then merged into one.
[18:10:12 CEST] <Fyr> each piece can be converted with FFMPEG.
[18:10:20 CEST] <Fyr> a simple bash wrapper can do that.
[18:10:49 CEST] <k_sze> And have you looked at your CPU usage?
[18:10:56 CEST] <k_sze> Maybe the CPUs are not the bottleneck.
[18:10:58 CEST] <Fyr> unfortunately, the developers don't want the users to be happy.
[18:11:29 CEST] <k_sze> Maybe it's the I/O, if you don't have fast enough storage to saturate 16 cores.
[18:12:26 CEST] <Fyr> 16-CORE INTEL XEON CLUSTER DOESN'T HAVE FAST ENOUGH STORAGE?
[18:12:36 CEST] <Fyr> how about my Banana Pi?
[18:12:53 CEST] <k_sze> Is your Xeon cluster running on SSD?
[18:13:29 CEST] <Fyr> it has an SSD, but for users it uses an HDD.
[18:13:50 CEST] <k_sze> And your audio file is on the spinning HDD.
[18:14:02 CEST] <Fyr> I/O has never been a problem on the cluster.
[18:14:07 CEST] <k_sze> measure
[18:14:15 CEST] <Fyr> but, the Banana Pi seems to have one.
[18:14:21 CEST] <catchin> is this ffmpeg at work? https://www.instagram.com/p/BHwDgD0AU4E/
[18:20:03 CEST] <durandal_1707> Fyr: I wrote patch ...
[18:21:08 CEST] <kepstin> as far as mpeg-ts goes, the timestamp format only holds ~26.5 hours iirc; they're designed to be able to hold a 24-hour realtime timestamp, wrapping each day. If you want to also store the date, it would have to be in metadata - which i'm sure there's a way to do, I just don't know it :)
[18:22:56 CEST] <k_sze> Fyr: how's the build quality of the banana pi? What model do you have btw.
[18:23:07 CEST] <Fyr> Banana Pro
[18:23:18 CEST] <k_sze> I have a Raspberry Pi 3 and an incoming Raspberry Pi Zero.
[18:23:19 CEST] <Fyr> I'm pretty much satisfied with the purchase.
[18:23:42 CEST] <k_sze> Maybe I'll try the Banana Pi sometime later.
[18:23:55 CEST] <Fyr> if you try to use it hard, you should buy something more reliable such as Intel NUC, ZBOX or whatever else.
[18:24:09 CEST] <k_sze> Fyr: can't see a Banana Pro listed here: http://www.banana-pi.com/eindex.asp
[18:24:16 CEST] <k_sze> "hard"?
[18:24:26 CEST] <Fyr> Banana Pi may be hot under full load of the CPU.
[18:24:31 CEST] <k_sze> ah
[18:24:47 CEST] <Fyr> k_sze, https://en.wikipedia.org/wiki/Banana_Pi#Banana_Pi_Pro
[18:26:57 CEST] <k_sze> ah
[18:35:09 CEST] <friki2015> Some days ago i asked here for help about rtmp ingest to Akamai. Finaly i got it working using ffmpeg rtmp implementation (non librtmp). ffmpeg ... -f flv "rtmp://USER:PASSWD@p.epCPCODE.i.akamaientrypoint.net/EntryPoint/STREAMNAME_ANGLE_BITRATE@STREAMID"
[18:36:16 CEST] <friki2015> just take care about avoid open new connection simultaneously (akamai reject logins in this situation, as they have internaly documented)
[18:52:49 CEST] <kbarry> I'm trying to grab only the audio from a single programs in an HLS stream. I'm a little confused with the syntax. I believe I need to be using -map, ie, 0:p:2 should refer to the streams from the program "1", but I am  unable to select just the audio stream from this particular program.
[18:56:24 CEST] <furq> Fyr: you could use flaccl if your encoding box has a decent gpu
[18:56:30 CEST] <__jack__> p ? use -map 0:a:0
[18:56:32 CEST] <furq> it's .net though which sucks
[18:56:44 CEST] <__jack__> (as in : first input, first audio stream)
[18:56:53 CEST] <Fyr> furq, flacCL doesn't have descent compression.
[18:57:08 CEST] <Fyr> however, it's 60x times faster, than the cluster.
[18:57:20 CEST] <furq> er
[18:57:25 CEST] <furq> neither of those tally with my experience
[18:57:38 CEST] <furq> although i haven't used it on a decent gpu
[18:58:31 CEST] <furq> last time i tried it was roughly on a par with flac -8 and ~3x as fast
[18:59:26 CEST] <Fyr> FFMPEG has the best FLAC compression with -multi-dim on.
[18:59:35 CEST] <DHE> kbarry: 0:a means all audio streams (there should only be one). you can also use ffprobe to report stats on the stream
[19:00:04 CEST] <kbarry> DHE: Well, ther eis one other problem.
[19:00:31 CEST] <kbarry> This is a TV signal. I'm trying to strip out the audio only, but it appears that the signal configurations is changing drastically when commercials appear
[19:01:12 CEST] <kbarry> http://pastebin.com/S0eqJWN3 during News Broadcast
[19:03:06 CEST] <DHE> oh... yeah that's always a pain when the PMT changes
[19:10:04 CEST] <friki2015> kbarry_: -map p:<program_id>:<audio_stream_index> or just -map p:5 -vn
[19:11:35 CEST] <kbarry_> well, the stream configuration keeps changing,
[19:12:37 CEST] <kbarry_> between  "commercial" and "non-commercial" periods.
[19:13:08 CEST] <friki2015> i never faced this problem, sorry
[19:16:01 CEST] <friki2015> could be you can fix it after recording all audios (detect commercial timeranges and remove them -ss start -to firstcommercial)
[19:48:14 CEST] <maziar> hi i want to know everything about converting MP4 to m3u8 and other formats, what should i read? i need something like book, pdf and etc.
[19:52:53 CEST] <Fyr> maziar, some short manuals on "how to convert MP4 into MP3" or "how to remux an MP4".
[19:53:03 CEST] <Fyr> there are plenty of them.
[19:53:46 CEST] <maziar> Fyr, i need a perfect document perfect and full
[19:53:49 CEST] <Jnorthrup> what is the mp4 encoding mojo to get "422 hq" ?
[19:54:34 CEST] <Fyr> maziar, https://wiki.archlinux.org/index.php/FFmpeg
[19:55:27 CEST] <Fyr> that one is good too:
[19:55:28 CEST] <Fyr> http://trac.ffmpeg.org/
[19:55:41 CEST] <maziar> fry how about this : https://www.ffmpeg.org/documentation.html
[19:56:02 CEST] <maziar> Fyr, which part should i read fot that ?
[19:56:05 CEST] <furq> Jnorthrup: isn't that prores
[19:56:27 CEST] <furq> maziar: https://www.ffmpeg.org/ffmpeg-formats.html#hls-1
[20:08:16 CEST] <Fyr> maziar, every.
[20:08:43 CEST] <maziar> Fyr, sorry, every what ?
[20:08:51 CEST] <Fyr> every part
[20:09:02 CEST] <Fyr> you needed full manual.
[20:09:56 CEST] <Fyr> those ones cover conversion broadly.
[20:10:17 CEST] <furq> is there anyone in the entire world who's read all of the ffmpeg docs
[20:10:37 CEST] <furq> not including the libav* docs
[20:14:37 CEST] <maziar> Fyr, for full manual, what should i do ?
[20:15:54 CEST] <llogan> man ffmpeg-full
[20:16:12 CEST] <llogan> *man ffmpeg-all
[20:19:08 CEST] <wiak> the windows compile guide for av1 is a whreck so i made a better one https://nwgat.ninja/compile-av1-using-msys2-mingw/
[20:19:09 CEST] <wiak> ^^
[21:47:11 CEST] <Mista-D> "-xerror" exits when using "-f concat" with in/out timestamps,  at  "[ac3 @ ...] frame sync error".  Anyway to set concat to not break AC3 frame sync or force xerror to ignore the AC3 error please?
[21:51:20 CEST] <Mista-D> http://pastebin.ca/3657801
[23:31:50 CEST] <Bray90820> Can someone tell me what this error means?
[23:31:50 CEST] <Bray90820> http://pastebin.com/raw/qnGDmMk4
[23:36:32 CEST] <jkqxz> "codec not currently supported in container"?  It means that the codec is not currently supported in the container.
[23:36:49 CEST] <Bray90820> No shit
[23:36:51 CEST] <jkqxz> (Try writing to a different container instead (Matroska, say), or just transcode.)
[23:37:24 CEST] <Bray90820> so something other than MP4?
[23:37:38 CEST] <jkqxz> Yeah.
[23:39:05 CEST] <Bray90820> I just transcoded it
[23:39:15 CEST] <friki2015> is there a recommended way to seek to the end of a "growing" video file? somethink like -ss "-00:00:20"
[23:39:35 CEST] <friki2015> of course i can get the duration and seek acoddingy
[23:41:58 CEST] <jkqxz> Bray90820:  To that?  Some random MPEG-4-alike is not really a sensible thing to use nowadays - you're probably better off transcoding again to something more modern, such as H.264 or VP9.
[23:45:56 CEST] <Bray90820> jkqxz: I just used -c:a instead of -c
[23:48:39 CEST] <jkqxz> And now it's transcoding the video to whatever the default video codec is, using some default settings you probably don't want, and that works in mp4.
[23:49:58 CEST] <jkqxz> (While copying the audio stream directly.)
[23:52:53 CEST] <Danimal> I have a thing that's streaming mpegts out to a multicast address. Everything is fine. it seems that once I have 6 of them sending to different multicast addresses on the same network, the audio sync gets off by about a second(a lot).
[23:53:48 CEST] <Danimal> this seems very weird to me. Anyone have an idea what might be going on?
[23:56:00 CEST] <Bray90820> jzHow would I transcode it to h.264 then
[23:56:05 CEST] <Bray90820> jkqxz:
[23:57:14 CEST] <Bray90820> Or copy it
[23:57:54 CEST] <pfelt> afternoon all.  how does the native hevc performance compare with libx265 ?
[23:58:11 CEST] <pfelt> (at least as i understand it, that's an apples-to-apples comparison)
[23:59:46 CEST] <jkqxz> Bray90820:  Transcode would be something like "ffmpeg -i input.avi -c:a copy -c:v libx264 -crf 20 output.mp4" (with a suitable value for 20).  For copy, you need to pick a different container, so "ffmpeg -i input.avi -c copy output.mkv" or similar.
[00:00:00 CEST] --- Wed Jul 13 2016