[Ffmpeg-devel-irc] ffmpeg.log.20170415

Sun Apr 16 03:05:01 EEST 2017

[00:00:01 CEST] <thebombzen> it's total overkill. doing that has the primary benefit of "now I have bash, git, and coreutils"
[00:00:05 CEST] <thebombzen> "and I can run shell scripts"
[00:00:07 CEST] <thebombzen> etc. etc.
[00:00:13 CEST] <furq> i'm serious. i intend to argue this point until everyone agrees with me
[00:00:20 CEST] <furq> because otherwise i have to and align wall textures
[00:00:37 CEST] <furq> +go
[00:01:58 CEST] <thebombzen> align wall textures?
[00:01:58 CEST] <thebombzen> wut
[00:02:05 CEST] <thebombzen> why would you have to do that
[00:02:12 CEST] <furq> because otherwise they'll be misaligned
[00:02:16 CEST] <thebombzen> so
[00:02:53 CEST] <thebombzen> I'm going to wake up to a sulfuric atmosphere as lava rains down and volcanic eruptions echo throughout the land
[00:03:13 CEST] <thebombzen> a weary old man will arrive and speak of the apocalypse
[00:03:28 CEST] <furq> he's a bit late isn't he
[00:03:28 CEST] <thebombzen> the day has come when wall textures were misaligned
[00:03:56 CEST] <thebombzen> if only furq had aligned the wall textures, this disappointingly cliche end-of-the-world would not have happened
[00:04:05 CEST] <thebombzen> but alas, this did not come to be, for furq neglected on his duties
[00:04:11 CEST] <furq> if you're happy with giant seams in walls then good for you
[00:14:25 CEST] <djk> hmmm
[00:14:33 CEST] <djk> [v4l2 @ 0x23c50f0] ioctl(VIDIOC_G_FMT): Invalid argument
[00:14:33 CEST] <djk> Could not write header for output file #0 (incorrect codec parameters ?): Invalid argument
[00:14:33 CEST] <djk> Error initializing output stream 0:0 --
[00:14:33 CEST] <djk> [video4linux2,v4l2 @ 0x23c3380] Some buffers are still owned by the caller on close.
[01:13:05 CEST] <mavi>   I'm trying to do one command with complex filters that allow for denoising, hardsub, scaling (to 480 height) and watermark.  I've gotten 3/4 filters to work in combination with each other
[01:13:14 CEST] <mavi> ffmpeg -i test720.mp4 -vf hqdn3d=1.5:1.5:6:6,scale=w="if(gt(dar\,854/480)\,min(854\,iw*sar)\,2*trunc(iw*sar*oh/ih/2))":h="if(gt(dar\,854/480)\,2*trunc(ih*ow/iw/sar/2)\,min(480\,ih))"\:flags=bicubic,setsar=1\,subtitles=add.ass[out]" -c:v libx264 -crf 18 -preset veryslow -tune film -refs 8 -bf 6 -aq-mode 2  -c:a copy -f mp4 720encoded.mp4
[01:13:24 CEST] <mavi> but when trying to do all 4 my syntax doesn't work.
[01:13:33 CEST] <mavi>  ffmpeg -i test720.mp4 -vf hqdn3d=1.5:1.5:6:6,scale=w="if(gt(dar\,854/480)\,min(854\,iw*sar)\,2*trunc(iw*sar*oh/ih/2))":h="if(gt(dar\,854/480)\,2*trunc(ih*ow/iw/sar/2)\,min(480\,ih))"\:flags=bicubic,setsar=1\,"movie=actorimage.png[wm];[in][wm]overlay=30:main_h-overlay_h-120\,subtitles=add.ass[out]" -c:v libx264 -crf 18 -preset veryslow -tune film -refs 8 -bf 6 -aq-mode 2  -c:a copy -f mp4 720encoded.mp4
[01:13:59 CEST] <mavi> Any ideas what may be wrong with my syntax? When i add in the watermark, it all goes to hell and gives me a Too many inputs specified for the "movie" filter.  error.
[01:36:50 CEST] <c_14> you escape too much stuff
[01:37:51 CEST] <c_14> try ffmpeg -i test720.mp4 -filter_complex "hqdn3d=1.5:1.5:6:6,scale=w=if(gt(dar\,854/480)\,min(854\,iw*sar)\,2*trunc(iw*sar*oh/ih/2)):h=if(gt(dar\,854/480)\,2*trunc(ih*ow/iw/sar/2)\,min(480\,ih)):flags=bicubic,setsar=1[we];movie=actorimage.png[wm];[we][wm]overlay=30:main_h-overlay_h-120,subtitles=add.ass[out]"
[01:38:08 CEST] <c_14> oh, you need the filter_complex instead of -vf because you're using a filtergraph instead of a single filterchain
[01:39:53 CEST] <c_14> getting rid of the movie filter and using another input might make it a bit simpler too
[01:45:50 CEST] <utack_> Is there a way to get the real time of the "benchmark" mode, other than using unix time? it only shows CPU time
[01:48:09 CEST] <Threads> anyone know what 100ms is to frames ?
[01:55:35 CEST] <BtbN> to frames? Around 100ms I'd say. What do you mean?
[02:04:42 CEST] <Threads> BtbN what is a frame to audio ?
[02:05:23 CEST] <BtbN> A format dependant number of samples
[02:05:31 CEST] <BtbN> something like 1024 or 4096 samples
[02:09:03 CEST] <Threads> ok thanks
[02:26:15 CEST] <mavi> just noticed the update c14
[02:26:17 CEST] <mavi> let me try it now
[02:26:48 CEST] <mavi> I was using the escapes due to using it in a php script, it was needed.
[02:27:47 CEST] <mavi> so I tried the  ffmpeg -i test720.mp4 -filter_complex "hqdn3d=1.5:1.5:6:6,scale=w=if(gt(dar\,854/480)\,min(854\,iw*sar)\,2*trunc(iw*sar*oh/ih/2)):h=if(gt(dar\,854/480)\,2*trunc(ih*ow/iw/sar/2)\,min(480\,ih)):flags=bicubic,setsar=1[we];movie=actorimage.png[wm];[we][wm]overlay=30:main_h-overlay_h-120,subtitles=add.ass[out]" -c:v libx264 -crf 18 -preset veryslow -tune film -refs 8 -bf 6 -aq-mode 2  -c:a copy -f mp4 720encoded.mp4
[02:28:00 CEST] <mavi> and im getting this error. "Filter subtitles has an unconnected output" any idea why?
[02:28:36 CEST] <mavi> found this resource on it but trying to figure it out: https://superuser.com/questions/1088757/filter-boxblur-has-unconnected-output
[02:40:18 CEST] <c_14> you need to map [out]
[02:40:37 CEST] <c_14> just add -map '[out]' somewhere before 720encoded.mp4
[02:47:45 CEST] <mavi> perfect that did the trick, thanks!
[02:48:05 CEST] <mavi> I thought the [out] part of it after the add.ass was enough.
[02:57:01 CEST] <c_14> oh, if you add -map [out] you'll also need to map the audio track
[02:57:11 CEST] <c_14> since adding a manual mapping gets rid of the automatic ones
[02:57:39 CEST] <c_14> But yeah, when you have a filtergraph you need to add the mapping manually. Because it doesn't know which it should use
[02:57:57 CEST] <c_14> though I think you could possible remove the [out] and it would create an automatic one and use that
[03:13:00 CEST] <mavi> ill play with it when i get back from work, thanks again for the help c_14. Was stuck for a few hours on it.
[07:28:54 CEST] <thebombzen> #tfw you find a 16G uncompressed y4m file on your computer
[09:19:30 CEST] <mavi> ffmpeg -i test720.mp4 -filter_complex "hqdn3d=1.5:1.5:6:6,scale=w=if(gt(dar\,854/480)\,min(854\,iw*sar)\,2*trunc(iw*sar*oh/ih/2)):h=if(gt(dar\,854/480)\,2*trunc(ih*ow/iw/sar/2)\,min(480\,ih)):flags=bicubic,setsar=1[we];movie=actorimage.png[wm];[we][wm]overlay=30:main_h-overlay_h-120,subtitles=add.ass"  -c:v libx264 -crf 18 -preset veryslow -tune film -refs 8 -bf 6 -aq-mode 2  -c:a copy -f mp4 720.mp4  works with auto mapping for audio/
[09:23:08 CEST] <kerio> does anyone know off the top of their heads what i have to install on raspbian to compile ffmpeg with mmal?
[09:23:59 CEST] <tdr> the configure script should let you know what its missing
[09:24:09 CEST] <kerio> "mmal"
[09:24:25 CEST] <zap0> so you don't want someone to provide a link to it.. has to be off the top of my head?
[09:24:33 CEST] <kerio> zap0: yes, otherwise it's cheating
[09:24:40 CEST] <kerio> anyway i'm trying with libomxil-bellagio-dev
[09:24:41 CEST] <zap0> ;)
[09:25:07 CEST] <kerio> but that seems more related to the omx part rather than the mmal part
[09:25:08 CEST] <kerio> so idk
[09:25:54 CEST] <kerio> nope, ERROR: mmal not found
[09:27:03 CEST] <tdr> ./configure --enable-mmal --enable-omx-rpi --enable-omx  ?
[09:27:41 CEST] <kerio> yes
[09:27:50 CEST] <kerio> this is not raspbian proper, this is OSMC
[09:27:53 CEST] <kerio> so maybe it's not preinstalled
[09:32:28 CEST] <kerio> fatal error: interface/mmal/mmal.h: No such file or directory
[09:34:11 CEST] <kerio> oh sonuvabitch
[09:34:12 CEST] <kerio> it was rbp-userland-dev-osmc
[09:41:01 CEST] <nichego> hi. something strange happened just now. -to switch stopped working when -ss is used in fast input seeking mode
[09:41:42 CEST] <nichego> i used this command successfully before ffmpeg -ss <start-time> -i input.s -to <end-time> -c copy -copyts clip.ts
[09:42:07 CEST] <nichego> now it copies out 0 frames to clip.ts. i have tested with multiple different inputs, including inputs that worked perfectly fine before
[09:42:21 CEST] <nichego> i have not updated or otherwise made any change to ffmpeg
[11:54:54 CEST] <kerio> how do i stream a thing between two ffmpegs
[11:55:20 CEST] <kerio> more specifically, how do i stream mjpeg
[11:56:05 CEST] <kerio> is mjpeg in mpegts not a thing
[11:56:51 CEST] <kerio> oh i can just stream the raw mjpeg
[11:56:52 CEST] <kerio> ;o
[11:58:23 CEST] <furq> you can just stream the raw mjpeg
[11:58:24 CEST] <furq> hth
[12:00:18 CEST] <furq> also rpi-update should install all the videocore stuff in /opt/vc/
[12:00:31 CEST] <furq> and at least for h264_omx ffmpeg will add that to the include path by default
[12:01:51 CEST] <kerio> furq: how do i tell mpv to play frames as soon as they're available
[12:02:01 CEST] <furq> shrug
[12:02:01 CEST] <kerio> oh i guess that's just what's happening
[12:02:41 CEST] <kerio> furq: the eyetoy is pretty shit
[12:02:47 CEST] <kerio> but not as shit as you made me think
[12:02:57 CEST] <kerio> or possibly someone else in here
[12:03:18 CEST] <JEEB> lol eyetoy. I remember playing around with the PS2 thing
[12:03:33 CEST] <furq> the only thing i said about it was "it's only 240p"
[12:03:35 CEST] <kerio> mjpeg doesn't fit in mpegts ._.
[12:03:39 CEST] <kerio> furq: yea but it's 480p
[12:03:46 CEST] <furq> oh
[12:03:47 CEST] <kerio> so CLEARLY
[12:03:48 CEST] <kerio> you were wrong
[12:03:49 CEST] <furq> did they make more than one
[12:03:51 CEST] <JEEB> 480p with very low rate IIRC
[12:03:57 CEST] <kerio> ye it's 7.5fps
[12:04:25 CEST] <furq> is it any good
[12:04:34 CEST] <JEEB> on a modern level, no
[12:04:36 CEST] <kerio> i mean i can see my face
[12:04:37 CEST] <furq> the caps i saw looked surprisingly ok
[12:04:43 CEST] <furq> by which i mean not noisy but still blurry as fuck
[12:04:45 CEST] <kerio> 240p at 15fps apparently
[12:05:07 CEST] <furq> i've had worse cameras anyway
[12:05:19 CEST] <kerio> and i'm kind of a low light situation here
[12:05:22 CEST] <kerio> so props
[12:05:31 CEST] <JEEB> 720p30 is attainable with a 20-30 eur logitech so the eyetoy only makes any sense if you already have it
[12:05:49 CEST] <furq> i assume this isn't for anything serious
[12:06:00 CEST] <kerio> ofc not
[12:07:14 CEST] <kerio> does it make sense to set the framerate if the camera output is mjpeg?
[12:07:38 CEST] <kerio> "The driver does not permit changing the time per frame" o well
[12:08:53 CEST] <kerio> 240p15 or 480p7.5
[12:08:56 CEST] <kerio> quality
[12:15:28 CEST] <kerio> furq: how do i losslessly save a 1fps mjpeg stream from this
[12:15:30 CEST] <kerio> with ffmpeg
[12:16:19 CEST] <kerio> ok apparently -framerate 1 did it i guess?
[12:16:38 CEST] <kerio> nope
[12:29:42 CEST] <TommyC> Hello, I'm wondering what the option to explicitly build with jack-audio-connection-kit is? `configure --help' doesn't seeem to list any thing when searching for the term "jack"
[12:30:17 CEST] <TommyC> (and there's no configure.ac or configure.in to look at as well)
[12:32:05 CEST] <furq> TommyC: https://github.com/FFmpeg/FFmpeg/blob/master/configure#L6007-L6008
[12:33:07 CEST] <furq> anything listed as autodetect in configure --help will just silently fail if you explicitly enable it and the check fails
[12:33:25 CEST] <furq> or not silently, it won't show up in the configure output of enabled options, but it won't throw an error
[12:44:34 CEST] <kerio> 3 hours later, pi0 still compiling
[12:45:29 CEST] <kerio> it's at libavcodec/v* tho
[12:46:34 CEST] <kerio> i should've thinned this out
[12:46:44 CEST] <kerio> i will literally never need vp9 software decoding
[12:50:54 CEST] <furq> just cross compile it
[12:51:08 CEST] <kerio> that would involve running linux on a real computer
[12:51:17 CEST] <furq> no it wouldn't
[12:51:21 CEST] <furq> it would involve running linux in a vm
[12:51:27 CEST] <kerio> that might be slightly more acceptable
[12:55:45 CEST] <kerio> wavpack hype
[12:59:50 CEST] <kerio> xan :o
[13:01:43 CEST] <kerio> i thought it was done but libswresample ._.
[14:53:01 CEST] <TommyC> furq: Tahnk you (sorry for the extremely late reponse)
[17:58:24 CEST] <StickyNipples> hello, i need help concatinating some flac files
[17:59:05 CEST] <StickyNipples> ....anyone??
[18:28:44 CEST] <durandal_1707> StickyNipples: use concat filter or demuxer
[19:26:26 CEST] <djk> For those interested I discovered a failing switch that was cause a problem for the streaming. This is the current seeming to be working well version for streaming the screen to Facebook. I did switch from Hawkeye to mjpg-streamer which seems be smoother.
[19:26:26 CEST] <djk> nohup ffmpeg  -thread_queue_size 2048 -probesize 32M -framerate 30 -r 30 -f x11grab -s 1280x720  -i :0.0+1280,5 -f lavfi -i anullsrc -c:v libx264 -pix_fmt yuv420p -g 60 -c:a aac -ar 44100 -b:a 128k -preset veryfast -tune zerolatency -maxrate 4000k -r 30 -f flv "rtmp://rtmp-api.facebook.com:80/rtmp/$KEY" &
[19:35:06 CEST] <dystopia_> what are you loading as an input after scaling to 720p?
[19:35:18 CEST] <dystopia_> also don't you need to define a scaler
[19:37:55 CEST] <furq> djk: don't use -tune zerolatency
[19:38:13 CEST] <djk> no?
[19:38:16 CEST] <furq> no
[19:38:29 CEST] <furq> and you can get rid of both -r 30
[19:41:49 CEST] <djk> so this would be better?
[19:41:49 CEST] <djk> nohup ffmpeg  -thread_queue_size 2048 -probesize 32M -framerate 30 -f x11grab -s 1280x720  -i :0.0+1280,5 -f lavfi -i anullsrc -c:v libx264 -pix_fmt yuv420p -g 60 -c:a aac -ar 44100 -b:a 128k -preset veryfast -maxrate 4000k -f flv "rtmp://rtmp-api.facebook.com:80/rtmp/$KEY" &
[19:43:21 CEST] <furq> i mean it'd still be much better to capture on the pi instead of this
[19:43:22 CEST] <furq> but sure
[19:44:46 CEST] <djk> yes yes I realize that but I didn't have luck with the compile yesterday. It is on the list to keep trying but time is limited and there are some personal task that need to get done too ;-)
[19:53:48 CEST] <djk> no good with those changes repeating
[19:53:48 CEST] <djk> Past duration 0.947334 too large
[19:55:51 CEST] <furq> you can normally ignore that
[19:56:24 CEST] <djk> if i put the -r 30 back it goes away
[19:56:42 CEST] <furq> both of those -r 30s should be noops
[19:56:58 CEST] <djk> noops?
[19:57:20 CEST] <furq> no operation
[19:57:23 CEST] <furq> they don't do anything
[19:57:31 CEST] <furq> you're already setting -framerate 30 on the input
[19:57:56 CEST] <furq> i'm pretty sure -r is mapped to -framerate for x11grab, so those are identical
[19:58:05 CEST] <furq> and then setting it on the output doesn't do anything because it's the same as the input framerate
[19:58:35 CEST] <djk> odd that they get rid of the repeating past duration message
[19:58:45 CEST] <furq> shrug
[19:58:50 CEST] <furq> like i said, you can normally ignore that message
[19:59:16 CEST] <furq> but yeah definitely don't use -tune zerolatency
[19:59:52 CEST] <furq> that's specifically for if you need less than 100ms of latency, and if you're going over rtmp to a host on the internet, that's not achievable anyway
[20:00:04 CEST] <furq> and it really badly fucks up the video quality because it turns loads of stuff off
[20:00:11 CEST] <djk> ok
[20:01:42 CEST] <Croolman> Hi everyone,
[20:02:45 CEST] <Croolman> found my way here after desperate attempts to look for an answer. If I am creating a filter for the FFMPEG how can I add padding or transpose the whole plane that is being processed?
[20:03:40 CEST] <djk> i'll try it with the -r 30 a bit noisy on the log but we'll see how it goes. Facebook keeps expiring it even though the stream does look to fail and of course they don't provide and error
[20:03:49 CEST] <furq> add -v error if you want to shut the warnings up
[20:04:53 CEST] <durandal_1707> Croolman: for padding see vf_pad and for transpose vf_transpose
[20:09:04 CEST] <djk> that is just warnings the errors/failures/bad thing will still show?
[20:10:22 CEST] <Croolman> durandal_1707: Thanks for the advice. I was hoping for a function from libavutil or libswscale that woudl do such a thing
[20:14:13 CEST] <furq> djk: yes
[20:16:06 CEST] <djk> furq: Thank you. fingers crossed
[00:00:00 CEST] --- Sun Apr 16 2017