[Ffmpeg-devel-irc] ffmpeg.log.20151120

[burek]

[00:51:54 CET] <jafacakes2011> Hey, anyone free to help?
[00:52:31 CET] <jafacakes2011> im having an issue with .ogg
[00:56:30 CET] <jafacakes2011> ok, thanks :D
[00:59:51 CET] <jafacakes2011> once i have done that can i ask for help here with the patsie link?
[01:03:11 CET] <jasom> jafacakes2011: yes, and in general on IRC, don't ask toask, just ask the question.
[01:03:28 CET] <jafacakes2011> I am trying to stream audio only via ogg using ffmpeg to ffserver. here are the ffserver errors: http://www.pastie.org/10568648
[01:04:12 CET] <jafacakes2011> it is telling me no extradata present, and i cant find any info anywhere on how to fix this.
[01:13:56 CET] <llogan> oh, ffserver. you might be on your own for that.
[01:15:35 CET] <jafacakes2011> is there anywhere i can get help with it or not at all?
[01:16:46 CET] <furq> jafacakes2011: you might be better off using icecast
[01:17:08 CET] <jafacakes2011> cool, will take a look thanks
[01:17:48 CET] <Flerb_> Hi. I have an issue, when I run ffserver, telling it to launch ffmpeg pointing at an MJPEG server of mine, it says it exits with error code 256. I've checked to see what ffmpeg does when launched directly, and it does nothing. I set the verbosity level to debug and I still see nothing from it
[01:18:38 CET] <llogan> jafacakes2011: the wiki may have some info too: https://trac.ffmpeg.org/wiki/ffserver
[01:19:11 CET] <Flerb_> And incidentally, I can run this from VLC
[01:19:15 CET] <Flerb_> *open
[01:28:08 CET] <pepee> is there a way to speed up vp8 encoding?  I want to mix an image with an audio file as a video, but vp8 is really slow. I run ffmpeg like this:   ffmpeg -loop 1 -i img.jpg -i audio.ogg -c:v libvpx -c:a libvorbis -shortest out.webm
[01:30:04 CET] <Flerb_> this is such a pain - all I wanted to do was to convert an MJPEG stream to an MPG stream
[01:49:19 CET] <llogan> pepee: maybe try the -cpu-used libvpx-vp8 option
[01:52:23 CET] <llogan> you can also use a lower frame rate by using -framerate as an input option
[01:53:47 CET] <pepee> llogan, what would be a "good enough" framerate for this?
[01:54:20 CET] <llogan> not sure. as low as you can go until the plyer doesn't support it.
[01:56:08 CET] <pepee> llogan, thanks, I'm trying that now
[02:01:16 CET] <llogan> and libvpx-vp9 will likely give you a significantly smaller file size, but will be much slower
[02:22:08 CET] <waressearcher2> schnee, zu viel schnee dort hin
[02:26:58 CET] <pepee> llogan, so, the trade-off is between speed and file size, right?
[02:29:14 CET] <furq> pepee: you probably want to use -threads with vp8
[02:29:41 CET] <furq> also do you have a good reason to use vp8 over x264
[02:30:45 CET] <pepee> some sites only support webm, and not even vp9, but vp8
[02:59:24 CET] <pepee> I wish I understood how video formats work... how come you can't just concatenate an image and the audio, and tell the video player to show that image in every frame while it's reproducing the audio?
[03:00:32 CET] <pepee> how come there aren't specific solutions for this specific problem: making a video based on a static image and a audio stream? or, can you do this? are there solutions?
[03:02:10 CET] <pepee> in this case, video encoding seems like a huge waste of resources
[03:04:06 CET] <llogan> pepee: -cpu-used is a "quality/speed ratio modifier"
[03:04:52 CET] <llogan> some players for web playback allow you to display an image with an audio file
[03:05:08 CET] <llogan> instead of making an "image video"
[03:06:05 CET] <pepee> perhaps that's what I should do in my website, I hadn't thought of that. the idea was to let users export the videos, though
[03:07:16 CET] <llogan> perhaps you can just add the image as album art to the audio file (container and player support dependent)
[03:07:36 CET] <llogan> ...for the downloaders, not necessarily for the web player
[03:10:33 CET] <pepee> but I'm still wondering, no one thought about having some specific solution for this in the video formats?
[03:12:24 CET] <llogan> album art metadata provided that functionality in most sane audio players. and if you use a low enough frame rate with a sane encoder you might get what you want
[03:12:58 CET] Action: llogan will return tomorrow, probably
[03:13:33 CET] <pepee> thanks llogan
[03:35:44 CET] <prelude2004c> hey everyone good evening
[03:36:01 CET] <prelude2004c> can i get some help with deinterlacing... i am totally clueless  what is going on
[03:36:18 CET] <prelude2004c> so when i don't use yadif i get huge picture jitter and just bad framerate at 30fps
[03:36:37 CET] <prelude2004c> the source says its 29.97 frames .. but somehow the system is getting the frames that are interlaced
[03:37:17 CET] <prelude2004c> i used yadif=1 and it fixes it but... i see frames dropped  so the output logs is saying that frames are going at 30fps but the drops are also 30fps .. weird right ? how can it drop 30fps of a source that only has 30fps ?
[03:37:22 CET] <prelude2004c> can anyone help me to understand this
[03:44:15 CET] <pepee> err, nm my silly rant, I'm stupid :P
[03:48:12 CET] <prelude2004c> ehh yadif=2 doesn't double frames
[03:56:02 CET] <jasom> prelude2004c: even numbered yadif will double the framerat odd numbered modes won't
[03:56:35 CET] <prelude2004c> i think oposite.
[03:56:38 CET] <prelude2004c> 1 , 3 will double
[03:56:43 CET] <prelude2004c> even wont double
[03:56:46 CET] <jasom> prelude2004c: sorry, yes
[03:56:48 CET] <prelude2004c> currently set at 2 and i don't get the double :)
[03:56:56 CET] <prelude2004c> so i learned somethign today :P
[03:57:10 CET] <jasom> 30fps interlaced is 60 fields per second
[03:57:12 CET] <prelude2004c> smooth playback now
[03:57:30 CET] <jasom> so a deinterlacer has the option of generating 60fps to preserve the temporal resolution of the original
[03:58:05 CET] <prelude2004c> only problem now is the damn cpu power usage using the vf :)
[03:58:56 CET] <jasom> prelude2004c: if you are okay halving the vertical resolution you can just drop an entire field; that's super fast
[03:59:53 CET] <jasom> -vf field
[04:03:01 CET] <jasom> and -fv kerndeint will be faster than yadif but slower than -vf field
[04:07:21 CET] <prelude2004c> really .. cool let me try it
[04:07:29 CET] <prelude2004c> do you know why i get errors somethings like  [mpeg2video @ 0x22e7a60] invalid mb type in I Frame at 0 25
[04:07:51 CET] <prelude2004c> a bunch come though teh system. and i get pixelation on the tv... its like the cpu can't keep up or somethign and it starts to fall behind or drop thing :(
[04:07:54 CET] <prelude2004c> i dont know how to fix it
[04:08:21 CET] <jasom> Is this from a broadcast source?  It's possible there were errors in the transmission
[04:09:08 CET] <prelude2004c> wow.. field uses like 1/2 the power of the other
[04:09:13 CET] <prelude2004c> omg.. that could be it : )
[04:11:22 CET] <prelude2004c> ya when the cpu gets close to 200% i start to get those drops
[04:11:32 CET] <prelude2004c> but now with field i get 60% only :)
[04:12:49 CET] <jasom> yadif isn't the fastest thing in the world *and* it's single-threaded only... Now I wonder if I could recode yadif to be multithreaded...
[04:14:33 CET] <prelude2004c> jasom, you can do it :)
[04:14:45 CET] <prelude2004c> i am sooo much happier now with this.. freed up my cpu by laf
[04:14:46 CET] <prelude2004c> half*
[04:14:55 CET] <prelude2004c> and now i don't get the drops right now .. which is very very good
[04:17:02 CET] <prelude2004c> damn.. i enabled refs 2 and now the cpu errors are back
[04:23:47 CET] <prelude2004c> does refs improve quality ?
[04:35:07 CET] <prelude2004c> hey is it normal that the tv now has black bars at the top and bottom ? after i used field
[04:35:20 CET] <prelude2004c> is ther some way to make sure it's full screen
[04:44:41 CET] <jasom> prelude2004c: add a scale filter to get it back to the original size
[04:47:15 CET] <jasom> I think -vf field,h=2*ih will do it
[04:47:55 CET] <Reelee> Can someone let me know if this bug (https://trac.ffmpeg.org/ticket/4200) was included in FFmpeg 2.8.1? If not, any idea when it's expected to be released?
[04:56:47 CET] <prelude2004c> h is giving me an error saying no filter
[04:58:42 CET] <prelude2004c> help ? :)
[05:07:17 CET] <jasom> oh
[05:07:27 CET] <pepee> can you use "vframes 1" to make a 1 frame video of any length/duration?
[05:07:42 CET] <pepee> err, -vframes 1
[05:07:49 CET] <prelude2004c> jasom ? any idea ? my stuff is letterbox now :)
[05:07:52 CET] <jasom> prelude2004c: -vf field,scale=h=2*ih will do it
[05:08:08 CET] <jasom> I forgot to include the filter name :)
[05:08:17 CET] <prelude2004c> tried that but let me try again
[05:09:02 CET] <jasom> you may need quotes, but probably not
[05:09:53 CET] <prelude2004c> hum.. it took it but still looks the same
[05:10:03 CET] <prelude2004c> box is at top and bottom
[05:10:09 CET] <prelude2004c> looks like SD content
[05:10:24 CET] <prelude2004c> or not like sd. .but like a movie wide screen but squished
[05:10:32 CET] <prelude2004c> its just the hight
[05:15:19 CET] <jasom> huh
[05:19:41 CET] <jasom> maybe try -vf field,scale=iw:2*ih
[05:25:34 CET] <prelude2004c> did the sar, dar settings
[05:25:54 CET] <prelude2004c> that seems to do it.... setsar=1:1,setdar=16:9 seems to have done that
[05:26:00 CET] <prelude2004c> does that seem right ?
[05:27:50 CET] <prelude2004c> seems on motion that hte interlace is back though .. is that possible ?
[05:31:14 CET] <jasom> prelude2004c: it shouldn't if you have -vf field
[05:31:17 CET] <jasom> anyway I gotta go
[05:32:28 CET] <prelude2004c> hey thnk you soooooo much for your help
[05:32:35 CET] <prelude2004c> you have helped me A LOT !!!
[07:22:31 CET] <visiot> is quick sync support addded into ffmpeg. i read , it was added in ffmpeg-2.8, but i dont found a way to compile its support into ffmpeg
[07:29:05 CET] <waressearcher2> visiot: hallo
[07:29:32 CET] <visiot> waressearcher2: hey
[07:29:55 CET] <waressearcher2> visiot: wie geht's ?
[07:31:10 CET] <visiot> waressearcher2: gut
[07:31:53 CET] <waressearcher2> wie steht's ?
[07:32:28 CET] <visiot> waressearcher2: kann Ihnen helfen, mw
[07:34:12 CET] <waressearcher2> was ist "mw" ?
[08:08:09 CET] <meadhikari> guys, I need to add a little water mark text in the lower left side of the video, can anyone help me with the command please
[08:14:17 CET] <waressearcher2> meadhikari: hallo
[08:14:24 CET] <meadhikari> Hi
[08:14:40 CET] <waressearcher2> meadhikari: wie geht's ?
[08:15:00 CET] <meadhikari> I wanted to draw text in a video
[08:15:11 CET] <meadhikari> and show it at the bottom left corner
[08:46:44 CET] <hayden> helo ? anyone up for a couple of questions ?
[09:08:34 CET] <waressearcher2> hayden: hallo, hast du irgendwelche Probleme, Schwierigkeiten, Fragen oder Aufgaben ob groß oder klein, bei denen wir dir behilflich sein könnte ?
[09:09:31 CET] <durandal_1707> hayden: ask, do not ask to ask.
[09:11:02 CET] <hayden> hi
[09:11:15 CET] <hayden> i know that, more just checking if anyone here
[09:11:16 CET] <meadhikari> Thanks guys I figure it out
[09:13:16 CET] <hayden> in the ffmpg docs when it says an arg : broadcast=1|0 does it mean the first option is default ?
[10:15:17 CET] <ubitux> durandal_1707 ^
[12:48:39 CET] <aquarat> Is there a simple method of using ffmpeg on it's on to stream raw h.264 ?
[12:48:44 CET] <aquarat> without using ffserver
[12:49:07 CET] <aquarat> *own
[12:49:12 CET] <fruitbag> Hi.
[12:50:12 CET] <fruitbag> I have recovered MXF files by signature from a CF card with a damaged filesystem. Some of them are unsound.
[12:50:17 CET] <BtbN> You don't want to stream raw h264.
[12:50:56 CET] <aquarat> ok
[12:51:02 CET] <aquarat> what would you suggest ?
[12:51:03 CET] <fruitbag> I wan't to either repair them, unwrap them or convert them.
[12:51:37 CET] <aquarat> I have an HDMI to ethernet device which outputs mjpeg... I'm currently converting that to a stdin input into ffmpeg
[12:51:50 CET] <aquarat> and I want a way of getting the output, over a wifi network to a few tablets
[12:51:53 CET] <BtbN> h264 in some container.
[12:51:59 CET] <aquarat> ok
[12:52:31 CET] <aquarat> the issue so far hasn't been the container or the video codec, it's been the transport between ffmpeg and the client devices/consumers
[12:52:42 CET] <aquarat> ffserver seems to want to re-encode everything
[12:53:07 CET] <BtbN> Well, converting mjpeg to h264 without re-encoding will be interesting.
[12:53:30 CET] <aquarat> ffmpeg grabs the mjpeg, encodes it to h.264 and then hands it off to ffserver
[12:53:36 CET] <aquarat> which re-encodes it again
[12:53:45 CET] <aquarat> it doesn't seem to support "copy"
[12:54:38 CET] <BtbN> Hand over mjpeg then, and let ffserver handle the transcoding?
[12:54:57 CET] <aquarat> mmm
[12:55:06 CET] <aquarat> seems reasonable
[12:55:08 CET] Action: aquarat tries
[13:01:37 CET] <aquarat> that seems to be working... somewhat
[13:02:00 CET] <aquarat> just need to tune for low latency and multi-threading, if applicable
[13:14:30 CET] <aquarat> latency/delay is still 7 seconds
[13:14:31 CET] <aquarat> ergh
[14:10:10 CET] <aquarat> can't get this stream latency below 3 seconds :/
[15:09:03 CET] <bove> I'm using the following parameters to change frame rate and stretch the video: -r 25 -vf 'setpts=N/(25*TB)' What are my options to sync the audio?
[15:09:44 CET] <bove> Either by resampling or changing the samplerate the same way as the framerate
[15:11:39 CET] <fritsch> what's the difference between: resampling and "changing the samplerate"?
[15:15:30 CET] <bove> Let's say my original video is 10seconds in 50fps 48000Hz. The output will be 20seconds in 25fps with either 24000Hz (not resampled) or 48000Hz (resampled)
[15:17:17 CET] <fritsch> what i wanted to say: both needs to be resampled
[16:03:11 CET] <pgunnars> what would cause the ?date functionality in ffserver to stop working
[16:04:23 CET] <pgunnars> or more precisely
[16:04:29 CET] <pgunnars> no matter the date, go to the beginning
[16:32:05 CET] <pgunnars> hmm
[16:32:26 CET] <pgunnars> when using ?date with stored stream from ffserver, how come -c copy takes almost as long as transcoding it?
[16:44:28 CET] <Zeranoe> NVENC is currently the only GPU encoder FFmpeg supports correct? OpenCL only does filtering.
[16:45:34 CET] <c_14> There's QSV, but that's not GPU
[16:45:40 CET] <c_14> Other than that, none I know of.
[17:07:30 CET] <DHE> x264 has basic opencl offload but it's hit and miss as to whether it helps or not
[17:08:00 CET] <DHE> sometimes using opencl will slow it down
[17:08:04 CET] <kepstin> from what i've read from the devs, it's slower and lower quality than the pure cpu encode in most situations
[17:08:52 CET] <kepstin> NVENC isn't really a gpu encode either; it uses a dedicated hardware encoder block on the gpu chip rather than compute resources.
[17:09:13 CET] <DHE> I've seen it give a 10% improvement for high resolution encodes (1080p) with preset=medium or so
[17:09:21 CET] <DHE> can't tell you about the quality though
[17:09:32 CET] <furq> it gives a bigger improvement for sd encodes iirc
[17:09:37 CET] <DHE> but at the end of the day that's barely worth it
[17:09:40 CET] <furq> speed improvement, that is
[17:09:56 CET] <furq> apparently the quality is slightly worse but not enough for me to notice
[17:10:53 CET] <DHE> meh, nvenc is good enough usually
[17:12:39 CET] <kepstin> (QSV is also only available on intel chips that have the gpu, since the encoder block is on the gpu side of the chip; that excludes some xeons iirc)
[17:12:50 CET] <furq> did opencl bicubic scaling make it into ffmpeg yet
[17:23:53 CET] <fruitbag> Hello, everyone.
[21:02:00 CET] <votz> The left and right audio channels get reversed when wav output is piped to various other programs. Strangely, this only happens when piped to some programs (e.g. sox) and not others (e.g. lame). What could cause this?
[21:02:37 CET] <votz> This 'cat audio.aac | ffmpeg -i - -vn -sn -acodec pcm_s16le -ac 2 -r 44100 -f wav - | sox -t raw -e signed-integer -b 16 -c 2 -r 44100 - sox.wav' will cause sox.wav to have the left and right audio channels backwards.
[21:04:08 CET] <votz> However, the resultant lame.mp3 of 'cat audio.aac | ffmpeg -i - -vn -sn -acodec pcm_s16le -ac 2 -r 44100 -map_metadata -1 -f wav  -t 265.0 - | lame - lame.mp3' has the audio channels correctly placed.
[21:09:58 CET] <Kota> Question: Is there a way to remove metadata (specifically "Title") from .mp4 containers without generating a brand new file?
[21:10:04 CET] <Kota> Basically, can I do it in place?
[21:10:21 CET] <furq> not with ffmpeg
[21:10:40 CET] <Kota> hmmm....
[21:11:19 CET] <Kota> I've been having plex trouble and found out it's reading the metadata in the files to name the episodes, and the metadata is useless "series_disk_1_episode_2,mov" or similar
[21:11:28 CET] <Kota> and I saw that ffmpeg can deal with metadata
[21:11:47 CET] <Kota> Damn, was hoping to be able to script this quick
[21:12:15 CET] <furq> you can do it with mp4box
[21:12:34 CET] <furq> i imagine there is some cli tool which can do it but i don't know of one
[21:44:56 CET] <Flerb> So if, with ffserver, I have a stream set up with the name stream1.mpg, and the port set to 8090, any requests to it should display the stream?
[21:45:13 CET] <Flerb> so http://localhost:8090/stream1.mpg should be it?
[23:03:18 CET] <ChocolateArmpits> Hello, I'm rewrapping a ts file to hls segments, but all video frames before the first video frame get discarded. Any way to preserve them?
[23:03:32 CET] <ChocolateArmpits> I mean frames before the first keyframe
[23:03:59 CET] <JEEB> why on earth would you want them?
[23:04:06 CET] <JEEB> I mean, they can't be decoded correctly
[23:04:21 CET] <fritsch> and kill some hw decoders
[23:04:24 CET] <JEEB> although huh... I'm not sure if that's how it works
[23:04:26 CET] <ChocolateArmpits> The reference frames are located at the end of another file
[23:04:37 CET] <JEEB> then just provide that as the input before?
[23:04:45 CET] <JEEB> there's N ways to concat shit
[23:05:30 CET] <ChocolateArmpits> There are problems with timestamps, ffmpeg/ffplay doesn't playback such concatenated files nicely
[23:05:58 CET] <ChocolateArmpits> The first file has the timestamps in the 90000+ range, while the second has them in 1000+
[23:06:14 CET] <JEEB> make sure you try out all the ways to concat, and try other means of remuxing as well
[23:06:17 CET] <JEEB> have fun
[23:06:36 CET] <ChocolateArmpits> Well other means are not feasable, and this is not fun :(
[23:09:26 CET] <TD-Linux> I thought ts wasn't supposed to have absolute timecodes. how did they screw this up?
[23:11:37 CET] <TD-Linux> err right it does, but it also has a flag for resetting it
[23:12:19 CET] <ChocolateArmpits> TD-Linux: Please give more clues if possible
[23:15:30 CET] <ChocolateArmpits> there's a 'showall' decoder flag that passes frames before the first keyframe. Obviously it can't work with a stream that's being copied
[23:31:48 CET] <TD-Linux> ChocolateArmpits, this is pretty much outside of my expertise. But you're going to have to either join the streams into one continuous one or reencode the first bit before the keyframe
[23:51:36 CET] <votz> How can the RIFF/WAV metadata chunks (LIST and INFO) be disabled in output?
[23:56:24 CET] <waressearcher2> votz: hallo
[23:57:59 CET] <votz> waressearcher2: guten tag
[23:58:44 CET] <waressearcher2> votz: wie geth's wie steht's ?
[00:00:00 CET] --- Sat Nov 21 2015