[Ffmpeg-devel-irc] ffmpeg.log.20170923

Sun Sep 24 03:05:01 EEST 2017

[00:03:41 CEST] <martyrs> guys thanks for the help
[00:04:15 CEST] <martyrs> now i need to process 700+ video files without killing my server cpu1
[00:11:46 CEST] <sabusanta> Outsource to the cloud
[00:14:57 CEST] <sabusanta> alexpigment: you were right about them comments. good tip
[00:14:57 CEST] <martyrs> sabusanta: pricey
[00:15:48 CEST] <alexpigment> sabusanta: glad to hear it :) I've been checking the site every day since the early 2000s - when I actually listened to the music they covered. I get the distinct impression that everyone who comments is in the same boat
[00:15:51 CEST] <Sebas_> limit threads to half your cores?
[00:16:13 CEST] <sabusanta> martyrs: maybe. I guess if you factor in burning your CPU, electricity costs, time, and cooling might even out
[00:16:19 CEST] <sabusanta> or under
[00:16:52 CEST] <sabusanta> alexpigment: Nu-metal will always have a place in my heart tho. my teenage angst heart \
[00:18:18 CEST] <alexpigment> I can relate
[00:18:44 CEST] <alexpigment> they also cover Deftones constantly on that site, which is always good imho
[00:25:31 CEST] <furq> https://www.youtube.com/watch?v=7W6dDqEksKc
[00:25:33 CEST] <furq> what do u think
[00:32:04 CEST] <thebombzen> BtbN: it's 10-bit
[00:33:11 CEST] <thebombzen> I'm not sure if it's a bug in lavc's hevc_cuvid or if it's someone else's fault
[00:33:31 CEST] <thebombzen> FWIW, the sample is here: http://0x0.st/vfK.mp4
[00:53:08 CEST] <lindylex> I am using this https://pastebin.com/RuhupF49 to place to videos side by side.   I want to place a black line on top the video.  This can be a black video dynamically created with the width of 2 pixel and height of 720.  Or it can be a black image.  Any suggestions on how to do this?
[03:20:16 CEST] <AndrewMock> I want to mix 5.1 to stereo WITH LFE. Can't find an ideal way online.
[03:21:20 CEST] <AndrewMock> the stereo will go through a crossover. but i want prevent a bassy mix. can't find a good matrix
[03:35:07 CEST] <kepstin> the trick you have to know is that in most 5.1 systems, the sub channel is actually stored at a different gain from the other channels
[03:35:15 CEST] <kepstin> so you have to adjust that when mixing it
[03:37:20 CEST] <kepstin> although iirc the lfe is usually stored 10dB *quieter* than the other tracks, so mixing it straight in shouldn't normally make stuff be "too bassy"
[03:40:16 CEST] <kepstin> I'd expect simply centre-panning it into a stereo mix shouldn't be unreasonable, plus or minus some gain to taste.
[03:45:22 CEST] <AndrewMock> it'll pass thru a live sound console on its way out. i'll tell the engineer to EQ to taste. thx.
[05:06:16 CEST] <lindylex> I am using this https://pastebin.com/RuhupF49 to place to videos side by side.   I want to place a black line on top the video.  This can be a black video dynamically created with the width of 2 pixel and height of 720.  Or it can be a black image.  Any suggestions on how to do this?
[12:09:17 CEST] <kubast2> Hey how can I read a font style used in softsub? When I cut a video the font style of softsubs seems to change
[12:28:49 CEST] <c_14> kubast2: ass?
[12:28:53 CEST] <c_14> output it and open in a text editor?
[12:31:08 CEST] <kubast2> yeah I see now thx
[13:24:48 CEST] <norbert> hi, I'm transcoding DOSBox AVI to MP4 for usage in kdenlive 720p 60fps, using: ./ffmpeg -r 70.086 -i in.avi -vf "scale=1152:720,pad=1280:720:64:0:black" -b 4000k -ab 160k -r 60 -acodec libvo_aacenc out.mp4
[13:25:42 CEST] <norbert> in the project bin (clips overview) of kdenlive, when I click the MP4 and play it, it plays in its entirety, but when I move the clip onto a track it only plays and shows about 2/3 of the clip
[13:26:34 CEST] <norbert> while this might primarily be a kdenlive issue, I'm wondering if anyone here has a suggestion how I could transcode the clip differently to make it more likely that kdenlive will properly display/play the clip inside one of its tracks?
[13:54:44 CEST] <norbert> nobody?
[14:07:03 CEST] <blap> hi
[14:48:03 CEST] <Sharkigator> i want to encode a video with h.264 and preserve true black
[14:48:19 CEST] <Sharkigator> when i encode it with libx264 it becomes gray
[14:48:38 CEST] <Sharkigator> but with libx264rgb it keeps it
[14:49:06 CEST] <Sharkigator> but is there another way, without using the rgb color format?
[14:53:05 CEST] <c_14> you can try encoding yuv444p
[14:53:18 CEST] <c_14> or using the zscale filter to do the color conversion
[14:54:26 CEST] <Mavrik> hmm
[14:54:36 CEST] <Mavrik> Sounds like HDMI black level issue?
[14:54:47 CEST] <Mavrik> It encodes to 14-255 range?
[15:04:09 CEST] <Sharkigator> yuv444p still makes it gray
[15:05:19 CEST] <utack> are you sure the ENcoding is wrong, not the hardware decoder used for playback? some gpus like to mess things up or do odd processing. have you decoded a frame to something like png again with ffmpeg and looked at it in a image tool
[15:05:41 CEST] <utack> my old amd card had some post processing by default that did this for example
[15:05:53 CEST] <Sharkigator> i'm looking at it with vlc
[15:06:09 CEST] <Sharkigator> but i'll try
[15:06:44 CEST] <utack> probably not the issue wehn other videos in that format look ok, but it is easy enough to exclude as potential problem
[15:07:45 CEST] <Sharkigator> well, the .pngs look fine
[15:10:59 CEST] <Sharkigator> but vlc says hardware acceleration is disabled
[15:11:30 CEST] <Mavrik> Sharkigator, can you pastebin ffmpeg output for the encode?
[15:15:58 CEST] <Sharkigator> Mavrik <https://pastebin.com/Fn38Y8sp>
[15:17:04 CEST] <Mavrik> Sharkigator, can you do -pixfmt yuvj444p ?
[15:17:27 CEST] <Mavrik> I'm 90% sure that yuv444p implies HDMI black levels where "black" is actually "14" color value instead of "0"
[15:18:03 CEST] <Sharkigator> still gray
[15:18:35 CEST] <Mavrik> *shrug* Out of ideas then :)
[15:18:48 CEST] <Sharkigator> but it also says [swscaler @ 00000000026b9fe0] deprecated pixel format used, make sure you did set range correctly
[15:20:03 CEST] <Sharkigator> hm, thank you though
[15:21:43 CEST] <Mavrik> Yeah, I'm still thinking it's a color range isse
[15:21:55 CEST] <Mavrik> But the fact that you have a funny input format means that it can be anything :/
[15:24:41 CEST] <Sharkigator> unrelated: when i try to set the profile option, it always says the profile is not supported
[15:25:02 CEST] <Sharkigator> [libx264 @ 00000000001ec740] Error setting profile main.
[15:25:02 CEST] <Sharkigator> [libx264 @ 00000000001ec740] Possible profiles: baseline main high high10 high422 high444
[15:29:51 CEST] <Mavrik> IIRC you can't have 4:4:4 in main profile
[15:31:11 CEST] <Sharkigator> i don't set the pixel format, shouldn't it choose a correct one?
[15:31:27 CEST] <furq> it'll pick 4:4:4 if the input is rgb
[15:32:27 CEST] <furq> you can try -pix_fmt yuvj420p
[15:32:38 CEST] <furq> that's not an ideal solution but it'll at least tell you if it's a full/limited range issue
[15:32:57 CEST] <Sharkigator> btw, sorry, i just found another checkbox in vlc... "use hardware yuv->rgb conversions"
[15:33:07 CEST] <Sharkigator> not it is actual black...
[15:33:11 CEST] <Sharkigator> *now
[15:33:14 CEST] <furq> oh
[15:33:15 CEST] <furq> fun
[15:34:04 CEST] <Mavrik> hmm, fun :)
[15:36:22 CEST] <furq> on which note, how does one actually do rgb to yuv with zscale
[15:36:32 CEST] <furq> every time i've tried it i get "no path between colorspaces"
[15:37:07 CEST] <furq> i assume i need to set the input primaries and whatnot but idk what those should be for a screen capture
[18:55:58 CEST] <thebombzen> when I use ffmpeg -f x11grab to grab the screen, the Y offset isn't working correctly
[18:56:57 CEST] <thebombzen> if I use: ffmpeg -f x11grab -video_size 1280x720 -i :0.0+320+180, the area it grabs is indeed 1280x720, but it starts at X=320 and Y=0, not X=320 and Y=180
[18:57:12 CEST] <thebombzen> I believe this is a bug
[18:57:16 CEST] <c_14> use , instead of the second +
[18:57:49 CEST] <thebombzen> oh thanks
[18:57:53 CEST] <thebombzen> did that change? :O
[18:57:58 CEST] <c_14> it's always been that way?
[18:58:21 CEST] <thebombzen> weird
[18:58:27 CEST] <thebombzen> I swear I remember doing it the other way /shrug
[18:58:56 CEST] <thebombzen> hm, the docs agree with you. idk wow
[19:11:04 CEST] <rjp421> furq, JEEB, my pi's sdcard went belly up while rebuilding the /opt/vc.... i was not able to test the raspivid piped to ffmpeg-git... hopefully someone else can, i would like to know if it still works
[19:24:13 CEST] <faLUCE>  1) in a mpegts audio+video stream is the PCR necessary to sync audio and video?  2) should the PCR be sent on the first PES of the stream?
[19:24:18 CEST] <faLUCE> (hello)
[19:25:10 CEST] <rjp421> anyone with a pi and pi-cam, and updated packages+kernel+ffmpeg-git:  can you please test piping raspivid to ffprobe or ffmpeg?  confirm whether or not ffmpeg crashes
[19:26:10 CEST] <Mavrik> faLUCE, PCR is pretty much critical for any player yeah
[19:26:19 CEST] <Mavrik> a lot of them sync on it
[19:27:12 CEST] <faLUCE> Mavrik: do the players sync on it when the stream is audio+video or even audio or video only?
[19:27:28 CEST] <Mavrik> It's pretty much the input mux that does it
[19:27:33 CEST] <Mavrik> Which doesn't care about the content yet :)
[19:27:40 CEST] <Mavrik> In other words:
[19:27:51 CEST] <Mavrik> Without valid PCR you might have a really bad time playing back your stream.
[19:28:27 CEST] <faLUCE> Mavrik: but my question was different
[19:31:06 CEST] <faLUCE> [19:27] <faLUCE> Mavrik: do the players sync on it when the stream is audio+video or even audio or video only?
[19:31:49 CEST] <Mavrik> And I told you it doesn't matter what is inside the stream.
[19:31:54 CEST] <Mavrik> Meaning "on all".
[19:34:21 CEST] <faLUCE> Mavrik: I'm not sure you are right. It seems for example that gstreamer uses it as offset for syncing audio and video. But for audio/video only, it uses the first useful frame as offset.
[19:34:57 CEST] <Mavrik> Hence the word *might have a bad time*
[19:35:13 CEST] <Mavrik> I'm sure there are players that will attempt to play broken TS streams anyway.
[19:35:18 CEST] <Mavrik> I guarantee you there are players that will not.
[19:35:40 CEST] <Mavrik> TS stream needs a PCR on container level no matter what's in it.
[19:36:11 CEST] <Mavrik> Without it, your stream might still play (especially in software players)
[19:36:25 CEST] <Mavrik> But certanly not on all players and sync miight be completely off as well
[19:36:47 CEST] <faLUCE> let's ask in another way. If I have a video-only stream, I don't need the PCR for the FIRST frame, in order to START PLAYING the stream. But I need the PCR in order to start playing an audio+video stream, right?
[19:39:00 CEST] <faLUCE> Mavrik: in other words, given what you said:  is it true that the PCR is necessary 1) for syncing audio and video   AND    2) for having good timestamps on all the elementary streams ?
[19:44:57 CEST] <Mavrik> faLUCE, yes and yes :)
[19:45:16 CEST] <Mavrik> and you'll probably need PCR on some players for video-only stream as well
[19:49:27 CEST] <faLUCE> Mavrik: sorry, had connection problem:
[19:49:35 CEST] <faLUCE> [19:36] <faLUCE> let's ask in another way. If I have a video-only stream, I don't need the PCR for the FIRST frame, in order to START PLAYING the stream. But I need the PCR in order to start playing an audio+video stream, right?
[19:49:42 CEST] <faLUCE> [19:39] <faLUCE> Mavrik: in other words, given what you said:  is it true that the PCR is necessary 1) for syncing audio and video   AND    2) for having good timestamps on all the elementary streams ?
[21:59:25 CEST] <dviola> hi
[21:59:47 CEST] <dviola> I want to increase the audio volume of a video (re-encoding the audio) but without re-encoding the video itself, how can I do this?
[22:00:15 CEST] <dviola> more like normalize the audio
[22:02:12 CEST] <Mavrik> pass -codec:v copy
[22:02:16 CEST] <Mavrik> which will just copy video stream
[22:02:24 CEST] <Mavrik> and set audio filters / encoding parameters as usual
[22:03:28 CEST] <dviola> I see, thanks
[22:08:29 CEST] <dviola> -filter:a "volume=20dB"
[22:08:34 CEST] <dviola> would something like that work?
[22:10:04 CEST] <dviola> there's also -filter:a loudnorm -- not sure which one I should use
[22:12:07 CEST] <dviola> "Loudness Normalization (loudnorm): This is recommended for most applications, as it will lead to a more uniform loudness level compared to simple peak-based normalization."
[22:12:10 CEST] <dviola> I guess this is what I need
[23:42:30 CEST] <molecman> hi i am not sure of function of file used in feed section (just a buffer ?) so why not use memory with /dev/shm instead of /tmp ?
[00:00:00 CEST] --- Sun Sep 24 2017