[Ffmpeg-devel-irc] ffmpeg.log.20180708

[burek]

[01:21:14 CEST] <TheAMM> Given a series of images (or image2pipe), is there any way to specify the timestamp for each frame (other than writing your own NUT muxer)?
[01:21:58 CEST] <TheAMM> As in doing VFR instead of CFR with input -r
[01:23:22 CEST] <atomnuker> no
[01:25:25 CEST] <TheAMM> Oh, actually
[01:25:33 CEST] <TheAMM> I could do some absurd if chains with setpts
[01:26:07 CEST] <atomnuker> yep
[01:28:00 CEST] <TheAMM> Any already-known performance concerns over this?
[01:28:35 CEST] <TheAMM> My use case is a hundred frames at most, but I don't know how the expressions work on the inside
[01:35:14 CEST] <DHE> well, the encoder is likely going to use way more CPU then anything you could feed to the filter pipeline
[01:35:54 CEST] <TheAMM> Another idea I've had is copying timestamps from another input stream
[01:36:21 CEST] <TheAMM> Either directly somehow (haven't found a way) or abusing overlay, but that didn't work last I tried, either
[02:40:41 CEST] <TheAMM> 24 frames (just under a second) means a 519 character expression ala if(eq(N,0),0.000/TB,if(eq(N,1),0.042/TB,if(eq(N,2),0.083/TB, ...
[02:40:52 CEST] <TheAMM> But it does work
[02:41:14 CEST] <TheAMM> And I guess I can force the TB and remove it from the exp
[02:42:13 CEST] <TheAMM> Which gets me down to 395 chars
[02:42:52 CEST] <TheAMM> I'll hit the argument length cap at some point but good enough, good enough
[02:48:29 CEST] <TheAMM> Nevermind, hit the stack limit with 288 frames
[02:48:35 CEST] <TheAMM> sigh
[02:51:12 CEST] <DHE> aww..
[02:51:22 CEST] <DHE> because if it worked you could write a filter to a file
[02:52:29 CEST] <TheAMM> I can?
[02:52:57 CEST] <TheAMM> I work around the stack with eq(N,0)*0+eq(N,1)*42+ ... which works
[02:54:16 CEST] <TheAMM> O, filter_complex_script
[08:22:57 CEST] <yagiza> Hello!
[08:23:52 CEST] <yagiza> Can any1 help me with AVCodecContext
[08:23:56 CEST] <yagiza> ?
[10:26:09 CEST] <Whiskey> C:\Users\Anton\Desktop\Cover\ffmpeg\ffmpeg -y -i $destination\$SourceFile.jpg -q:v 10 $destination\$SourceFile-New.jpg <- this one compress the JPEG by 200KiB, is "-q:v 10" the higest compression method?
[10:31:22 CEST] <furq> Whiskey: iirc 2 is best, 31 is worst
[12:07:47 CEST] <akaWolf> hi, guys
[12:07:58 CEST] <akaWolf> does ffmpeg have option for no trust PCR?
[16:06:44 CEST] <Hello71> also ffmpeg's jpeg compressor is... not great
[16:18:22 CEST] <furq> it's not going to be comparable to mozjpeg or something
[16:18:41 CEST] <furq> but honestly it's sad we're still using jpeg in 2018 anyway
[16:19:43 CEST] <furq> mozilla need to hurry up and support webp already
[16:21:31 CEST] <JEEB> the webp thing is they dislike it being a single-vendor thing. I guess they didn't have such heavy feelings with libvpx or VP8/9
[16:24:58 CEST] <atomnuker> webp is horribly engineered cancer no on wants to support
[16:25:58 CEST] <wyseguy> is ffmpeg cpu intensive?
[16:26:01 CEST] <Hello71> avif?
[16:26:36 CEST] <wyseguy> im going to have a macbook air in my backpack while trying to live stream to youtube and wondering if its gonna over heat
[16:27:03 CEST] <Hello71> did they fix isobmff or is it still more flexible than ipsec
[16:41:37 CEST] <furq> JEEB: yeah exactly
[16:47:00 CEST] <akaWolf> can I fix that by passing some parameters to ffmpeg?
[16:47:01 CEST] <akaWolf> ffmpeg/video: h264: co located POCs unavailable
[17:55:10 CEST] <Whiskey> I got this line working "C:\Users\Anton\Desktop\Cover\ffmpeg\ffmpeg -i $SourceFullName -i $temp\$SourceFile-New.jpg -map_metadata 0 -map 0 -map 1 $DestinationFullName" but why does it covert my JPGs to PNGs?`
[17:58:04 CEST] <furq> does $destinationfullname end in .png
[17:58:21 CEST] <furq> actually nvm this is cover art isn't it
[17:58:35 CEST] <furq> use -c:v copy if you don't want to reencode the cover art
[17:58:43 CEST] <furq> otherwise it'll use the default encoder
[17:58:57 CEST] <furq> i suspect you just want -c copy or else it'll reencode the audio as well
[18:08:38 CEST] <DHE> wyseguy: it can be. defaults are usually a decent quality/speed tradeoff. but there are options to reduce CPU usage (at the expense of image quality)
[18:10:34 CEST] <Whiskey> i mange to make the JPG be a JPG
[18:10:48 CEST] <Whiskey> Problem is that i compress it but its inserted uncompressed
[18:11:47 CEST] <Whiskey> The coide i use now, https://paste.ubuntu.com/p/r3P8WjK6zR/
[18:14:45 CEST] <wyseguy> DHE okay. so with settings I was recommend to try its still a really bad quality... ffmpeg -f mpegts -i "udp://10.5.5.100:8554?fifo_size=10000" -c:v copy -f flv "rtmp://a.rtmp.youtube.com/live2/xxxx-xxxx-xxxx-xxxx
[18:32:30 CEST] <dan9er> Hey. 3D metadata guy again. I submitted my question to the mailing list 2 days ago, but i've still yet to receive a response. Please answer it. My email is dan "nine" erthegamer "at" gmail "dot" com.
[18:33:24 CEST] <dan9er> Gonna AFK in here for a bit.
[18:41:37 CEST] <wyseguy> ill be back
[19:23:22 CEST] <dan9er> Erm, I got disconnected?
[19:24:59 CEST] <atomnuker> dan9er: which patchset is yours?
[19:25:44 CEST] <dan9er> Patchset? No, this is not a bug. I'm asking how to add 3D metadata to a SBS Full Frame MP4
[19:25:53 CEST] <dan9er> atomnuker
[19:27:00 CEST] <dan9er> I already sent details to the mailing list.
[19:27:49 CEST] <atomnuker> oh, the user mailing list
[19:30:10 CEST] <atomnuker> right, just add AV_PKT_DATA_SPHERICAL to all packets leaving the encoder
[19:36:05 CEST] <dan9er> atomnuker No, it's a 3D video. not a 360 video -_-
[19:40:28 CEST] <atomnuker> fine, AV_PKT_DATA_STEREO3D then, just read the bloody docs
[19:40:57 CEST] <dan9er> atomnuker Where are the docs?
[19:41:57 CEST] <dan9er> The video is 648x720p per eye, does that complicate things?
[19:42:55 CEST] <dan9er> And it's SBS Full Frame. NOT SBS Half Frame. Big difference.
[19:46:22 CEST] <atomnuker> dan9er: libavcodec/avcodec.h and libavutil/stereo3d.h
[19:48:30 CEST] <dan9er> atomnuker Dude... this is #ffmpeg, not #ffmpeg-devel. I just need a command flag, not the header file that the flag activates.
[19:58:53 CEST] <dan9er> Hello?
[20:01:16 CEST] <atomnuker> nope, no command line flags
[20:01:25 CEST] <atomnuker> you have to do it yourself
[20:01:57 CEST] <atomnuker> and yes, this is #ffmpeg, a place where people ask for API use as well
[20:07:38 CEST] <dan9er> Why can't ANYONE just give me a fukin command?!
[20:07:51 CEST] Action: dan9er table flip
[20:50:11 CEST] <bray90820> Would there be an easy way to cut something at each chapter
[20:50:15 CEST] <bray90820> It's an MKV And I would like to have teh audio only
[20:54:45 CEST] <alealeale> hello guys, a question -> I have several mp3 files and some marked "regions" that I want to take out of them and concat them together
[20:55:10 CEST] <alealeale> I'm using the concat filter and a list file with all the inpoint and outpoint's
[20:56:26 CEST] <alealeale> the thing is that the output file has an unpredictable offset (the concat filter does not exactly cut/paste on the exact locations specified)
[20:57:22 CEST] <alealeale> in the documentation (for concat inpoint) there is: "This directive works best with intra frame codecs, because for non-intra frame ones you will usually get extra packets before the actual In point and the decoded content will most likely contain frames before In point too."
[20:57:28 CEST] <TheAMM> I haven't used the concat filter or demuxer in ages, but from what I remember it wasn't that nice when taking segments from source files
[20:57:49 CEST] <TheAMM> I recommend grabbing the regions separately and then joining them afterwards
[20:58:12 CEST] <alealeale> how does one do that?
[20:58:26 CEST] <TheAMM> :shrug:
[20:58:37 CEST] <TheAMM> I know how to use mpv for this, because with it segments are easy
[20:59:26 CEST] <TheAMM> mpv --\{ a.mp3 --start=30 --end=35 --\} --\{ b.mp3 --start=10 --end=20 --\} --oac=flac --o joined.flac
[20:59:49 CEST] <TheAMM> Surely someone will guide you with ffmpeg as well
[20:59:57 CEST] <alealeale> let me try (with mpv) and I'll be back, thanks :D
[21:00:00 CEST] <TheAMM> I just have bad luck with -ss and -to
[21:00:39 CEST] <alealeale> yeah, apparently it's "expected" behavior, but I need to find a way to achieve this thing anyway
[21:09:46 CEST] <leonardus> is it possible to take the movement data from a video, and place that onto an image with ffmpeg?
[22:15:10 CEST] <guru_> Could someone please help me compile ffmpeg for Rockchip RK-3328 with hardware encoding capabilities? So far, I have installed the rockchip-mpp library and configured ffmpeg with ./configure --enable-version3 --enable-gpl --enable-libdrm --enable-rkmpp --enable-nonfree but I seem to be missing the hardware encoding capability.
[00:00:00 CEST] --- Mon Jul  9 2018