[Ffmpeg-devel-irc] ffmpeg.log.20181208

[burek]

[00:12:32 CET] <jeaye> I've 50 frames of an animation with which I'm building a webm file; I've done it this way many times before, with no issue, but this particular set of files seems to be resulting in a corrupted output. I don't get any errors from ffmpeg and the video plays entirely and then has visual corruption at the end. Any ideas what might be going on?
[00:12:37 CET] <jeaye> CLI usage and output: https://gist.github.com/jeaye/cc4db72103a5a121df01e012e78780cb
[00:20:37 CET] <jeaye> Ok, this is due to the size of the frames. The 30x30 and 80x80 versions I have show the issue, but not the 125x125 version.
[01:50:21 CET] <aicra> hi. this is kind of a "user" question. I have been admiring your DX - your website - and I am curious about what you used... sphinx? jekyll? something else?
[01:52:01 CET] <furq> aicra: https://github.com/FFmpeg/web
[02:16:02 CET] <aicra> furq: thanks!
[11:44:38 CET] <dv_> does anybody here know anything about vp8? I have a hw codec here that I am wrapping. the problem is that I dont get information from it about the frame type. specifically, I don't know if the frame is a golden or alternate reference frame. but these I have to skip, so I need to know if it is an altref/golden frame.
[11:44:57 CET] <dv_> so, I need to check myself if it is a golden or altref frame. but I didn't find out how. the specs I found only speak of the golden/altref frames themselves, not how to identify them.
[13:12:50 CET] <pink_mist> loads of people here know lots about vp8 ... I'm not one of them though. suggest you hand around until one of them shows up
[13:13:39 CET] <pink_mist> *hang
[13:13:45 CET] <pink_mist> though I guess you've already been here for an hour and a half, so I'm presuming hanging around won't be a problem
[13:17:40 CET] <atomnuker> dv_: just parse the field yourself then, its in the packet header
[14:38:07 CET] <dv_> atomnuker: when you say packet, what packet do you mean? webm packet? or vp8 bitstream packet? sorry I'm not familiar at all with how vp8 is encoded
[14:38:37 CET] <atomnuker> vp8 bitstream packets
[14:39:22 CET] <atomnuker> you could use webm packets too I suppose if you trust the muxer has done its job at marking keyframes correctly but I'm not familiar with those
[15:07:57 CET] <dv_> atomnuker: so the part that is relevant for me would be from https://tools.ietf.org/html/rfc6386#section-9.1 the bit that defines whether or not the frame is to be displayed, right?
[15:16:52 CET] <atomnuker> lol no
[15:36:57 CET] <dv_> atomnuker: then where?
[15:37:32 CET] <dv_> because it does sound a lot like it. as said, I need to know whether or not the decoded frame is to be displayed..
[15:37:47 CET] <dv_> altref frames I have to skip instead
[15:43:40 CET] <atomnuker> the key frame flag is what you were asking about
[16:05:19 CET] <p0a> Hello I want to video capture part of my screen
[16:05:26 CET] <p0a> the commands I've seen online capture the whole screen
[16:05:27 CET] <p0a> is it possible?
[16:10:30 CET] <p0a> I think I found some information online
[16:21:42 CET] <jkqxz> dv_:  For reference frame use in the decoder, there are stored frames - "last", "golden" and "altref".  These need not be distinct - they can all contain the same frame.  When a new frame is decoded, it always gets stored in the "last" slot.
[16:22:26 CET] <jkqxz> There are also bits in the frame header to indicate tat it should be stored in the "golden" and "altref" slots (independently, so it can be in both).
[16:22:54 CET] <dv_> that's what I saw when digging through the libvpx code. slots for golden and altref.
[16:23:30 CET] <jkqxz> Further, the "golden" and "altref" slots can be filled by a frame already in one of the other slots, so for example you can copy the previous "last" frame to "golden" , or any other combination.
[16:25:33 CET] <dv_> well, okay. but what I need is to know when to skip / not-display a frame.
[16:25:34 CET] <jkqxz> Since you can always change the type later and rearrange slots, there isn't really any sense in which a single decoded frame is "altref" or "golden" because at any given time it can be none or one or both of those.
[16:26:14 CET] <dv_> do I really have to skip all keyframes ?
[16:26:49 CET] <jkqxz> Sounds like you actually want the show_frame flag, which is completely unrelated to the reference frame storage.
[16:27:41 CET] <dv_> yeah, I guess so. I did conflate "is it an altref frame" with "should I skip it earlier", so I caused some confusion I guess
[16:42:57 CET] <dv_> jkqxz: yeah, this seems to be it, thanks! one last question: are these "invisible" frames unique to vp8, or do they also occur in vp9 and vp6?
[16:46:08 CET] <jkqxz> VP9 and AV1 both have them too, and in those there is a special "display old invisible frame" command.
[20:47:01 CET] <kab0m> Hello everyone
[20:49:17 CET] <kab0m> Im having troubles using ffmpeg on the cli. "ffmpeg -i old.flac -acodec alac new.m4a" does not copy every metatag present in the FLAC file to the converted ALAC file. Please help me, i have quite a lot of files to convert. tia
[20:50:23 CET] <alone-x> ct
[20:50:28 CET] <alone-x> kab0m, hello
[20:50:38 CET] <alone-x> may be u know how to encode with qsv?
[20:51:06 CET] <kab0m> alone-x: no, what is qsv?
[20:51:29 CET] <alone-x> gpu acceleration
[20:51:34 CET] <alone-x> with intel CPU
[20:52:23 CET] <klaxa> kab0m: i think you want -map_metadata https://www.ffmpeg.org/ffmpeg.html#Advanced-options
[20:52:45 CET] <kab0m> alone-x: i dont really understand what gpu acceleration has to do with the tags?
[20:53:15 CET] <kab0m> klaxa: thanks for this info, i allready tried "-map_metadata 0 -id3v2_version 3" but without success
[20:53:22 CET] <alone-x> well, with that will be faster.
[20:53:59 CET] <kab0m> alone-x: thanks for this tipp ;) but the speed is not the problem..the problem is about missing metatags after conversion
[20:54:37 CET] <furq> which tags are missing
[20:55:03 CET] <alone-x> well, all i need reducing CPU usage with screen capturing
[20:57:08 CET] <kab0m> furq: quite a lot...like BARCODE, CD-NUMBER, ITUNESADVISORY, etc...
[20:58:33 CET] <kab0m> I just want ffmpeg to copy over every metatag, not just a few...but i dont know how to do this. "-map_metadata 0" wasnt successful
[20:59:31 CET] <furq> i don't think the m4a muxer will write any itunes tags
[20:59:55 CET] <furq> maybe try the latest ffmpeg if you're not already on it
[21:00:11 CET] <alone-x> furq hello
[21:00:23 CET] <alone-x> may be you know how to usd QSV?
[21:00:26 CET] <furq> i do not
[21:00:30 CET] <alone-x> i have no success
[21:00:40 CET] <alone-x> but with bandicam it's working
[21:01:41 CET] <kab0m> furq: hmmm, i dont understand where the problem is. i mean, i can successfully copy every missing tags by hand via kid3, but this is no option for a few thousand files...
[21:02:31 CET] <furq> all the tags you mentioned are itunes-specific extensions, they're not part of the spec
[21:02:35 CET] <furq> so ffmpeg just doesn't bother with them
[21:02:42 CET] <pink_mist> kab0m: I always used foobar2000 on windows to manage tags
[21:02:43 CET] <furq> you could write a mutagen script or something that runs after ffmpeg
[21:03:25 CET] <furq> http://vpaste.net/peWR6
[21:03:27 CET] <furq> something like that
[21:03:34 CET] <furq> i'm pretty sure mutagen respects itunes-specific metadata
[21:03:48 CET] <kab0m> pink_mist: thanks for the tip, but im on linux/mac and need automated script for this
[21:05:05 CET] <kab0m> furq: thank you, i never worked with phyton before...i tried to get this working with kid3-cli but had no success so far
[21:07:51 CET] <furq> ffmpeg -i foo.flac -c:a alac bar.m4a && copytags foo.flac bar.m4a
[21:07:53 CET] <furq> should work
[21:09:09 CET] <kab0m> is copytags a package i have to install or the phyton-script you posted before?
[21:09:14 CET] <furq> copytags is the script i posted
[21:09:22 CET] <furq> you'll need to install mutagen for python 3 but that should be in your repos
[21:09:42 CET] <kab0m> furq: thanks for the help, i will test and report :)
[21:15:29 CET] <ariyasu> hello
[21:16:10 CET] <ariyasu> "ffmpeg -t 60 -i input.mkv -c copy output.mkv"
[21:16:26 CET] <ariyasu> this is only copying tracks set to default and skipping all other tracks
[21:16:34 CET] <ariyasu> is there someway to keep them all?
[21:16:42 CET] <klaxa> add -map 0
[21:16:49 CET] <ariyasu> thanks
[21:16:55 CET] <klaxa> ffmpeg -t 60 -i input.mkv -map 0 output.mkv should work
[21:17:29 CET] <ariyasu> yes that worked :)
[21:17:31 CET] <ariyasu> thank you klaxa
[21:17:40 CET] <klaxa> :)
[21:22:11 CET] <kab0m> furq: it did not work. when i call teh script via "python copytags.py in.flac out.m4a" nothing happens
[21:24:03 CET] <pink_mist> did the tags not get copied?
[21:24:20 CET] <kab0m> pink_mist: nope
[21:45:58 CET] <kab0m> furq: where do you have this script from? Do you if can get any support for this script or have you written this by yourself? Does anyone else has an idea how i can achieve this task? tia
[21:46:51 CET] <furq> i wrote it myself
[21:47:01 CET] <furq> i checked mutagen and it doesn't look like their tag mapping is as good as i thought
[21:49:20 CET] <kab0m> furq: ahh i see. So i guess this mean that the script wont work? Im a total noob in this, so i don't know any script/tool which does exactly what i need except kid3-cli ... but i dont understand the syntax from kid3-cli exactly :(
[21:50:26 CET] <kab0m> Does anyone in here know about how to work with kid3-cli?
[23:09:11 CET] <GuiToris> hello, how can I check how many frames a specific video clip has?
[23:09:46 CET] <durandal_1707> by demuxing and/or decoding all of them
[23:10:17 CET] <GuiToris> oh :(
[23:12:13 CET] <GuiToris> and what's wrong with yuvj420p? ffmpeg says: deprecated pixel format used, make sure you did set range correctly
[23:12:37 CET] <durandal_1707> everything is wrong
[23:13:11 CET] <GuiToris> which one should I use? I'm converting still jpeg images
[23:14:01 CET] <durandal_1707> just ignore it, if you are not changing color range
[23:14:51 CET] <GuiToris> well okay, thanks :)
[00:00:00 CET] --- Sun Dec  9 2018