[Ffmpeg-devel-irc] ffmpeg.log.20160829

[burek]

[00:05:29 CEST] <ozette> what are alternatives to hls?
[00:06:14 CEST] <ozette> i feel like not all browsers like hls
[00:07:39 CEST] <Mavrik> HLS is pretty much the best you can do.
[00:07:54 CEST] <Mavrik> There's also DASH but it's a standard from hell that's even less supported.
[00:08:16 CEST] <Mavrik> (I suggest using a streaming server that can do HLS and DASH)
[00:11:20 CEST] <DHE> there is a javascript library that will convert HLS to MP4 in-browser. still requires the browser capable of H264 playback
[00:14:05 CEST] <ozette> hmm
[00:15:58 CEST] <ozette> i will see if I can find such a js library, sounds useful, at least as a fallback
[00:16:34 CEST] <Mavrik> Flash is a more useful fallback in most cases we deployed that
[00:18:07 CEST] <ozette> also an idea
[05:15:56 CEST] <ytan> Hello there
[05:16:11 CEST] <ytan> I need some help with ffmpeg rtsp streaming
[05:16:46 CEST] <ytan> Whenever I start streaming, ffmpeg.exe crashes.
[05:18:58 CEST] <ytan> Here are the messages that appeared when the crash happens http://pastebin.com/CHvedJ8g
[05:19:23 CEST] <ytan> Any ideas?
[06:40:06 CEST] <madprops> so
[06:40:08 CEST] <memeka> hi, how can i output to caca ?
[06:40:12 CEST] <madprops> is libav in debian the same as ffmpeg?
[06:40:15 CEST] <memeka> with ffplay?
[06:44:38 CEST] <memeka> anyonw?
[06:44:44 CEST] <madprops> nvm i read about it
[06:44:52 CEST] <madprops> now the confusion is that supposedly debian is returning to ffmpeg
[06:56:38 CEST] <furq> it returned to ffmpeg ages ago
[06:56:52 CEST] <furq> if you're on stable for some reason then you can install it from backports
[07:01:29 CEST] <memeka> how can i play a video to caca output with ffmpeg??
[07:11:31 CEST] <durandal_170> memeka: add -f caca  smthing I don't remember
[07:11:42 CEST] <memeka> durandal_170: doesn't work :(
[07:12:23 CEST] <durandal_170> memeka: pastebin full output you tried
[07:13:20 CEST] <memeka> durandal_170: tried: ffplay -an ./sintel_trailer-720p.mp4 -f caca
[07:13:26 CEST] <memeka> output: http://paste.debian.net/794474/
[07:15:15 CEST] <durandal_170> memeka: ffplay supports caca via sdl
[07:15:34 CEST] <memeka> --enable-libcaca
[07:15:46 CEST] <memeka> should be libcaca, not sdl,  rite?
[07:15:51 CEST] <durandal_170> so compile sdl with caca support
[07:16:26 CEST] <durandal_170> memeka: no, that option is for ffmpeg command
[07:16:47 CEST] <memeka> can i then use ffmpeg?
[07:17:02 CEST] <durandal_170> yes...
[07:17:14 CEST] <memeka> durandal_170:  ffmpeg -an -i ./sintel_trailer-720p.mp4 -pix_fmt rgb24 -f caca ncurses
[07:17:21 CEST] <memeka> got it to work
[07:17:38 CEST] <memeka> durandal_170: now, can i set the decoder manually?
[07:18:12 CEST] <durandal_170> ? I don't understand
[07:19:20 CEST] <shkm> Hi guys, a question: Im trying to place one or several overlays on a video. These overlays can be images or videos themselves. No problem, but I need to scale these overlays before applying them to the base video (which can also be an image). How can I do that? If I scale, for example, [1:v] as the last filter_complex operation, that ends up being the entire output. Command example: http://pastebin.com/BPXKTRQC
[07:19:21 CEST] <memeka> durandal_170: e.g. i wanna add -vcodec h264
[07:20:43 CEST] <durandal_170> shkm: then add scales before overlay
[07:21:03 CEST] <durandal_170> memeka: for encoding?
[07:21:05 CEST] <shkm> Will give that a try, durandal_170. Thanks :)
[07:21:09 CEST] <memeka> durandal_170: i want to use a hw-accelerated decoder
[07:21:31 CEST] <memeka> but the decoder output format is not rgb24
[07:21:37 CEST] <memeka> which is the input for caca
[07:22:02 CEST] <durandal_170> yes, caca supports only that
[07:22:02 CEST] <memeka> so i need to specify decoder, and then convert nv12 (decoder output) to rgb24 ... how to do that?
[07:23:48 CEST] <durandal_170> -vf format_-_rgb32
[07:23:54 CEST] <durandal_170> ?
[07:24:04 CEST] <durandal_170> never tried
[07:30:19 CEST] <memeka> durandal_170: something like:  ffmpeg -an -i ./sintel_trailer-720p.mp4 -vcodec h264_v4l2m2m -device /dev/video0 -f caca ncurses -vf format=pix_fmts=rgb24 ?
[07:31:47 CEST] <durandal_170> put vf format before -f caca ncurses
[07:33:06 CEST] <memeka> tried it like that ... segm fault in any case
[07:33:45 CEST] <memeka> [swscaler @ 0x89600] No accelerated colorspace conversion found from yuv420p to rgb24.
[07:33:58 CEST] <memeka> [h264_v4l2m2m @ 0x12d0f0] Suggested pixel format rgb24 is not accepted on output pool, will guess one.
[07:34:04 CEST] <memeka> then segm fault :(
[07:34:18 CEST] <memeka> maybe the h264_v4l2m2m  has issues
[07:34:25 CEST] <durandal_170> heh, no
[07:35:04 CEST] <durandal_170> perhaps you need to add hwdownload or something
[07:35:26 CEST] <memeka> what's hwdownload? :D
[07:35:50 CEST] <durandal_170> another filter
[07:36:36 CEST] <durandal_170> actually, you cant use hw decoders like that
[07:37:06 CEST] <memeka> durandal_170:  h264_v4l2m2m had a pull request more than 1 year ago, but was not accepted - i have a patched ffmpeg with it - so it might be that
[07:37:34 CEST] <memeka> in fact, this is what i wanna do - try h264_v4l2m2m  :D
[07:38:03 CEST] <durandal_170> I don't think that can work
[07:38:36 CEST] <durandal_170> hw accels are not decoders in ffmpeg any more
[07:39:12 CEST] <durandal_170> or thay are, dunno
[07:40:20 CEST] <memeka> so how should thinks work?
[07:40:34 CEST] <memeka> this one is accessed via v4l2
[07:41:11 CEST] <memeka> so u put in h264 in /dev/video0 and decoded nv12 comes out of it
[07:42:55 CEST] <durandal_170> uh, that is hacky
[07:43:22 CEST] <durandal_170> use pipe
[07:44:05 CEST] <durandal_170> with another ffmpeg instance to take nv12
[07:44:24 CEST] <durandal_170> and output it to caca
[07:45:07 CEST] <durandal_170> if there's container that supports nv12
[07:45:23 CEST] <durandal_170> as raw format
[08:29:01 CEST] <memeka> durandal_1707: still here?
[08:41:22 CEST] <memeka> durandal_1707: back?
[08:57:17 CEST] <shkm> Im overlaying a video onto something else (e.g. a jpg). This kind of works, but the overlaid video behaves like a static image. Is there any reason for this? Command: ffmpeg-loop 1 -i base.jpg -i overlay1.png -i overlay2.mp4 -filter_complex "[0:v] overlay=0:0:enable='between(t, 0, 10)' [stream1]; [stream1] overlay=0:0:enable='between(t, 10, 20)'" -c:v libx264 -pix_fmt yuv420p -t 20 -r 25 -y out.ts
[09:22:51 CEST] <durandal_1707> memeka: ?
[09:26:59 CEST] <memeka> durandal_1707: so, looks like the decoding works - i can get libcaca picture
[09:27:16 CEST] <memeka> the problem is that the hw decoder only likes annexB h264
[09:28:01 CEST] <memeka> durandal_1707: basically this works: ffmpeg -i ./sintel_trailer_1080p.mp4 -codec:v copy -codec:a none -bsf:v h264_mp4toannexb -f rawvideo - | ffplay -an -vcodec h264_v4l2m2m -
[09:28:18 CEST] <memeka> is there a way to make it annexb in ffplay directly?
[09:33:31 CEST] <soulshock> kind of unrelated: the website seems slow today
[09:33:48 CEST] <soulshock> meaning it takes over 10 seconds to get a response
[09:39:01 CEST] <durandal_1707> DOS attack?
[10:44:10 CEST] <Kadigan_KSB> Hey. I was wondering: since ffmpeg doesn't have "-target xdcamhd422" like ffmbc does, can someone please enlighten me what does that target mode entail, so I can set it for ffmpeg? (since ffmbc doesn't seem to have a Mac port)
[10:44:38 CEST] <Kadigan_KSB> Would anyone here even know?
[10:48:28 CEST] <furq> Kadigan_KSB: https://github.com/AndyA/ffmbc/blob/cb114a8bf3d2362757d8ced6dec3b58bd7f3501b/ffmbc.c#L4728-L4755
[10:49:22 CEST] <furq> https://github.com/bcoudurier/FFmbc/blob/ffmbc/ffmbc.c#L5201-L5224
[10:49:28 CEST] <furq> that one looks more up to date
[11:00:29 CEST] <Kadigan_KSB> furq: thank You kindly!
[11:00:46 CEST] <Kadigan_KSB> Now, I'm trying to figure out what to map color_transfer, color_primaries and color_matrix, if any
[11:26:04 CEST] <Fa1th> moin moin everyone :)
[11:27:47 CEST] <Kadigan_KSB> Okay, I had to figure out that flags2 changed shape into actual flags at some point.
[11:56:43 CEST] <Kadigan_KSB> Okay. I need a little more help here... I'm trying to convert 1080p25 to 1080i50. I'm doing '-top 1' (TFF), '-flas +ilme+ildct', and MediaInfo still claims the file is Progressive. What am I missing?
[11:56:53 CEST] <Kadigan_KSB> -flags*
[12:01:07 CEST] <Mavrik> Are you actually generating interlaced frames with interlace filter?
[12:03:34 CEST] <Kadigan_KSB> If I need to set up a filter, then no. I had no idea I have to use a filter, though it makes sense...
[12:04:12 CEST] <Kadigan_KSB> So what, -vf interlace ?
[12:11:00 CEST] <Fa1th> May I ask if its possible to create a vhs effect in a video without scripts in ffmpeg?
[12:14:42 CEST] <brontosaurusrex> Fa1th: That would be a combined effort of mayn filters i imagine, probably scaling width to around 200px would be first one...
[12:14:47 CEST] <brontosaurusrex> many*
[12:14:57 CEST] <Kadigan_KSB> Mavrik: so now I'm attempting to use -vf interlace, and I get a "Timecode frame rate 12/1 not supported". When I set -r 25 the video looks weird (like half of the frames were dropped and then duplicated)... unfortunately, so does -r 50.
[12:15:35 CEST] <Kadigan_KSB> Oh... RTFM. So it actually halves the FPS.
[12:15:54 CEST] <Kadigan_KSB> Is there a way to simply encode Progressive as TF/BF?
[12:29:02 CEST] <Kadigan_KSB> Funny thing... I specify -tag:v xd5c, and I always get xd5e.
[13:08:00 CEST] <ytan> Has anyone got streaming to work using TLS protocol?
[13:22:07 CEST] <shkm> Is there anything I should know when overlaying a video on top of an image? The overlaid video isnt played; it just shows statically.
[13:27:45 CEST] <kepstin> shkm: you'll probably have to turn the image into a video by looping it. Add the "-loop 1" and "-framerate <fps>" options before the still image input.
[13:29:14 CEST] <shkm> @kepstin I already have those set :(
[13:36:29 CEST] <shkm> @kepstin sorry, had to remove some private project stuff from the input/output. But its something like this: http://pastebin.com/GwNy5Hej
[13:39:46 CEST] <kepstin> hmm, i'd recommend using the -framerate input option, not -r (they do behave differently), but that shouldn't make a difference here :/
[13:40:57 CEST] <shkm> OK, switched it to -framerate anyway :)
[13:41:12 CEST] <kepstin> shkm: how long is the video in 'overlay.mp4'? that filterchain will be taking the section from 10s to 20s in that video.
[13:41:32 CEST] <kepstin> if that video isn't at least 20s long, it'll just show the last frame i think
[13:42:04 CEST] <shkm> Aha! Its well below 20s.
[13:42:18 CEST] <shkm> What I actually want is for the overlays to loop for those 10 seconds.
[13:42:20 CEST] <kepstin> shkm: sounds like you might want to use the concat filter :)
[13:42:27 CEST] <shkm> Ill go look at that. Thanks!
[14:23:25 CEST] <Kadigan_KSB> Huff. Puff. Turns out I can't use ffmpeg in a straightforward fashion for this encoding after all - the recipient does technical validation based on FourCC, not content, so all frames can be interlaced, but the file will be rejected on grounds of having xd5e (progressive) instead of xd5c.
[14:40:11 CEST] <c_14> Have you tried -vtag xd5c ?
[15:02:41 CEST] <brontosaurusrex> Kadigan_KSB: that mpeg2 version of xdcam right?
[15:03:48 CEST] <brontosaurusrex> Kadigan_KSB: I have this in my script: ffmpeg -hide_banner -i "$file" -i "$tmpdir/$base.L.wav" -i "$tmpdir/$base.R.wav" -map 0:v -map 1:a -map 2:a -vf setfield=tff -vcodec mpeg2video -pix_fmt yuv422p -b:v 50000k -minrate 50000k -maxrate 50000k -bufsize 17825792 -rc_init_occupancy 17825792 -sc_threshold 1000000000 -bf 2 -g 12 -intra_vlc 1 -non_linear_quant 1 -dc 10 -qmin 1 -qmax 12 -s 1920x1080 -flags
[15:03:50 CEST] <brontosaurusrex> +ilme+ildct -aspect 16:9 -acodec pcm_s24le -ar 48000 -timecode 00:00:00:00 -f mxf "$tmpdir/$base.mxf" -loglevel panic -stats </dev/null
[15:05:02 CEST] <Kadigan_KSB> c_14: yes, I have tried -vtag xd5c, -tag:v xd5c, both before and after specc'ing the format etc.
[15:05:35 CEST] <Kadigan_KSB> brontosaurusrex: suppose I need an XDCAM .mov file. :)
[15:05:54 CEST] <brontosaurusrex> Kadigan_KSB: mpeg2 or what?
[15:07:23 CEST] <Kadigan_KSB> Is there an XDCAM format OTHER than MPEG-2?
[15:07:54 CEST] <Kadigan_KSB> (if there is, that would be interesting to hear btw.)
[15:08:20 CEST] <brontosaurusrex> than thats it, just replace mxf with mov
[15:08:25 CEST] <Kadigan_KSB> Yes, I need an XDCAM HD422 mpeg2video in a .mov container for a delivery. The codec specified must be specifically xd5c.
[15:08:39 CEST] <Kadigan_KSB> Yes, well, no matter what I set up, I end up getting xd5e.
[15:09:17 CEST] <Kadigan_KSB> Even if I forcibly interlace the video - and it SHOWS - I still end up with xd5e. As I understand it, I would need to split into separate files and then remux.
[15:09:25 CEST] <Kadigan_KSB> (to be clear, this is a BUG)
[15:09:34 CEST] <JEEB> are you actually coding interlaced?
[15:09:44 CEST] <JEEB> and if yes, how you are configuring that?
[15:09:54 CEST] <Kadigan_KSB> MediaInfo and idet are claiming it's interlaced.
[15:10:02 CEST] <JEEB> because ffmpeg will happily encode interlaced content as progressive
[15:10:10 CEST] <JEEB> depending on the encoder
[15:10:19 CEST] <Kadigan_KSB> But it doesn't matter. I don't care for it being interlaced
[15:10:27 CEST] <Kadigan_KSB> (in fact, I need it claimed TFF but otherwise progressive)
[15:10:42 CEST] <Kadigan_KSB> I do care for it not being set to xd5c when I want to.
[15:10:50 CEST] <brontosaurusrex> Kadigan_KSB: how do you read that tag?
[15:11:13 CEST] <Kadigan_KSB> Okay, wait.
[15:11:59 CEST] <JEEB> Kadigan_KSB: well I'm just wondering if it (libavformat muxer) was selecting the flag depending on how it's coded :P
[15:15:51 CEST] <brontosaurusrex> ok, how do i read that flag, dont see anything with mediainfo or ffprobe
[15:17:11 CEST] <JEEB> if it's MOV-like use l-smash's boxdumper
[15:18:39 CEST] <JEEB> btw, I don't find the string xd5e in libavformat
[15:19:14 CEST] <JEEB> can you note the full command line used somewhere?
[15:22:03 CEST] <Kadigan_KSB> https://paste.ee/r/oZfHW
[15:29:14 CEST] <JEEB> yeah, it picks the xd5c from somewhere
[15:29:24 CEST] <JEEB> couldn't find it with a simple grep so I might take a look at that after work
[15:29:27 CEST] <michaelshroyer> im having a problem when uploading videos i record with my dslr to youtube. the audio is muffled when the video is played on a mobile phone without headphones
[15:30:01 CEST] <BtbN> that's what poor phone speakers do.
[15:30:09 CEST] <JEEB> because if there's separate identifiers for progressive and interlaced content and I *think* those flags should be enabling interlaced coding
[15:30:56 CEST] <michaelshroyer> ?
[15:31:06 CEST] <brontosaurusrex> Must be blind, but I dont see anything xd*
[15:31:45 CEST] <Kadigan_KSB> JEEB, brontosaurusrex: ex. "    Stream #0:0(und): Video: mpeg2video (xd5c / 0x63356478), yuv422p, 1920x1080 [SAR 1:1 DAR 16:9], q=1-28, 50000 kb/s, 25 fps, 12800 tbn, 25 tbc (default)
[15:31:45 CEST] <Kadigan_KSB> "
[15:31:57 CEST] <Kadigan_KSB> Video: mpeg2video (xd5c / ...
[15:32:03 CEST] <JEEB> yeah, I noticed that. anyways, will take a look after work
[15:32:07 CEST] <Kadigan_KSB> Claims to be encoding xd5c
[15:32:11 CEST] <Kadigan_KSB> but when I ffmpeg idet,
[15:32:12 CEST] <JEEB> that's the muxing
[15:32:13 CEST] <Kadigan_KSB> it says xd5e.
[15:32:36 CEST] <JEEB> what does boxdumper say?
[15:32:50 CEST] <Kadigan_KSB> What's that?
[15:33:06 CEST] <JEEB> a very useful tool for looking at what's actually written in a MOV-like file (MP4 etc. go as well)
[15:33:11 CEST] <JEEB> comes in the L-SMASH project
[15:33:23 CEST] <JEEB> https://github.com/l-smash/l-smash
[15:33:33 CEST] <bencoh> libavformat/isom.c:214:    { AV_CODEC_ID_MPEG2VIDEO, MKTAG('x', 'd', '5', 'e') }, /* XDCAM HD422 1080p25 CBR */
[15:33:50 CEST] <Kadigan_KSB> JEEB: know the equivalent for OS X?
[15:33:59 CEST] <JEEB> Kadigan_KSB: in OS X it should be pretty simple to build
[15:34:06 CEST] <JEEB> ./configure and make
[15:34:11 CEST] <JEEB> it doesn't have real dependencies
[15:34:15 CEST] <JEEB> other than a compiler
[15:34:27 CEST] <JEEB> and in OS X you get that pretty easily with XCode/tools
[15:34:52 CEST] <Kadigan_KSB> Yeah, running config nowe
[15:34:55 CEST] <Kadigan_KSB> now*
[15:35:21 CEST] <JEEB> after make you should be able to call `./cli/boxdumper --help` I think
[15:36:08 CEST] <Kadigan_KSB> jesus that's a lot of info
[15:36:17 CEST] <Kadigan_KSB> Which switches? Or do I paste everything?\
[15:36:19 CEST] <JEEB> and then you can dump the structure of a file with `./cli/boxdumper --box INPUT_FILE > boxdumper_info.txt`
[15:36:29 CEST] <brontosaurusrex> Kadigan_KSB: Interesting, I have nothing in my mxf xdcams that would look like xd*
[15:36:31 CEST] <JEEB> and then you can search for xd5e
[15:36:39 CEST] <brontosaurusrex> both generated by ffmpeg or Adobe
[15:36:41 CEST] <JEEB> brontosaurusrex: MXF is separate
[15:36:44 CEST] <JEEB> completely different
[15:36:45 CEST] <JEEB> this is MOV
[15:36:48 CEST] <brontosaurusrex> oh
[15:36:56 CEST] <Kadigan_KSB> [xd5e: Visual Description]
[15:37:01 CEST] <Kadigan_KSB> (no match for xd5c)
[15:37:04 CEST] <brontosaurusrex> Ok, that makes sense :)
[15:37:08 CEST] <JEEB> yeah, then the file only has xd5e
[15:37:11 CEST] <Kadigan_KSB> That's on OUTPUT_1080i25.mov
[15:37:25 CEST] <JEEB> and you have checked the actual contents of the file instead of trusting some app ;)
[15:37:33 CEST] <Kadigan_KSB> Yeah, I used my eyes to see the comb pattern.
[15:37:40 CEST] <JEEB> that's completely separate
[15:37:49 CEST] <JEEB> you don't need to have combing for interlaced coding
[15:37:51 CEST] <JEEB> coding != content
[15:38:15 CEST] <Kadigan_KSB> Okay. Please propose a good way to "check" without "trusting some app" and yet not using my eyes.
[15:38:35 CEST] <Kadigan_KSB> I can't read rawhex GOP or anything.
[15:38:42 CEST] <Kadigan_KSB> :)
[15:38:49 CEST] <_jason_> hello
[15:40:02 CEST] <JEEB> Kadigan_KSB: the written MOV ID thing can be checked with boxdumper (and yes, ffmpeg cli also shows it), and the coding mode should be check'able with ffprobe/ffmpeg
[15:40:08 CEST] <JEEB> content is then something you have to check with eyes
[15:41:01 CEST] <Kadigan_KSB> Well, all I can say is this: the source file has no combing, the resultant i25 file has combing, at least visually.
[15:41:22 CEST] <Kadigan_KSB> I do realize that it may simply be w/ some fill, or multiplexed or something
[15:41:26 CEST] <Kadigan_KSB> but then, how do I trust it?
[15:41:29 CEST] <JEEB> well my point was that coding mode has nothing to do with the actual content
[15:41:41 CEST] <JEEB> you can code interlaced content with progressive modes in encoders, and vice versa
[15:41:49 CEST] <Kadigan_KSB> I know.
[15:41:58 CEST] <Kadigan_KSB> My problem is that I consistently get xd5e,
[15:42:08 CEST] <Kadigan_KSB> and the verification software consistently rejects it for not being xd5c.
[15:42:09 CEST] <JEEB> yes, and I already told you that I'll poke at this after work :P
[15:42:24 CEST] <Kadigan_KSB> Oh, I apologize if I'm scanning a tiny tiny bit
[15:42:33 CEST] <Kadigan_KSB> (I'm kinda at work >.>)
[15:42:44 CEST] <Kadigan_KSB> Sorry.
[15:43:16 CEST] <JEEB> thankfully I don't have to use MPEG-2 encoding any more for anything at work, but if it's fix'able I'll try to do it. I'm just never sure when those things get set so you might or might not have to manually set the interlacism in some way in the command line
[15:43:39 CEST] <Kadigan_KSB> All I know is that ffmbc does it right.
[15:43:54 CEST] <JEEB> and that code we can't see because it's GPL and we can't take those changes back as LGPL :P
[15:44:23 CEST] <JEEB> but it hopefully shouldn't be too hard
[15:44:36 CEST] <Kadigan_KSB> Well, the way I see it, it's a bug in ffmpeg (... or so I think - there's an entry about it in the bug tracker)
[15:44:41 CEST] <JEEB> yes, it is a bug
[15:44:57 CEST] <Kadigan_KSB> so any fix ffmbc guys did ... theoretically contributes to core ffmpeg ;P
[15:45:16 CEST] <durandal_1707> Nope
[15:45:26 CEST] <JEEB> the ffmbc author made the decision to make his stuff GPL, which means that unless we make stuff GPL we can't apply those changes
[15:45:30 CEST] <durandal_1707> ffmbc is gpl only
[15:45:42 CEST] <JEEB> and yes, he has the right to do that since you can switch LGPL to GPL
[15:45:49 CEST] <Kadigan_KSB> In any case, I'm sorry for being a bit of a spoiled princess, but I'm kinda backed into a corner of my boss wondering why I'm still spending time on this. (he's the kind of guy that "will cross that bridge when he comes to it", or "we'll worry about HDD data when they break")
[15:45:52 CEST] <JEEB> it just means that it's a one-way street in that sense :P
[15:46:10 CEST] <JEEB> Kadigan_KSB: hah, then I wonder why you're not looking into the movenc muxer then :P
[15:46:23 CEST] <JEEB> anyways, back to work for me
[15:46:36 CEST] <Kadigan_KSB> JEEB: I'm just saying that nobody says You should take their code, but since it's o/s, You're more than welcome to see HOW they fixed it, and WHERE the issue is
[15:46:38 CEST] <Kadigan_KSB> or sth.
[15:46:59 CEST] <JEEB> as soon as you look at it or make a "derivative" of that code it's GPL
[15:47:02 CEST] <JEEB> :P
[15:47:27 CEST] <JEEB> although I think this can be solved in a similar manner to the colorimetry-specific fourccs for AVI and Ut Video
[15:47:30 CEST] <Kadigan_KSB> So what, if someone GPLs something and I think of the exact same thing (but not in the exact same code), I still can't claim it?
[15:47:52 CEST] <bencoh> nobody talked about "claiming"
[15:48:03 CEST] <JEEB> basically we're LGPL and we want to keep our stuff LGPL
[15:48:10 CEST] <JEEB> he has used his right to make his changes GPL only
[15:48:12 CEST] <bencoh> just that once you've looked at an implementation, then ....
[15:48:18 CEST] <JEEB> ^this
[15:48:20 CEST] <Kadigan_KSB> Besides, like I said - it's a bug in ffmpeg. It's not like You will have to go out of Your way to fix it in a way that's different from ffmbc, if the solution is trivial - right?
[15:48:30 CEST] <JEEB> no, I just won't look at ffmbc
[15:48:34 CEST] <Kadigan_KSB> Okay.
[15:48:39 CEST] <Kadigan_KSB> Fair enough.
[15:48:57 CEST] <bencoh> Kadigan_KSB: you could look at ffmbc, have a look at ffmpeg, and point at the bug in ffmpeg, though
[15:49:10 CEST] <bencoh> and then someone would write a patch for ffmpeg
[15:49:16 CEST] <JEEB> but he can't post any code from ffmbc
[15:49:29 CEST] <JEEB> anyways, I hope at some point you will in such situations take a look at the FFmpeg code yourself :P
[15:49:42 CEST] <JEEB> because it's not magical pixie dust
[15:50:29 CEST] <Kadigan_KSB> I assume it's not, but I can only do that on my own time... And this is kind of a pain, because we're again 2 days to deadline, and I still need to keep running between two PCs to make stuff happen. It's never a crisis until it is, here.
[15:50:51 CEST] <JEEB> hah
[15:51:30 CEST] <Kadigan_KSB> Anyway, full-on crisis mode hat, on you go
[15:51:33 CEST] <Kadigan_KSB> back to work.
[15:51:35 CEST] <Kadigan_KSB> Thanks!
[16:27:06 CEST] <deivid___> Hello! I have a video that, by default, doesn't detect the codec parameters, running ffprobe with -analyzduration 16M gives me the correct values. Can I fix without re-encoding?
[16:27:28 CEST] <deivid___> I'm getting: [mov,mp4,m4a,3gp,3g2,mj2 @ 0x1bbf800] Could not find codec parameters for stream 0 (Video: h264 (avc1 / 0x31637661), none, 1280x720, 1000 kb/s): unspecified pixel format
[16:28:04 CEST] <kepstin> weird that it works with -analyzeduration 16M, that means it must be a segmented file of some sort?
[16:28:31 CEST] <kepstin> you can probably mix it by just remuxing to a new file, use '-c copy' to copy the video/audio streams without re-encoding.
[16:29:30 CEST] <deivid___> I edited a few h264 files with kdenlive; it seems like it doesn't output "good" files. This has happened on 12 large files that's why I'm trying to avoid reencoding
[16:29:48 CEST] <deivid___> Remuxed to another mp4 and an mkv, same thing
[16:29:52 CEST] <BtbN> raw h264 is not good by any definition
[16:30:13 CEST] <deivid___> nah it's not raw; a combination of h264+.. in different formats. flv/mkv/mp4 from different sources
[16:34:48 CEST] <deivid___> Ok; copying the streams directly doesn't work. Splitting audio and video to different files and remuxing works. Do I report this?
[16:42:31 CEST] <Kadigan_KSB> Sounds kind of like "working as intended".
[16:45:54 CEST] <deivid___> Why does copying audio+video at the same time (from a broken file) result in a broken file, but copying them separately and merging them work?
[16:46:43 CEST] <Kadigan_KSB> Garbage in, garbage out.
[16:47:22 CEST] <Kadigan_KSB> The error is somewhere in the mux. Demuxing to two separate (perfectly valid) files removes the bullshit info. Remuxing them creates a new, valid mux.
[16:47:27 CEST] <Kadigan_KSB> Or that's how it looks to me.
[16:48:06 CEST] <deivid___> Then I'd like to request the feature "-force_remux" or something like that to avoid this
[16:50:55 CEST] <Kadigan_KSB> Which would -probably- work... except ffmpeg might need to write to a number of different temp streams/files on the way... Not sure that's how it should behave...
[17:02:52 CEST] <_jason__> hello
[17:03:13 CEST] <_jason__> i have download ffmpeg and complied
[17:03:38 CEST] <_jason__> running this command to capture desktop
[17:04:06 CEST] <_jason__> ./ffmpeg -f x11grab -r 25 -s 361x176 -i :0.0+0,24 -vcodec libx264 -threads 0 video.mkv
[17:04:28 CEST] <_jason__> nothing is showing in video.mkv just black screen
[17:05:32 CEST] <Kadigan_KSB> (asdide from my suspicion about overlay video modes,
[17:05:38 CEST] <Kadigan_KSB> aside*
[17:06:08 CEST] <Kadigan_KSB> I would suggest You encode to something less demanding, like motion jpeg, first)
[17:06:20 CEST] <_jason__> and how to do that
[17:08:48 CEST] <_jason__> anyone?
[17:08:53 CEST] <_jason__> hello?
[17:09:23 CEST] <_jason__> i'm recording screen in ubuntu and nothing appears
[17:09:37 CEST] <durandal_1707> _jason__: use better player
[17:09:45 CEST] <_jason__> better player?
[17:10:09 CEST] <durandal_1707> or encode to yuv420p
[17:10:33 CEST] <durandal_1707> There should be message
[17:10:52 CEST] <durandal_1707> Post full logs to pastebin
[17:12:11 CEST] <_jason__> ok
[17:17:08 CEST] <_jason__> http://pastebin.com/3qLM0sn2
[17:17:17 CEST] <_jason__> @durandal here you go
[17:20:58 CEST] <mrelcee> GI_Jack  ping?
[17:21:02 CEST] <mrelcee> nope he isn't here
[17:21:31 CEST] <mrelcee> figured out where freebsd 4 kept it's noodles..
[17:21:51 CEST] <mrelcee> hit me as i was near unconsciousness last night going to bed..
[17:27:35 CEST] <durandal_1707> _jason__: those are configure logs
[17:32:48 CEST] <_jason__> let me clear logs and then create again
[17:34:27 CEST] <_jason__> i cleared the logs and then started again but it seems nothing is writing into config.log... now what?
[17:38:04 CEST] <durandal_1707> _jason__: ffmpeg logs!
[17:39:02 CEST] <_jason_> oh ok
[17:42:14 CEST] <_jason__> http://pastebin.com/Gfk1aait
[17:42:59 CEST] <afnj> What is better for decoding (h264) in ffmpeg, a higher CPU frequency or more CPU cores?
[17:43:34 CEST] <nonex86> afnj: how many streams plan to decode simultaneously?
[17:44:03 CEST] <afnj> multiple.. its 10bit stuff too
[17:44:21 CEST] <nonex86> afnj: multiple... how many exactly ? :)
[17:45:00 CEST] <nonex86> afnj: also, resolution/bitrate does matter
[17:45:16 CEST] <afnj> four 10bit h264 streams... 1080p content
[17:45:33 CEST] <afnj> Looking at Xeon's trying to decide between more CPU cores or higher frequency for the same price
[17:46:34 CEST] <nonex86> just fyi, on my quad core haswell 4.0Ghz (4 physical cores/8 ht) i decoded 9 1080p 30fps h264 streams without any noticeable problems
[17:47:39 CEST] <nonex86> i guess i5 like cpu will be able to decode 4 streams without any problems
[17:47:53 CEST] <nonex86> also, does xeon support intel quick sync?
[17:48:24 CEST] <nonex86> if so, you can offload all decoding work to mfx engine
[17:48:27 CEST] <afnj> Not 10bit content, its the 10bit thats nasty, a lot of optimizations on 8bit processing don't work
[17:48:54 CEST] <nonex86> well, cant say anything about 10bit per color component
[17:50:01 CEST] <nonex86> http://ic.pics.livejournal.com/vedmysh/21083959/13708/13708_original.png
[17:50:12 CEST] <nonex86> 16 hd streams, stable fps on all
[17:50:32 CEST] <nonex86> decoded using intel quick sync
[17:51:06 CEST] <afnj> Nice! QSV has gotten really good
[17:51:08 CEST] <nonex86> around 5% cpu load
[17:51:18 CEST] <afnj> What CPU is that?
[17:51:35 CEST] <nonex86> same haswell
[17:51:43 CEST] <nonex86> let me check the model
[17:52:15 CEST] <nonex86> honestly speaking dxva decoder over ffmpeg works well too
[17:52:48 CEST] <nonex86> i7 4790k
[18:01:30 CEST] <afnj> cool thanks yeah thats a nice CPU
[18:51:05 CEST] <pxl_> hey everyone, need help on syntax if anyone has a minute?
[18:52:27 CEST] <pxl_> I have a subtitle file that needs utf-8 encoding for special characters. I am using a video filter though and cant find an example online to show how to use -sub_charenc with a vf
[18:52:46 CEST] <pxl_> here is my command  http://pastebin.com/ecjNNRbj
[18:55:43 CEST] <c_14> -vf subtitles=blah.srt:sub_charenc=UTF-8
[18:55:47 CEST] <c_14> though isn't utf-8 default?
[18:56:17 CEST] <c_14> eh
[18:56:25 CEST] <c_14> -vf subtitles=blah.srt:charenc=UTF-8
[19:02:15 CEST] <pxl_> thanks, going to give it a try. I think ansi is default
[21:06:15 CEST] <pgorley> can h263p be accelerated using the h263_vaapi and h263_videotoolbox hwaccels?
[22:11:58 CEST] <deweydb> is there any way to do this faster:
[22:11:59 CEST] <deweydb> ffmpeg -i input.flv -vf "select='eq(pict_type,PICT_TYPE_I)'" -vsync vfr thumb%04d.png
[22:12:12 CEST] <deweydb> i.e. extract I fromes from video
[22:12:32 CEST] <deweydb> it seems incredibly slow. even when i limit it to a 90 frame chunk of video.
[22:13:23 CEST] <c_14> try adding -skip_frame nokey
[22:17:17 CEST] <deweydb> thanks i'll try that.
[22:18:49 CEST] <kepstin> also, png encoding is pretty slow
[22:20:29 CEST] <deweydb> oh. good to know, should i use jpg?
[22:20:33 CEST] <deweydb> whats fastest?
[22:24:27 CEST] <c_14> if the input is yuv, raw yuv will be fastest
[22:26:14 CEST] <kepstin> it really depends what you're going to be doing with the images after - if you're just feeding them into another ffmpeg later, some raw yuv format is great. If you need to open them in a graphics editor, but have lots of disk space, consider maybe bmp?
[22:26:41 CEST] <kepstin> if you're e.g. finding thumbnails to use on a website, you probably want jpg
[22:26:44 CEST] <deweydb> i'm computing the laplacian transfer to find the least blurry ones, then saving those to disk
[22:27:06 CEST] <kepstin> deweydb: so it really depends what image formats your analysis tool supports :)
[22:27:30 CEST] <deweydb> yeah, i gotta look into if i can feed raw yuv into python PIL. i think i can.
[22:27:36 CEST] <deweydb> actually it might save some processing power
[22:27:45 CEST] <deweydb> cause i think right now i'm compressing, then expanding
[22:43:56 CEST] <deweydb> when i tried the -skip_frame nokey
[22:43:58 CEST] <deweydb> i get this error:
[22:43:58 CEST] <deweydb> Codec AVOption skip_frame (skip decoding for the selected frames) specified for output file #0 (./test-%02d.jpg) is not an encoding option.
[22:44:10 CEST] <c_14> It's an input option, put it before -i
[22:44:18 CEST] <deweydb> ohhh
[22:48:46 CEST] <deweydb> hmm now i'm just getting no output at all "Output file is empty, nothing was encoded (check -ss / -t / -frames parameters if used)"
[22:48:52 CEST] <deweydb> "ffmpeg -skip_frame nokey -i "../video_in/1472486031995_UFE2MyBWaWRlbyA4LTI3LTE2Lk1PVg.MOV" -vf "select=eq(pict_type\,I)*between(n\,739\,929)" -vsync vfr ./test-%02d.jpg"
[22:57:22 CEST] <deweydb> O.O
[22:57:50 CEST] <deweydb> i know why its slow... this command is processing the full video beofre taking those frames from that small segment.
[22:57:59 CEST] <deweydb> but if i cut it first...
[22:58:01 CEST] <deweydb> aha!
[23:40:37 CEST] <deweydb> guys is there anything i can do to speed up this command:
[23:40:40 CEST] <deweydb> ffmpeg -i video_temp/1472506615-no-audio.mp4 -i assets/image/bid13-logo-150-transparent.png -filter_complex "overlay=x=(main_w-overlay_w-20):y=(main_h-overlay_h-20)" video_temp/1472506615-watermarked.mp4
[23:40:50 CEST] <deweydb> basically i want to add a watermark to the whole video.
[23:44:13 CEST] <relaxed> deweydb: -preset veryfast
[23:47:13 CEST] <deweydb> sorry, yeah, i usually pastebin, i was being lazy cause it was just one line.
[23:47:16 CEST] <deweydb> i will do that next time.
[23:49:19 CEST] <mgraczyk> Hello, I noticed that much of the same functionality is implemented in libavformat/oggparseopus.c and libavcodec/opusdec.c, but that the behavior in the two implementations differs in certain cases.
[23:49:32 CEST] <mgraczyk> Would it be desirable or possible to consolidate those two implementations?
[23:50:03 CEST] <mgraczyk> I actually meant to say libavformat/oggparseopus.c and libavcodec/opus.c
[23:50:19 CEST] <DHE> there are a lot of codecs with parsers and decoders.
[23:51:47 CEST] <mgraczyk> In this case it seems that the header parsing logic in oggparseopus.c is much less detailed than the logic in libavcodec/opus.c. Is that also common?
[23:57:58 CEST] <llogan> deweydb: it contains info that may reveal why it may be slow
[00:00:00 CEST] --- Tue Aug 30 2016