[Ffmpeg-devel-irc] ffmpeg.log.20150410

burek burek021 at gmail.com
Sat Apr 11 02:05:01 CEST 2015


[00:00:41 CEST] <DragonsLord> I asked this because I still have a black line on the left
[00:00:47 CEST] <DragonsLord> and I am not able to crop it
[00:02:17 CEST] <justinX> are there way to encode in a video that its size is some pixels smaller than the data really is? I mean so it crops the video say 3 pixel left and 7 bottom while playing, without having to change video data itself.
[00:02:59 CEST] <justinX> perhaps something in the container format or such...
[00:03:50 CEST] <llogan> justinX: ffplay -vf crop=w:h:x:y input
[00:03:53 CEST] <justinX> for all the people that thinks multiple is 8 is booring :-D
[00:03:56 CEST] <justinX> ok nice!
[00:04:41 CEST] <llogan> ffplay lets you take a crop
[00:04:50 CEST] <llogan> on your video
[00:05:21 CEST] <DragonsLord> llogan, it seems to be not working for me :(
[00:05:21 CEST] <justinX> but ffplay?  that would only be when you play the video with that right? I want something that tells every player (vlc, webpages etc) how to crop the video when they play it.
[00:05:51 CEST] <llogan> justinX: oh. i don't know.
[00:06:29 CEST] <llogan> DragonsLord: sorry, but i'm not sure what you're doing exactly.
[00:08:01 CEST] <DragonsLord> llogan, crop=w=iw-118:h=ih-4:x=52:y=4,scale=w=640:h=480:flags=bilinear
[00:09:28 CEST] <DragonsLord> I also received this "Warning: data is not aligned! This can lead to a speedloss"
[00:10:12 CEST] <DragonsLord> at the end
[00:10:13 CEST] <DragonsLord> I still have a black line on the left
[00:10:18 CEST] <DragonsLord> and I am not able to crop it
[00:12:53 CEST] <DragonsLord> :(
[00:13:36 CEST] <llogan> you can ignore that warning
[00:14:39 CEST] <sg90> Adding to my previous post regarding segments and timecode, I've actually noticed the second segment at t=30 doesnt have the correct timecode, its supposed to be 00:00:30:00 but it is just an int - 42384379 - and never increments
[00:15:31 CEST] <DragonsLord> llogan, OK but ... at the end I continue having a black line of about 2 pixels on the left
[00:15:33 CEST] <DragonsLord> :(
[03:23:04 CEST] <jookiyaya> is there big difference between AC3 and DTS?
[04:59:51 CEST] <noobert> http://i.imgur.com/279K7DC.png
[04:59:56 CEST] <noobert> pls halp
[05:40:04 CEST] <pentanol> noobert could you make grep -ri nvEncode ./ in yours ffmpeg sources folder
[05:40:19 CEST] <brotherBox> he's gone
[05:41:20 CEST] <pentanol> really, thanks
[05:41:38 CEST] <relaxed> if he comes back he needed --extra-cflags, and not cxxflags
[12:51:32 CEST] <sybariten> uh ok hai
[12:52:15 CEST] <sybariten> would anyone be so kind to give me a short crash course in what presets are, where ffmpeg on win expects to find them, and how i should refer to them on the command line
[12:53:02 CEST] <sybariten> I'm suspecting this is all being complicated by the fact that i'm trying to run ffmpeg from bash, also ... should maybe go with cmd32
[12:54:52 CEST] <c_14> sybariten: do you mean the ffpresets/avpresets?
[12:55:23 CEST] <sybariten> c_14: yes i think so.
[12:56:00 CEST] <sybariten> "file for preset 'normal' not found"
[12:56:07 CEST] <c_14> https://ffmpeg.org/ffmpeg.html#Preset-files
[13:01:38 CEST] <sybariten> i can do a hard refer using -fpre but then i get another error: invalid option or argument:  directpred=3
[13:04:05 CEST] <c_14> That doesn't look like a valid ffmpeg option to me. (isn't listed in ffmpeg -h full)
[13:04:50 CEST] <sybariten> ok
[13:04:57 CEST] <sybariten> i tried a simpler command lnie now anyhow
[13:05:05 CEST] <sybariten> but fuck if i dont get other errors... sigh
[13:15:38 CEST] <sybariten> " -sameq was removed , use -qscale 0 "
[14:52:15 CEST] <Naminor> Hello there
[14:52:33 CEST] <Naminor> I'm facing an issue with ffmpeg. I'll explain the situation
[14:53:46 CEST] <Naminor> We need for our client to make screencast of webpages. So we decided to create servers (Ubuntu) put ffmpeg on it and with an API system call the ffmpeg command to screencast all the screen and the audio system
[14:54:09 CEST] <Naminor> so it records the web page and the sound of the webpage.
[14:54:46 CEST] <Naminor> Also ffmpeg is used to both stream and record locally the screencast.
[14:55:20 CEST] <Naminor> But we are facing random audio/video desync
[14:55:38 CEST] <Naminor> I tried several parameters and even power upthe serversz
[14:55:43 CEST] <Naminor> servers*
[14:56:08 CEST] <Naminor> we are now with a 6 Ghz Quad Core Xeon for one screencast agent
[14:56:19 CEST] <Naminor> and some time, it is out of sync
[14:56:31 CEST] <Naminor> could you help me with this problem ?
[15:00:46 CEST] <Venti> check that the liquid nitrogen hasn't run out, because 6GHz is kind of a lot
[15:06:41 CEST] <Naminor> it actualy run on a vmware esx
[15:06:50 CEST] <Naminor> we rent these servers
[15:07:35 CEST] <Naminor> an on full charge it consumed about 6400MHz of CPU
[15:20:59 CEST] <Naminor> here is my ffmpeg command if you want to take a look: ffmpeg -f x11grab -s 1280x720 -r 25 -i :0+0,0 -f pulse -i default -c:v h264 -vb 1024k -preset ultrafast -tune zerolatency -pix_fmt yuv420p -async 1 -acodec libmp3lame -ar 22050 -ac 2 -ab 50k -r 25 -f flv 'rtmp://XXXXXXXXXXXX/live/XXXXXXX'
[15:21:30 CEST] <Naminor> and ffmpeg version 1.2.6-7:1.2.6-1~trusty1
[15:23:46 CEST] <Naminor> any idea ?
[15:24:35 CEST] <c_14> Naminor: what's the first line of the output of `ffmpeg -version' ?
[15:24:57 CEST] <Naminor> ffmpeg version 1.2.6-7:1.2.6-1~trusty1
[15:26:20 CEST] <c_14> First of all, you should probably update your version of ffmpeg. That version is very old.
[15:26:45 CEST] <c_14> Second, have you tried without -async 1?
[15:27:20 CEST] <Naminor> I will try without -async 1
[15:27:41 CEST] <Naminor> yeah but we are on ubuntu and there are no recent repository (or should i compile it from source ?)
[15:29:23 CEST] <c_14> If you can compile from source, please do. There have been lots of bug fixes and general improvements since 1.2.6.
[15:30:59 CEST] <Naminor> Okay
[15:31:25 CEST] <Naminor> i've removed async, it's a little better but the desync happen later
[15:33:39 CEST] <c_14> Other things you could try are changing the first -r to -framerate. And if that doesn't help adding -af aresample=async=1000
[15:37:59 CEST] <Naminor> it is worse with framerate so i add -af aresample=async=1000 and keep framerate or do i rollback to -r ?
[15:38:32 CEST] <c_14> It's worse with -framerate? Weird. Try both.
[15:38:41 CEST] <c_14> Updating will probably be your best bet though.
[15:38:49 CEST] <Naminor> framerate + -af ?
[15:39:02 CEST] <c_14> Both -framreate and -af and -r and -af
[15:39:06 CEST] <c_14> *framerate
[15:45:39 CEST] <Naminor> no it is worse in the both cases
[15:48:15 CEST] <Naminor> i think i'm gonna go with the sources but i never did it before, i cloned the git so i just have to compile it with gcc ?
[15:49:22 CEST] <c_14> https://trac.ffmpeg.org/wiki/CompilationGuide/Generic https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu
[15:49:45 CEST] <Naminor> thank you !
[16:40:49 CEST] <karika200> hi
[16:41:09 CEST] <karika200> I'm restreaming an rtsp source to rtmp push with ffmpeg
[16:41:20 CEST] <karika200> in the rtsp source i have sometimes broken frames i think
[16:41:31 CEST] <karika200> how can i drop this corrupted frames with ffmpeg?
[16:43:29 CEST] <synthecypher> Is there a quick way to split a video into 30 second chunks?
[16:44:59 CEST] <karika200> ffmpeg -ss 00:00:00 -i video.mp4 -to 00:30:00 -c copy cut.mp4 ?
[16:45:24 CEST] <synthecypher> will take take a 3 minute video and split it into 30 chunks?
[16:45:31 CEST] <synthecypher> or just the first 30 seconds?
[16:46:10 CEST] <karika200> just first 30, but with minimal scripting you can step forward the whole video
[16:47:40 CEST] <Prelude2004c> join #ffmpeg-dev
[16:47:53 CEST] <Prelude2004c> oops
[16:48:06 CEST] <c_14> synthecypher: look at the segment muxer
[16:48:15 CEST] <synthecypher> c_14: kk
[16:48:16 CEST] <Prelude2004c> :) good morning everyone.. using up a new nvenc with ffmpeg.. now they changed form fast , very fast, etc etc to hq hp , etc
[16:48:34 CEST] <Prelude2004c> anyone know  a link that tells me which is considered slow or normal or veryfast etc ?
[16:48:41 CEST] <Prelude2004c> should be comparable right?
[16:58:50 CEST] <synthecypher> c_14: I'm using this "ffmpeg -i scenarios.mp4 -codec copy -map 0 -f segment -segment_times 30,60,90,120,150,180 out%03d.mp4" the first segment is fine but the others don't play.
[17:02:38 CEST] <c_14> Just tested on my system with an mp4 I had lying around. Works perfectly for me. What version are you running?
[17:19:22 CEST] <synthecypher> c_14: Are all of the segments playable in a player like VLC?
[17:22:38 CEST] <c_14> I used mpv, but yes they're all playable.
[17:22:42 CEST] <Timster> Hey, guys. I have some trouble with concatenating files together - there are files recorded on two different computers, but reencoded with the same software again. Media Info shows, that everything is exactly the same but still, when there is a switch between files of different origins I get choppy video, artifacts and glitches - audio seems just fine. Any idea what this could be?
[17:23:00 CEST] <synthecypher> c_14: must be something wrong with my mp3
[17:23:02 CEST] <synthecypher> c_14: must be something wrong with my mp4*
[17:23:16 CEST] <Timster> Sure, c_14 - one sec.
[17:26:00 CEST] <Timster> http://pastebin.com/4ws0QKR4
[17:27:32 CEST] <Timster> Anything wrong there? Need a moment to change the setup to replicate the error again.
[17:27:32 CEST] <c_14> Are you sure all the input files have the same settings? Can you pastebin an ffprobe of all of them?
[17:27:57 CEST] <Timster> Of course, if you tell me how I use ffprobe on all flv's in a folder
[17:28:00 CEST] <Timster> Well
[17:28:25 CEST] <Timster> The only difference I am aware of is, before the reencode with the software, that one origin uses bilinear filtering, the others lanczos
[17:28:56 CEST] <c_14> for file in dir/*flv; do ffprobe $file; done |& curl -F 'sprunge=<-' http://sprunge.us
[17:29:02 CEST] <c_14> Or redirect into a file.
[17:29:44 CEST] <Timster> Could you tweak it so it does log into a file, please?
[17:29:58 CEST] <c_14> for file in dir/*flv; do ffprobe $file; done > ffprobe_logs
[17:30:06 CEST] <c_14> eh
[17:30:06 CEST] <Timster> Thanks, one sec
[17:30:09 CEST] <c_14> &>
[17:30:23 CEST] <c_14> Though the usefull info is on stdout anyway iirc.
[17:30:35 CEST] <Timster> for file in dir/*flv; do ffprobe $file; done &> ffprobe_logs
[17:30:37 CEST] <Timster> ?
[17:31:34 CEST] <Timster> May I query you about this?
[17:32:06 CEST] <c_14> ye
[17:55:32 CEST] <Prelude2004c> hey, with a quadro card...         Gpu                         : 8 %
[17:55:32 CEST] <Prelude2004c>         Memory                      : 2 %
[17:55:32 CEST] <Prelude2004c>         Encoder                     : 60 %
[17:55:32 CEST] <Prelude2004c>         Decoder                     : 0 %
[17:55:46 CEST] <Prelude2004c> any way to use up the GPU too as well as the nvenc hardware ?
[17:56:02 CEST] <Prelude2004c> wonder if there is some way to increase encoding process to GPU as well as the new hardware accel
[18:01:02 CEST] <BtbN> Prelude2004c, the nvenc encoder is a dedicated piece of hardware on the cards. The GPU is not involved at all.
[18:14:14 CEST] <Scorp1us> Hi all, I was wondering if there was a way to use ffmpeg to divide a video file up inso segments based on I and P frames?
[18:15:00 CEST] <Scorp1us> Example Video= [IPPPPIPPPIPPP] -> [IPPP], [IPPP], [IPPP]
[18:16:43 CEST] <loki_> Prelude2004c: for what you need utilize gpu core?
[18:17:36 CEST] <loki_> Prelude2004c: and in what terminal program you check your video card utilization?
[18:52:25 CEST] <crossf> hi guys, could you please help me to understand this error message: Failed to configure output pad on Parsed_concat_0
[18:53:14 CEST] <spaam> if you show us the commandline you trying to use. we might can help you.
[18:54:56 CEST] <crossf> hmmm, not sure how to extract it from this wrapper c#..., I have tried with native ffmpeg and args and I had other problems, let me paste it for you
[18:57:09 CEST] <crossf> my args: " -i Intro_31bf475c-ba8f-48de-bd59-83197fbcdea5 -i BodyPart_da822a64-2941-4317-b111-1c2c2a5ecd3f -filter_complex 'movie=Intro_31bf475c-ba8f-48de-bd59-83197fbcdea5, scale=1920:1080, setsar=1:1 [v1] ; movie=BodyPart_da822a64-2941-4317-b111-1c2c2a5ecd3f, scale=1920:1080, setsar=1:1 [v2] [v1] [v2] concat [outv]' -map '[outv]' intvid74127b8a-df2f-46ea-b818-302e25d6e574-c5a447a5-f42e-4475-
[18:57:10 CEST] <crossf> 8138-9d26e3786673"
[18:57:39 CEST] <crossf> Stream specifier ':1' in filtergraph description [0:0] [1:1] concat=n=2:v=1:a=1 [v] [a] matches no streams.
[18:58:06 CEST] <crossf> basically I would like to concat videos
[18:59:17 CEST] <crossf> guid are not really readable sorry
[19:00:35 CEST] <crossf> spaam is that help?
[19:06:07 CEST] <debianuser> crossf: It seems you miss a comma between "setsar" and "concat" filters in the "... setsar=1:1 [v2], [v1] [v2] concat ..." part.
[19:06:20 CEST] Action: debianuser isn't sure about movie=... syntax
[19:11:54 CEST] <c_14> semicolon
[19:11:56 CEST] <c_14> not comma
[19:12:15 CEST] <c_14> Like so [v2];[v1][v2]
[19:16:38 CEST] <jehar_work> Heyo
[19:16:55 CEST] <jehar_work> any map_channel gurus around today?
[19:17:04 CEST] <jehar_work> Specifically, prores with 16 channels on a single audio track (fucking blackmagic), pulled out a wav, similarly with 16 channels on one track.
[19:17:13 CEST] <jehar_work> However, any attempt to use map_channel 0.0.1 or whatever to produce a single-channel wav just gives me a blank file, of the correct length.
[19:17:20 CEST] <jehar_work> I know I'm accessing the correct channels, because 0.0.0-0.0.15 are fine, but anything above 15 is rejected
[19:27:29 CEST] <Roberth1990> I see that chrome uses ffmpeg, do they ship it builtin?
[19:28:32 CEST] <jehar_work> Full outputs from my tests: http://pastebin.com/z3HaHQM4
[19:30:11 CEST] <llogan> your ffmpeg build is old
[19:30:58 CEST] <jehar_work> Granted, will compare with a newer build
[19:31:42 CEST] <llogan> http://johnvansickle.com/ffmpeg/
[19:32:43 CEST] <kepstin-laptop> Roberth1990: yes, chrome builds a custom copy of ffmpeg included with the browser. I dunno the details of how they link it.
[19:33:24 CEST] <Roberth1990> kepstin-laptop: well that doesn't matter I just began to wonder why chrome continued to work without restart after going from libav to ffmpeg
[19:50:31 CEST] <debianuser> jehar_work: according to https://trac.ffmpeg.org/wiki/AudioChannelManipulation it should work. Have you tried it on that 16ch .wav file, e.g. `ffmpeg -i all.wav -map_channel 0.0.0 1.wav` ?
[19:55:03 CEST] <kyleogrg> I am converting a dvd vob file (non square pixels) to a 16:9 mkv video.  For -vf I (bascially) have yadif,mcdeint,framestep,crop black border,scale to square pixels,scale to resize.  Is this the right order of video filters?
[19:57:59 CEST] <RobotsOnDrugs> looks reasonable, though i wouldn't bother with squaring the pixels myself
[19:58:43 CEST] <jehar_work> debianuser, llogan was correct. the 2013 build still had those bugs.
[19:59:32 CEST] <kyleogrg> ok
[21:20:40 CEST] <blert> Is it possible to use ffmpeg with Cisco's OpenH264 binaries? I'd like to be able to drop the binary DLL into a folder and have ffmpeg recognize the binary, load it, and use it for decoding and encoding h264 streams. Thanks.
[21:21:23 CEST] <klaxa> i don't think openh264 does encoding
[21:21:50 CEST] <klaxa> oh wait
[21:21:53 CEST] <llogan>  --enable-libopenh264     enable H.264 encoding via OpenH264 [no]
[21:21:57 CEST] <klaxa> yes
[21:22:45 CEST] <blert> So, does it work "out of the box"? Or do I need to rebuild ffmpeg with those flags enabled? (btw, thanks for th quick response!)
[21:23:24 CEST] <klaxa> you will probably have to compile it, since it is disabled by default
[21:23:27 CEST] <blert> (by "out of the box", i mean the pre-built ffmpeg binaries.)
[21:23:34 CEST] <blert> ok. thanks!
[21:24:48 CEST] <blert> this may be OT: does anyone here know what the difference is between ffmpeg and Chrome's "ffmpegsumo" dll?
[21:30:38 CEST] <BtbN> It's every libav* library in one binary iirc
[21:31:08 CEST] <blert> oh. that makes sense. thanks!!
[22:08:18 CEST] <DragonsLord> Hello to everybody ... I have a cropping problem
[22:08:23 CEST] <DragonsLord> May someone help me?
[22:13:59 CEST] <ChocolateArmpits> DragonsLord: I can help
[22:14:09 CEST] <DragonsLord> Thank you so much, ChocolateArmpits
[22:14:36 CEST] <ChocolateArmpits> DragonsLord: I'm all ears
[22:14:50 CEST] <DragonsLord> I am proceeding with Pass-1
[22:15:01 CEST] <DragonsLord> I will pastebin the log
[22:17:03 CEST] <DragonsLord> Report is too big
[22:17:03 CEST] <DragonsLord> :|
[22:17:35 CEST] <ChocolateArmpits> Can you possibly trim it?
[22:18:47 CEST] <DragonsLord> What could I delete?
[22:19:16 CEST] <ChocolateArmpits> What is exactly in your log? Is it full verbose output?
[22:20:05 CEST] <DragonsLord> Here it is > http://pastebin.com/D1FRhPfh <
[22:20:25 CEST] <DragonsLord> The troubles I am experiencing are:
[22:22:17 CEST] <DragonsLord> 1) "Warning: data is not aligned! This can lead to a speedloss" (not visible probably because I set 00:02:00.000, as duration)
[22:23:19 CEST] <DragonsLord> 2) Also increasing NN in crop=w=iw-NN ffmpeg seems to be not able to crop a 2 pixels black line on the left
[22:23:28 CEST] <DragonsLord> am I doing something wrong?
[22:25:40 CEST] <DragonsLord> ChocolateArmpits, should I produce a different report?
[22:25:58 CEST] <ChocolateArmpits> DragonsLord: Nah, if this is enough
[22:26:20 CEST] <DragonsLord> The fact that it's not working is enough for me ;)
[22:29:03 CEST] <ChocolateArmpits> DragonsLord: You should try cropping to multiples of 16
[22:29:19 CEST] <DragonsLord> O_o
[22:29:27 CEST] <DragonsLord> so, is it a bug?
[22:29:31 CEST] <ChocolateArmpits> However does it crash at the warning error ?
[22:30:09 CEST] <DragonsLord> no, it simply seems to keem -56 from the left and -52 from the right
[22:30:31 CEST] <DragonsLord> also if i increase the NN number in crop=w=iw-NN
[22:30:57 CEST] <DragonsLord> c_14, told me that it should be crop from the left + crop from the right true?
[22:31:32 CEST] <ChocolateArmpits> As far as I understand the x y coordinate starts at 0,0 placement
[22:31:56 CEST] <DragonsLord> probably
[22:32:11 CEST] <ChocolateArmpits> so x=52 means the horizontal crop will start at pixel 52
[22:33:04 CEST] <ChocolateArmpits> "The horizontal position, in the input video, of the left edge of the output video. It defaults to (in_w-out_w)/2. This expression is evaluated per-frame. "
[22:33:37 CEST] <ChocolateArmpits> So in your case with x=52 it will crop 52 pixels from left and 56 from right
[22:33:49 CEST] <ChocolateArmpits> leaving 612 pixels horizontal
[22:33:53 CEST] <DragonsLord> true
[22:34:01 CEST] <DragonsLord> exactly
[22:34:37 CEST] <DragonsLord> but I have 2 pixels still present on the left
[22:34:52 CEST] <DragonsLord> so I increase 108 to 110 e.g.
[22:35:00 CEST] <DragonsLord> and the didn't disappear
[22:36:02 CEST] <ChocolateArmpits> hmm
[22:36:51 CEST] <ChocolateArmpits> I'll try testing it
[22:37:02 CEST] <DragonsLord> as you can see
[22:37:11 CEST] <DragonsLord> I am also upscaling the frame
[22:38:00 CEST] <DragonsLord> from 612x572 up to 640x480
[22:38:47 CEST] <ChocolateArmpits> Insert format=yuv420p right before the crop filter
[22:38:55 CEST] <ChocolateArmpits> That should clear up the speedloss error
[22:39:09 CEST] <ChocolateArmpits> wait
[22:39:15 CEST] <ChocolateArmpits> not yet
[22:39:29 CEST] <DragonsLord> ok
[22:40:07 CEST] <DragonsLord> I don't know if it's important ... but i run FFmpeg on WindowsOS :(
[22:40:29 CEST] <ChocolateArmpits> me too, however I didn't have any cropping related problems before
[22:40:48 CEST] <DragonsLord> wait wait, ChocolateArmpits
[22:40:58 CEST] <DragonsLord> I did a test with AVIdemux
[22:41:27 CEST] <DragonsLord> 52 L - 56 R
[22:41:46 CEST] <DragonsLord> the 2 pixels line on the left should not be present
[22:42:08 CEST] <DragonsLord> and I am basically doing the same using swsResize also
[22:44:25 CEST] <DragonsLord> and, as far as I know, AVIdemux should use LibAV, too
[22:44:45 CEST] <DragonsLord> just performed the same test with xmedia recode
[22:44:50 CEST] <DragonsLord> the result is the same
[22:45:09 CEST] <DragonsLord> 52 L - 56 R <- no crop left
[22:45:35 CEST] <DragonsLord> I cannot understand why FFmpeg is acting in this way
[22:45:41 CEST] <DragonsLord> (sorry for my bad English)
[22:47:31 CEST] <ChocolateArmpits> hmm problem cropping 2 pixels on my side
[22:48:32 CEST] <DragonsLord> :)
[22:48:38 CEST] <DragonsLord> I am almost happy
[22:48:43 CEST] <ChocolateArmpits> wait
[22:48:49 CEST] <ChocolateArmpits> sorry, wrote wrong
[22:48:58 CEST] <ChocolateArmpits> *No problems here cropping 2 pixels
[22:49:31 CEST] <DragonsLord> Did you start with my same source
[22:49:38 CEST] <DragonsLord> in terms of size and so on?
[22:49:50 CEST] <ChocolateArmpits> I found a 720x576 image
[22:49:58 CEST] <ChocolateArmpits> and applied same cropping settings
[22:50:17 CEST] <DragonsLord> to an image
[22:50:28 CEST] <DragonsLord> not to a video
[22:50:31 CEST] <llogan> blame your player
[22:50:44 CEST] <DragonsLord> hello, llogan
[22:50:45 CEST] <ChocolateArmpits> I'm testing the cropping with ffplay
[22:51:00 CEST] <ChocolateArmpits> you can paste your vf there
[22:51:06 CEST] <ChocolateArmpits> and see how the output looks
[22:52:57 CEST] <DragonsLord> encoding
[22:53:22 CEST] <DragonsLord> llogan, if you were referring to my player ... this is not the case
[22:54:01 CEST] <DragonsLord> the same video, interlaced, cropped and resized with both AVIdemux and Xmedia Recode is fine with the same player
[22:54:04 CEST] <DragonsLord> (VLC)
[22:55:43 CEST] <DragonsLord> [swscaler @ 000000000033c3a0] Warning: data is not aligned! This can lead to a speedloss
[22:56:09 CEST] <llogan> ignore that
[22:58:20 CEST] <DragonsLord> here is the image:
[22:58:22 CEST] <DragonsLord> http://postimg.org/image/xmh8dofhh/
[22:58:46 CEST] <ChocolateArmpits> You mean that dark line on the left ?
[22:58:57 CEST] <DragonsLord> yep
[22:59:34 CEST] <ChocolateArmpits> Is this the output of the command line you shared in the pastebin?
[23:00:01 CEST] <DragonsLord> yes
[23:00:05 CEST] <DragonsLord> oh
[23:00:07 CEST] <DragonsLord> not really
[23:00:11 CEST] <DragonsLord> that was pass-1
[23:00:23 CEST] <DragonsLord> this is of course the output for pass-2
[23:00:24 CEST] <ChocolateArmpits> ok, but the filter settings are the same right?
[23:00:28 CEST] <DragonsLord> yes
[23:00:31 CEST] <DragonsLord> exactly the same
[23:00:35 CEST] <llogan> you should provide a short input sample
[23:00:39 CEST] <ChocolateArmpits> Did you try setting crop x to 54 ?
[23:01:47 CEST] <DragonsLord> yes, increasing number 108 e.g. to 118
[23:02:31 CEST] <ChocolateArmpits> The first setting ("w") controls how wide the cropped image will be
[23:02:48 CEST] <ChocolateArmpits> It doesn't tell where the cropping should start
[23:03:12 CEST] <ChocolateArmpits> That's what "x" setting does for cropping width
[23:03:29 CEST] <DragonsLord> so probably I misunderstood something
[23:04:08 CEST] <DragonsLord> crop=w=iw-108:h=ih-4:x=52:y=4
[23:04:36 CEST] <ChocolateArmpits> here "w" is 720-108=612, so the width of the output will be 612 pixels
[23:04:53 CEST] <ChocolateArmpits> the width will be cropped starting from" x=52"
[23:05:07 CEST] <DragonsLord> OH
[23:05:08 CEST] <DragonsLord> so
[23:05:19 CEST] <DragonsLord> -52 if from the left of the frame
[23:05:23 CEST] <DragonsLord> or from the right
[23:05:24 CEST] <ChocolateArmpits> exactly
[23:05:26 CEST] <ChocolateArmpits> LEFT
[23:05:40 CEST] <ChocolateArmpits> 0,0 denotes upper left corner of the frame
[23:06:05 CEST] <DragonsLord> however
[23:06:12 CEST] <DragonsLord> I put these setting into AVIdemux
[23:06:16 CEST] <DragonsLord> -52 left
[23:06:25 CEST] <DragonsLord> -56 right
[23:06:30 CEST] <DragonsLord> -4 up
[23:06:34 CEST] <DragonsLord> 0 down
[23:07:59 CEST] <DragonsLord> and ...
[23:08:11 CEST] <DragonsLord> here is the resulting and corresponding frame:
[23:08:12 CEST] <DragonsLord> http://s10.postimg.org/97rdvla3t/vlcsnap_2015_04_10_23h07m16s64.png
[23:08:21 CEST] <DragonsLord> as you can see
[23:08:37 CEST] <DragonsLord> non black vertical line on the left, right?
[23:09:08 CEST] <ChocolateArmpits> will it crops slightly more than in your ffmpeg example
[23:09:12 CEST] <ChocolateArmpits> not just the line
[23:10:09 CEST] <DragonsLord> I don't really understand what you mean
[23:10:11 CEST] <DragonsLord> however
[23:10:17 CEST] <DragonsLord> the basic issue is that
[23:10:26 CEST] <DragonsLord> from my point of view setting are the same
[23:10:31 CEST] <DragonsLord> and the result is different
[23:10:34 CEST] <DragonsLord> furthermore
[23:11:16 CEST] <DragonsLord> I am trying to set crop=w=iw-160
[23:14:43 CEST] <DragonsLord> ChocolateArmpits, it can be that ffmpeg is trying to crop up to the next multiple of 16 so 112?
[23:15:12 CEST] <ChocolateArmpits> That is not possible
[23:15:22 CEST] <DragonsLord> I agree
[23:15:23 CEST] <ChocolateArmpits> It's working with uncompressed data
[23:15:27 CEST] <DragonsLord> I agree
[23:15:31 CEST] <DragonsLord> but, at the end
[23:15:41 CEST] <DragonsLord> this is what it's happening
[23:16:29 CEST] <DragonsLord> ChocolateArmpits!!!
[23:16:48 CEST] <DragonsLord> increasing NN in crop=w=iw-NN
[23:16:57 CEST] <DragonsLord> and leaving x-crop untouched
[23:17:10 CEST] <DragonsLord> ffmpeg should crop more on the left, right?
[23:17:17 CEST] <ChocolateArmpits> on the right
[23:17:27 CEST] <ChocolateArmpits> actually
[23:17:33 CEST] <ChocolateArmpits> it should crop LESS on the right
[23:17:44 CEST] <ChocolateArmpits> as you're expanding the output width leaving less room for the right portion
[23:18:17 CEST] <ChocolateArmpits> left will be left intact
[23:19:31 CEST] <DragonsLord> OK
[23:19:40 CEST] <DragonsLord> this is my misunderstanding
[23:20:32 CEST] <DragonsLord> just to recap
[23:20:44 CEST] <DragonsLord> x= 52 52 is cropping from the left?
[23:21:23 CEST] <ChocolateArmpits> YES
[23:22:05 CEST] <DragonsLord> <c_14> A "simple" formula is w=iw-[total width to cut]:h=ih-[total height to cut]:x=[pixels to cut from the right]:y=[pixels to cut from the top]
[23:22:09 CEST] <ChocolateArmpits> Well to be absolutely correct, that's where the first collumn of pixels will be passed
[23:22:31 CEST] <ChocolateArmpits> so 52nd pixel will be in the output
[23:22:40 CEST] <DragonsLord> c_14, stated x=[pixels to cut from the right]
[23:22:41 CEST] <ChocolateArmpits> while 51st no
[23:22:54 CEST] <c_14> left
[23:23:02 CEST] <ChocolateArmpits> well 0 is a pixel too :)
[23:23:09 CEST] <c_14> I think I corrected myself. I hope I corrected myself.
[23:23:17 CEST] <DragonsLord> hello c_14
[23:23:45 CEST] <DragonsLord> But, in this case I expect to have the same result of AVIdemux
[23:24:34 CEST] <DragonsLord> and furthermore
[23:25:13 CEST] <DragonsLord> setting w=iw-[total width to cut]=iw-160 and leaving x=[pixels to cut from the LEFT] untouched
[23:26:03 CEST] <ChocolateArmpits> Well considering avidemux presents the information differently than the barebones crop filter I think that there could be some different interpretation of the values
[23:26:04 CEST] <DragonsLord> OK
[23:26:06 CEST] <DragonsLord> I got it
[23:26:19 CEST] <DragonsLord> I set -56 left in AVIdemux
[23:26:38 CEST] <DragonsLord> while I set -52 left within ffmpeg
[23:26:39 CEST] <DragonsLord> :|
[23:26:44 CEST] <ChocolateArmpits> see
[23:27:11 CEST] <DragonsLord> sorry for causing you some lost time
[23:27:15 CEST] <DragonsLord> re-encoding
[23:27:26 CEST] <DragonsLord> in italy we would say "prova del 9"
[23:27:29 CEST] <DragonsLord> :-P
[23:27:36 CEST] <ChocolateArmpits> what does that mean?
[23:28:24 CEST] <DragonsLord> uhm
[23:28:37 CEST] <DragonsLord> can you remember when you were at the primary school?
[23:28:47 CEST] <DragonsLord> more or less when you were 5-6 years old?
[23:29:19 CEST] <ChocolateArmpits> that was kindergarten for me at that age
[23:29:23 CEST] <ChocolateArmpits> some parts
[23:29:43 CEST] <DragonsLord> the teacher was trying to teahc you additions
[23:30:11 CEST] <DragonsLord> OK
[23:30:16 CEST] <DragonsLord> in english it should be
[23:30:17 CEST] <DragonsLord> casting out nines
[23:30:33 CEST] <DragonsLord> Casting out nines is a sanity test to ensure that hand computations of sums, differences, products, and quotients of integers are correct. By looking at the digital roots of the inputs and outputs, the casting-out-nines method can help one check arithmetic calculations. The method is so simple that most schoolchildren can apply it.
[23:30:40 CEST] <DragonsLord> Wikipedia helps a lot ;)
[23:31:24 CEST] <DragonsLord> can you understand what I meant?
[23:31:29 CEST] <ChocolateArmpits> sorta
[23:31:34 CEST] <DragonsLord> hi Kolizer
[23:31:44 CEST] <DragonsLord> sorta?
[23:31:52 CEST] <Kolizer> hello, is there a function executing ignoring colors like chrome key?
[23:34:10 CEST] <BtbN> No chroma key.
[23:35:40 CEST] <DragonsLord> ChocolateArmpits, test completed successfully :)))
[23:35:53 CEST] <DragonsLord> crop was performed perfectly
[23:36:05 CEST] <DragonsLord> thank you very much for your time and your patience
[23:36:16 CEST] <DragonsLord> the same also "applies" for c_14 and llogan
[23:36:16 CEST] <ChocolateArmpits> DragonsLord: no probs
[23:36:19 CEST] <DragonsLord> :-P
[23:36:42 CEST] <DragonsLord> ChocolateArmpits, YES ... there was a problem: myself XD
[23:36:58 CEST] <Kolizer> That is, I won't be able to overlay an image with a green background and make the background transparent?
[23:39:24 CEST] <ChocolateArmpits> Kolizer: not with libav filters it seems. If you are working with files then try the Keying node in Blender or Chroma Key node in Natron, if you need a scripted solution maybe Avisynth, if streaming from an outside source then CasparCG suite.
[23:39:50 CEST] <ChocolateArmpits> all these are open source
[23:40:59 CEST] <DragonsLord> ChocolateArmpits, I am going to smoke ... when I will be back could I share another dubt with you?
[23:41:03 CEST] <Kolizer> I want tools ffmpeg
[23:41:09 CEST] <BtbN> Chroma/Color key is one major usefull feature that's just plain missing from ffmpeg.
[23:41:59 CEST] <ChocolateArmpits> Avisynth scripts can be fed into ffmpeg
[23:42:27 CEST] <Kolizer> cool =)
[23:44:31 CEST] <Kolizer> Thanks for the great answers )
[23:46:35 CEST] <Kolizer> I hope someday will add this feature
[23:49:27 CEST] <BtbN> It's not too complex, just nobody has done it.
[23:49:34 CEST] <Kolizer> I just stream using ffmpeg and I need real time to do the transparency
[23:49:56 CEST] <BtbN> Why not use something like OBS for that?
[23:50:14 CEST] <Kolizer> Oh no)
[23:50:32 CEST] <BtbN> It's way better suited for interactive live streams.
[23:50:38 CEST] <ChocolateArmpits> obs has a bad keyer
[23:50:51 CEST] <BtbN> a bad what?
[23:50:53 CEST] <Kolizer> There is also no such feature
[23:51:06 CEST] <BtbN> No such feature? OBS definitely does support chroma key.
[23:51:54 CEST] <ChocolateArmpits> well it's called color key
[23:51:59 CEST] <BtbN> No it's not.
[23:52:07 CEST] <BtbN> It's called color-key for image source.
[23:52:13 CEST] <BtbN> Video-Capture-Sources have chroma key.
[23:52:34 CEST] <ChocolateArmpits> monitor capture has color key
[23:52:44 CEST] <ChocolateArmpits> as does window capture
[23:52:44 CEST] <BtbN> Yes, because it makes much more sense for that.
[23:53:01 CEST] <BtbN> You get bit exact color captures from it.
[23:53:06 CEST] <BtbN> So color key is what you usualy want.
[23:59:22 CEST] <DragonsLord> Guys, if I have multiple videos and the duration is the same
[23:59:31 CEST] <Kolizer> I would have written the functionality of the color key for ffmpeg, if I was a programmer =)
[23:59:46 CEST] <DragonsLord> if I set the 2-pass medium bitrate the same
[00:00:00 CEST] <DragonsLord> I should obtain files with the same size, right?
[00:00:00 CEST] --- Sat Apr 11 2015


More information about the Ffmpeg-devel-irc mailing list