[Ffmpeg-devel-irc] ffmpeg.log.20170724

burek burek021 at gmail.com
Tue Jul 25 03:05:01 EEST 2017


[00:49:32 CEST] <Brian510> Hello,  I'm looking for some help with creating text overlays, from data files
[00:50:18 CEST] <Brian510> I have a list of still images and a list of images.data files that have the same name as the image file
[00:51:50 CEST] <Brian510> I already have a script that comes in and creates symbolic links of the images and renames the links in an appropriate format to be read in by -f image2 -i img%07d.jpg
[00:52:50 CEST] <Brian510> Id like to use the -reload option to feed new text to draw
[00:53:22 CEST] <Brian510> But I can't figure out how to trigger an update the textfile when the frame names changes
[00:56:24 CEST] <Brian510> Id like when  img0000001.jpg is being processed, that I can trigger an shell script to read img0000001.jpg.data and output its content to the text file of the draw text filter
[00:56:43 CEST] <Brian510> https://thepasteb.in/p/0ghJPxPm2PEh5
[00:57:06 CEST] <Brian510> how would I trigger a script at the processing of each frame ?
[01:04:00 CEST] Last message repeated 1 time(s).
[01:12:00 CEST] Last message repeated 1 time(s).
[01:15:28 CEST] <DHE> I might suggest making it into a pre-rendered subtitle file and then burn the subtitles into the video
[01:17:36 CEST] <Brian510> I thought about that, but,  subtitle seems to be driven by time
[01:17:39 CEST] <Brian510> and not frame
[01:18:16 CEST] <Brian510> data to image sync is important
[01:19:21 CEST] <Brian510> I thought about just writinthe data to the image but I would prefer not to make a copy of all the source images
[01:26:13 CEST] <bobo11> hey guys, is there a way to set all frequencies below 10khz to -6 db ?
[01:27:03 CEST] <cryptodechange> Not too sure if I've lost too much detail on my encode (look at line work, shadows and clouds) - https://imgur.com/a/Gf1b8
[01:45:33 CEST] <thebombzen> cryptodechange: the lineart is a bit soft. consider a small negative smartblur (negative sharpens)
[01:45:40 CEST] <thebombzen> keep in mind that sharpeners amplify noise
[02:16:57 CEST] <cryptodechange> thebombzen, thanks for the advice, use nlmeans at the moment, set rc=1 as it took away too much color
[02:22:18 CEST] <cryptodechange> will try the smartblur
[02:32:28 CEST] <techbomber> if you had 100k us dollars and you had to go all-in on ONE company stock, what company stock would you buy?
[02:34:32 CEST] <cryptodechange> trying smartblur=ls=-0.5:lt=10,nlmeans
[02:37:53 CEST] <furq> techbomber: what did you do with the other $150k
[02:38:11 CEST] <techbomber> furq nothing
[02:38:20 CEST] <furq> well then how are you going all in
[02:39:47 CEST] <techbomber> i am not techinically
[02:41:30 CEST] <furq> well if i were you, i'd put big money into uber
[02:41:33 CEST] <furq> those guys are on the up and up
[02:45:00 CEST] <techbomber> furq  uber is not public
[02:45:49 CEST] <furq> right but if you take 500 ubers you get a free calendar
[02:59:01 CEST] <cryptodechange> thebombzen the bitrate is super high still :(
[03:58:12 CEST] <cryptodechange> Figured it out, I think it looks much better, retained more detail too
[03:58:14 CEST] <cryptodechange> https://imgur.com/a/Gf1b8
[03:58:34 CEST] <cryptodechange> Would be nice to drop the bitrate a bit more though
[04:05:25 CEST] <furq> you might want to try some of the other denoise filters
[04:22:36 CEST] <cryptodechange> any recommendations furq? about to try hqdn3d
[04:24:45 CEST] <furq> hqdn3d with stronger temporal than spatial
[04:40:03 CEST] <cryptodechange> that's at default right?
[04:40:13 CEST] <cryptodechange> I'll try it without changing anything
[05:18:48 CEST] <cryptodechange> Is there any articles/comparisons floating about for nlmeans vs. hqdn3d?
[06:54:40 CEST] <zumba_addict> am I right that -preset ultrafast has better image quality than -preset slow?
[07:56:53 CEST] <kerio> no
[07:56:58 CEST] <kerio> not at all
[07:57:24 CEST] <kerio> couldn't be further from the truth
[10:42:11 CEST] <JEEB> I don't remember, was there a way to only set a specific module to do debug logging?
[11:39:46 CEST] <durandal_1707>  /whois sdggfddfgg
[13:01:20 CEST] <kerio> how do i output mjpeg without a specific framerate?
[13:01:37 CEST] <kerio> ie following the actual framerate of the input, rather than the nominal framerate
[13:25:20 CEST] <bencc> I saw a commit for fmp4 support in hls
[13:25:42 CEST] <bencc> but only for multiple files? not possible to have a single file with byte range?
[13:26:00 CEST] <bencc> https://github.com/FFmpeg/FFmpeg/commit/274bd1670b9c85e24f37d08f9043dd82e0a28c67
[13:28:26 CEST] <JEEB> I think the guy didn't care enough to implement it himself so far?
[13:34:51 CEST] <kerio> how do i generate a http mjpeg stream with ffmpeg
[13:34:54 CEST] <kerio> do i have to ffserver
[13:34:57 CEST] <kerio> i don't want to ffserver
[13:36:45 CEST] <bencc> JEEB: https://patchwork.ffmpeg.org/patch/4221/
[13:38:47 CEST] <JEEB> well yea, technically "I'm busy I might implement it later"
[13:39:12 CEST] <JEEB> it's still a "patch welcome" if you need it right now
[13:39:29 CEST] <bencc> JEEB: I have no idea how to help :)
[13:40:13 CEST] <bencc> I thought that he did 99% of the work
[13:40:47 CEST] <kerio> oh, mpjpeg
[13:40:50 CEST] <kerio> why does ffmpeg do EVERYTHING
[13:41:08 CEST] <kerio> except output mjpeg to v4l2loopback
[15:49:33 CEST] <cryptodechange> I wonder if my encode caused ringing, or it's just preserving some of the colour around the linework
[15:49:34 CEST] <cryptodechange> https://imgur.com/a/Xekf3
[15:49:53 CEST] <cryptodechange> Still a bit of noise/grain, but I'm happy where the bitrate is at
[16:34:41 CEST] <kepstin> cryptodechange: hmm, with that much denoising, i'd be worried about banding problems.
[16:36:22 CEST] <cryptodechange> will try and find a frame with some gradiet
[16:36:26 CEST] <cryptodechange> gradient
[16:45:30 CEST] <cryptodechange> kepstin: probably not the best sample but https://imgur.com/a/Cepqx
[16:46:54 CEST] <kepstin> hmm. the irregularities from denoising the grain make the banding less objectionable at least.
[16:48:33 CEST] <kepstin> pity the nlmeans filter doesn't support >8bit depth
[16:49:04 CEST] <cryptodechange> mm, the raw was 8bit anywho
[16:50:10 CEST] <kepstin> sure, but denoising the grain (which effectively acts like dithering) on an 8bit source can give you an effective higher bit depth.
[16:50:31 CEST] <kerio> why would ffmpeg not listen to sigterm anymore
[16:50:35 CEST] <kepstin> assuming you're ok with actually encoding at 10bit, your target audience has decoders which can handle it :)
[16:55:04 CEST] <cryptodechange> Mm, encoding this for plex so wanted to try and make it suitable for 'direct play', unless the bitrate limit is lower
[16:56:22 CEST] <Loop> Hello I have question regarding ffmpeg, is there a version of ffmpeg that doesnt write in the encoded file the stuff on the beginning of the file like
[16:56:53 CEST] <Loop>  ftypisom<bh:00><bh:00><bh:02><bh:00>isomiso2avc1mp41<bh:00><bh:00><bh:00><bh:08>free<bh:00><bh:83>v?mdat<bh:00><bh:00><bh:02><bh:8c><bh:06><bh:05><bh:ff><bh:ff><bh:88><bh:dc>E<bh:e9><bh:bd><bh:e6><bh:d9>H<bh:b7><bh:96>,<bh:d8> <bh:d9>#<bh:ee><bh:ef>x264 - core 119 - H.264/MPEG-4 AVC codec - Copyleft 2003-2011 - http://www.videolan.org/x264.html - options:
[16:57:12 CEST] <furq> x264 writes that, and no
[16:57:25 CEST] <Loop> Is there some clean version of x264
[16:57:32 CEST] <furq> no
[16:58:25 CEST] <Loop> Could I encode something with ffmpeg but in h264 or in some clean encoder?
[16:58:29 CEST] <kepstin> Loop: x264 does have a commercial license option, you're welcome to contact them and ask...
[16:58:41 CEST] <furq> what does "clean" mean here
[16:59:18 CEST] <Loop> Clean like free of tags metatags or info like videolan
[16:59:28 CEST] <furq> why do you need that
[16:59:43 CEST] Action: kepstin notes that x264 is also known to put other types of watermarking into the encoded video, it's not hard to identify x264-encoded video
[17:00:54 CEST] <Loop> hmm what type do you have some more info
[17:01:34 CEST] <Loop> currently I'm doing the manual hex editing and I would like to have some tool to clen the encodes
[17:01:51 CEST] <kepstin> ... why are you so interested in making it so people can't identify the encoder/settings used?
[17:02:13 CEST] <Loop> Its safer
[17:02:22 CEST] <furq> what
[17:02:43 CEST] <kepstin> for what, saving you from being sued for license violation or something?
[17:03:32 CEST] <Loop> for every possible situation like better to be safe than sorry
[17:03:41 CEST] <Loop> For my personal collection
[17:05:58 CEST] <kepstin> tbh, having an x264-video with metadata removed is a stronger identifying fingerprint than using a popular public build and leaving the info in.
[17:06:08 CEST] <kepstin> if that's what you're worried about
[17:10:25 CEST] <Loop> I'm writing some metadata but its not easy I need to calculate the blank space that I need to leave
[17:12:15 CEST] <Loop> I had idea to rewrite into the source of ffmpeg but I'm too lazy to do that, I'm certain that someone did that already
[17:13:26 CEST] <kepstin> what you're trying to do doesn't make sense. customizing ffmpeg/x264 to change metadata is only going to result in having a metadata pattern that uniquely identifies your files.
[17:16:24 CEST] <Threads> Loop what are you trying todo exactly
[17:21:24 CEST] <Loop> Exactly I don't want any metadata and encoder info like videolan lav free etc just clean mp4
[17:21:47 CEST] <Loop> maybe to add some metadata info
[17:21:53 CEST] <Loop> just maybe
[17:23:10 CEST] <Loop> I cleaned most of the stuff and I see that file is playable but it needs 5-10 seconds more to play
[17:24:01 CEST] <Loop> my initial question is, is there some custom ffmpeg version with some custom x264 that has such options
[17:24:30 CEST] <Loop> I can hex edit evrything manually but it would be easier if there is a tool
[17:26:12 CEST] <Loop> I see that I can inject into the youtube faked file with edited metadata and codec data this was just for a test, I'm just testing what work and what doesn't
[17:34:12 CEST] <DHE> Loop: bitexact mode will help with that, eliminating version-specific information. but it doesn't eliminate it entirely.
[17:36:38 CEST] <kepstin> Loop: just keep in mind the difference between "this video file was done with the same encoder and settings as a million other files" and "huh, strange, someone put a lot of work into stripping metadata from this file in a particular way that's uniquely identifiable"
[17:56:07 CEST] <Loop> @kepstin I never used bitexact option what arguments to use
[17:56:35 CEST] <Loop> bitexact
[17:56:35 CEST] <Loop>     Only write platform-, build- and time-independent data. (except (I)DCT). This ensures that file and data checksums are reproducible and match between platforms. Its primary use is for regression testing.
[18:06:43 CEST] <Loop> @kepstin  also I see only few versions of ffmpeg have bitexact most of the versions that I'm using personally doesn't have bitexact at all
[18:55:07 CEST] <superbia> good evening sirs
[18:56:36 CEST] <Gidian> I've been getting "Your videos will process faster if you encode into a streamable file format." when trying to upload my videos to youtube.
[18:56:54 CEST] <Gidian> ffmpeg -i <folder/image.png> -i <folder/audio.flac> -vf format=yuv420p <Output.mp4>
[18:57:28 CEST] <Gidian> The filesize is fantastic, and the video/audio quality is great too, but for some reason youtube hates it.
[18:57:29 CEST] <superbia> does anyone have a script that auto-converts all videos downloaded with youtube-dl to MP3 and later deletes original files?
[18:58:20 CEST] <furq> Gidian: -movflags +faststart
[19:01:33 CEST] <Gidian> Ty furq
[20:20:53 CEST] <ioan> hi. Is there a way to make this command work? ffmpeg -i rtsp://ip:554/axis-media/media.amp -vf "sendcmd=f=send.cmd,drawtext=fontsize=20:fontcolor=white:x=100:y=100:text='Hello World'" -vcodec copy -codec:a aac  -f flv  -r 30 "rtmp://a.rtmp.youtube.com/live2/<secret>"
[20:21:21 CEST] <ioan> I get the error: Filtering and streamcopy cannot be used together.
[20:25:17 CEST] <Blubberbub> what happens if you remove -vcodec copy?
[20:26:44 CEST] <kepstin> ioan: yeah, that error kind of says the problem... if you're copying the encoded video data, you can't modify it, and filtering it to add some text on top is a modification
[20:27:09 CEST] <kepstin> ioan: so you have to re-encode the video, replace the -vcodec with an actual encoder, e.g. libx264.
[20:28:38 CEST] <ioan> kepstin, I understand, thanks for explaining. I guess it won't be text overlay then, the video comes from an IP camera and I'm live streaming from a Raspberry Pi, so I don't want to make it harder for the Pi.
[20:30:02 CEST] <furq> ioan: https://support.google.com/youtube/answer/3068031?hl=en-GB
[20:30:04 CEST] <furq> might be an option
[20:32:38 CEST] <ioan> furq, thank you, I'll read about it to see if I can use that
[20:42:39 CEST] <Blubberbub> What i learned today: one cannot expect filters that use a time or duration to work properly, when one does not initialize the time_base in the stream.
[21:12:27 CEST] <durandal_170> Blubberbub: which filters?
[21:12:38 CEST] <Blubberbub> afade - for example
[21:13:30 CEST] <Blubberbub> totally my mistake, though.
[21:15:15 CEST] <MoSal> configure is creating ffbuild/config.mak files with NULs in them. Is anyone seeing this?
[21:21:19 CEST] <defanor> i don't
[21:27:27 CEST] <cryptodechange> compiled a 10-bit x264 binary and mv'd it to /usr/local/bin/x264-10bit
[21:27:56 CEST] <cryptodechange> trying to figure out how to compile a second ffmpeg binary with x264-10bit
[21:28:30 CEST] <defanor> are all the functions in libavformat supposed to be thread-safe?
[21:29:57 CEST] <JEEB> cryptodechange: FFmpeg will not use your x264 command line app :P
[21:30:40 CEST] <JEEB> you build, install into your prefix and when building FFmpeg set PKG_CONFIG_PATH=/your/prefix/lib/pkgconfig where there now should be x264.pc
[21:30:44 CEST] <cryptodechange> ;_;. so make install elsewhere, and specify the libx264 directory
[21:31:00 CEST] <cryptodechange> yes thankyou
[21:31:15 CEST] <JEEB> you should be able to test with PKG_CONFIG_PATH=/your/prefix/lib/pkgconfig pkg-config --libs x264
[21:31:22 CEST] <JEEB> you should be able to see if it catches the one in your prefix
[21:31:32 CEST] <JEEB> if yes, then configure script [hould find it, too
[21:31:34 CEST] <cryptodechange> Now to determine the best prefix for x264-10bit
[21:31:45 CEST] <cryptodechange> could just do /opt/x264-10bit
[21:32:01 CEST] <JEEB> whatever. I mean, I usually have a random prefix in my home dir if the thing's not supposed to be system global
[21:32:24 CEST] <JEEB> also there's a patch in the x264 ML that makes libx264 support both 8 and 10 bit coding
[21:32:34 CEST] <JEEB> koda did the great work to make it all work :3
[21:32:53 CEST] <JEEB> although I bet the libx264 wrapper will have to be changed when that change goes in :D
[21:32:57 CEST] <JEEB> in FFmpeg that is
[21:33:18 CEST] <cryptodechange> Will need to keep an eye out in the changelog for that
[21:33:21 CEST] <furq> cryptodechange: just use the static ffmpeg builds if you want one with 10-bit x264
[21:33:36 CEST] <furq> https://www.johnvansickle.com/ffmpeg/
[21:34:10 CEST] <cryptodechange> oh nice
[21:35:34 CEST] <cryptodechange> ty
[21:37:53 CEST] <iranen_> when using x265 what is better for tune setting PSNR or SSIM? (or select none tune?)
[21:38:26 CEST] <cryptodechange> In terms of 8bit > 10bit encodes, really, am I correct in assuming I should only take note of the crf?
[21:38:32 CEST] <cryptodechange> As that scales differently?
[21:43:51 CEST] <JEEB> cryptodechange: in x264 the CRF scale has been moved a bit so similar values should be OK (but I did some small adjustments in the end, although that often is because of just difference of content)
[21:44:04 CEST] <JEEB> basically crf 0 is no longer lossless and the values go to negative
[21:44:06 CEST] <furq> iranen_: psnr and ssim are for benchmarking. don't ever use them for anything else
[21:47:41 CEST] <cryptodechange> JEEB: oh I didn't know that, I read that the 8bit range was/is 0-51 and 10bit is 0-63
[21:48:13 CEST] <cryptodechange> but didn't find much documentation on the range comparisons yet
[21:48:53 CEST] <JEEB> cryptodechange: yea that's the quantizer range and what's the CRF range as well. but then it was later rolled in >8bit
[21:49:12 CEST] <JEEB> (so that it went lower than zero)
[21:49:28 CEST] <JEEB> that way the user wouldn't have to adjust the CRF value too much between bit depths
[21:50:14 CEST] <cryptodechange> Oh, they just applied an offset
[21:50:37 CEST] <JEEB> yea
[21:50:45 CEST] <cryptodechange> So -12 would be lossless
[21:50:46 CEST] <cryptodechange> makes sense
[21:51:50 CEST] <cryptodechange> Though I wouldve thought doing CRF*(63/51) would've made more sense
[21:52:15 CEST] <cryptodechange> So an 8-bit CRF of 18 would become 22
[21:54:44 CEST] Action: kepstin notes that if you actually want lossless, '-qp 0' will do that in all x264 bitnesses.
[21:58:43 CEST] <cryptodechange> Doing a 10bit encode test now, I upped the smartblur strength from -0.35 to -0.5 and dropped the CRF from 17 to 16
[21:58:54 CEST] <cryptodechange> upped/dropped
[22:00:02 CEST] <TheWild> hello
[22:00:30 CEST] <TheWild> running in virtual console: "ffmpeg http://rmfstream2.interia.pl/80SDISCO". How I can make it display metadata too?
[22:01:25 CEST] <TheWild> *ffplay
[22:23:57 CEST] <Loop> @kepstin could you tell me which ffmpeg version has bitexact which is working and could you tell me the arguments for the command because I'm getting ffmpeg: missing argument for option 'bitexact'
[22:36:54 CEST] <atomnuker> -flags +bitexact
[22:40:48 CEST] <c_14> Wasn't it -fflags?
[22:42:36 CEST] <Loop> Unrecognized option 'flags'
[22:43:03 CEST] <c_14> try -fflags +bitexact
[22:43:05 CEST] <Loop> [NULL @ 00dd41f0] [Eval @ 0028fb48] Undefined constant or missing '(' in 'bitexact'
[22:43:06 CEST] <Loop> [NULL @ 00dd41f0] Unable to parse option value "bitexact"
[22:43:06 CEST] <Loop> Invalid value '+bitexact' for option 'fflags'
[22:43:19 CEST] <c_14> hmm
[22:43:46 CEST] <Loop> I tried 14 versions of ffmpeg ony 3 version have bitexact at all
[22:43:49 CEST] <c_14> should definitely work (though apparently I always set both)
[22:44:04 CEST] <c_14> Loop: what versions?
[22:44:14 CEST] <c_14> (Please don't say 1.x
[22:44:15 CEST] <c_14> )
[22:44:34 CEST] <c_14> It should definitely be in anything > 2.7
[22:44:36 CEST] <c_14> and that's
[22:44:39 CEST] <c_14> years old now
[22:45:02 CEST] <Loop> FFmpeg version SVN-r26292, Copyright (c) 2000-2011 the FFmpeg developers
[22:45:02 CEST] <Loop>   built on Jan 24 2011 20:49:50 with gcc 4.4.5
[22:45:12 CEST] <c_14> yeeeeaaaah
[22:45:14 CEST] <c_14> don't do that
[22:45:18 CEST] <c_14> pleease don't do that
[22:45:31 CEST] <Loop> I cant find the version newer that have working bitexact
[22:45:36 CEST] <c_14> That's not supported anymore
[22:45:46 CEST] <c_14> What distro/operating system?
[22:45:53 CEST] <c_14> http://johnvansickle.com/ffmpeg/ <- for linux
[22:46:04 CEST] <Loop> Right know on vmware windows7
[22:46:11 CEST] <Loop> now
[22:46:27 CEST] <Loop> It must be on win7
[22:46:27 CEST] <c_14> https://ffmpeg.zeranoe.com/builds/
[22:46:42 CEST] <c_14> get the git version (or if you must release 3.3.2)
[22:46:54 CEST] <Loop> Ahh I tried that version and it doesnt work
[22:47:02 CEST] <c_14> doesn't work how?
[22:47:03 CEST] <Loop> let me choose some dev build
[22:48:15 CEST] <Loop> It says
[22:48:34 CEST] <Loop> Unrecognized option 'bitexact'.
[22:48:34 CEST] <Loop> Error splitting the argument list: Option not found
[22:48:42 CEST] <Threads> man your going along way just to either fake headers/versions or strip them
[22:49:00 CEST] <c_14> Loop: what's your commandline
[22:49:04 CEST] <Loop> yes you are right
[22:49:07 CEST] <c_14> sounds like something's eating an option where it shouldn't
[22:50:04 CEST] <Loop> give me sec to try dev build
[22:53:56 CEST] <atomnuker> c_14, Loop: -flags +bitexact works on git master
[22:54:59 CEST] <c_14> atomnuker: it's an option for both, apparently
[22:55:07 CEST] <c_14> not sure if they're aliased or if there's a difference though
[22:57:34 CEST] <kerio> one is for the format, the other is for the encoder?
[22:57:52 CEST] <Loop> ok seems shared version has working bitexact
[22:58:14 CEST] <Loop> I will try to see what results I'm getting
[22:58:40 CEST] <Loop> static definatlly soesnt work
[22:58:48 CEST] <Loop> static definitely doesnt work
[23:16:06 CEST] <Loop> @c_14 it encoded with -flags +bitexact and I see that it didnt write only at the end the writer info  lav version but all metadata from the x264 is present
[23:17:32 CEST] <kyleogrg> I want to join six videos into one, but I only want to take a section of each video.  I'd like to do this all in one command.  Reencoding (as opposed to concatenation) is okay.
[23:17:45 CEST] <kyleogrg> What's the easiest way/
[23:17:46 CEST] <kyleogrg> ?
[23:20:42 CEST] <kepstin> kyleogrg: list 6 inputs on the command line, with the '-ss' and '-t' input options, use -filter_complex with the concat filter.
[23:22:47 CEST] <kepstin> if you don't need exact seekpoints (starting on the nearest keyframe is ok), consider using the 'concat' format, which takes a playlist file as described here: https://www.ffmpeg.org/ffmpeg-formats.html#concat-1 (you can specify an inpoint and outpoint per-file)
[23:23:26 CEST] <kyleogrg> kepstin: with your first suggestion, does -ss and -t follow or precede every input?
[23:23:27 CEST] <kepstin> that's probably gonna be a bit iffy, tho, I'm not sure how well it'll work.
[23:23:38 CEST] <furq> you don't use -ss or -t
[23:23:49 CEST] <furq> you add the seek points to the concat list
[23:24:03 CEST] <kepstin> kyleogrg: input options precede the input that they're associated with
[23:24:14 CEST] <furq> oh nvm i missed that line
[23:24:30 CEST] <kyleogrg> kepstin: like -ss 00:00:10 -t 00:00:30 -i input1.mp4?
[23:24:45 CEST] <kyleogrg> furq: yeah, we're talking about the first method
[23:24:48 CEST] <kepstin> kyleogrg: that'll grabe 30 seconds of video starting 10 seconds in, yes.
[23:25:00 CEST] <kyleogrg> ok, imma go try this
[23:27:23 CEST] <kyleogrg> oh, and i need to scale some of them so they're all the same res
[23:27:56 CEST] <kyleogrg> can i put -vf "scale=X:X:X:X" after each input?
[23:28:18 CEST] <kepstin> kyleogrg: no, you'll have to put it all together into the single -filter_complex string
[23:29:54 CEST] <kyleogrg> kepstin: okay
[23:30:12 CEST] <kepstin> something like '[0:v]scale=X:X:X:X[0_scaled]; (repeat for each video) ; [0_scaled][1_scaled]...[6_scaled]concat=n=6:v=1:a=0'
[23:30:19 CEST] <kepstin> it's more complicated if you're doing audio too
[23:30:30 CEST] <kyleogrg> I need the audio.
[23:31:02 CEST] <kyleogrg> at this point it might be better for me to trim and scale them separately and  have a lossless intermediary
[23:31:07 CEST] <kepstin> then the concat part will look like [0_scaled][0:a][1_scaled][1:a]...concat=n=6:v=1:a=1 I think.
[23:31:35 CEST] <kyleogrg> okay
[23:32:24 CEST] <kepstin> if it gets too long on the command line, save the filter string in a text file, and use the -filter_complex_script <filename> option to load it :)
[23:32:43 CEST] <kyleogrg> kepstin: ah, never tried that method
[23:33:27 CEST] <kepstin> (if you were doing >6 videos, I probably would have recommended not using input files on the ffmpeg command line, but rather using the "movie" filter inside a filter_complex script file to load them)
[23:34:39 CEST] <kyleogrg> ok, i'll see which option seems simpler
[00:00:00 CEST] --- Tue Jul 25 2017


More information about the Ffmpeg-devel-irc mailing list