[Ffmpeg-devel-irc] ffmpeg.log.20160522

Mon May 23 02:05:01 CEST 2016

[01:58:39 CEST] <kynlem> so, i've got a mov file (h264 video; pcm audio) where each 12th frame is a key frame
[01:59:49 CEST] <kynlem> if i reencode just the video part and cut 60.06 seconds (1440 frames) from it, then everything's alright
[02:00:38 CEST] <kynlem> but if i reencode audio as well, and then merge the produced 60.06-second chunks, there's a glitch in sound *and* video where the pieces meet. thoughts?
[04:53:36 CEST] <hendry> I know one can make a slideshow like so: ffmpeg -framerate 1 -pattern_type glob -i '*.jpg' -c:v libx264 out.mp4
[04:53:50 CEST] <hendry> but I want the filenames of the jpgs in the video itself... how do I achieve that?
[07:48:35 CEST] <GimmeSpace> hello sir, how to watermarking videos with ffmpeg for specific frames??
[07:49:36 CEST] <GimmeSpace> like it just up to 10 sec and dissapear for 10 sec and up again for another 10 sec like so... thanks.
[07:54:48 CEST] <GimmeSpace> hello guys...
[08:13:11 CEST] <twistin> Hello. Can anyone help me use ffmpeg to combine a DashVideo (~489MB) and Dashaudio(~88MB) file fast? The command I am using now is `ffmpeg -i 1.dashAudio -i 2.dashVideo -shortest output.mp4` ... The length of the video is 3:12:29 (hr:min:sec) and same length audio. It has been nearly 2 hours and ffmpeg is still running.
[10:53:49 CEST] <cyzz> hi. its a pleasure to play round with your work
[10:55:32 CEST] <cyzz> twitch over ffmpeg at nvenc but..... :D it works
[13:01:44 CEST] <pandaologist> having some trouble getting ffplay to recognise a rawvideo stream piped from ffmpeg
[13:02:33 CEST] <pandaologist> it is NTSC yuv422p, can anyone advise what the ffplay parameters should be?
[13:03:08 CEST] <pandaologist> all I get at the moment, no matter what I try is
[13:03:11 CEST] <pandaologist> "Invalid data found when processing input"
[13:04:01 CEST] <pandaologist> it used to just work until I updated ffmpeg
[13:17:59 CEST] <pandaologist> no worries, seems to be working now
[13:18:07 CEST] <pandaologist> not even sure what i'm doing differently
[17:46:07 CEST] <lilibox> hi
[17:48:06 CEST] <ChocolateArmpits> hello
[17:52:27 CEST] <lilibox> i would like to know what -f switch is stand for, i have got some complicated .php which calls ffmpeg and uses switch -f sb16le
[17:52:53 CEST] <furq> format
[17:53:10 CEST] <lilibox> i wanna make static ffmpeg, so which lilbraries add to my build will be able handle sb16le
[17:53:14 CEST] <furq> you use it when the format can't be inferred from the filename
[17:53:35 CEST] <furq> s16le is pcm, it's built-in
[17:53:57 CEST] <lilibox> interesting
[17:59:07 CEST] <lilibox> and what is main difference between ffmpeg and ffmpeg_flv?
[17:59:55 CEST] <furq> i've never heard of ffmpeg_flv and neither has google
[18:00:02 CEST] <furq> so i'm going to assume someone's renamed an ffmpeg binary
[18:08:29 CEST] <DHE> sounds like someone's custom build with a narrow purpose
[18:08:45 CEST] <DHE> I have a few of those, but I keep them private. confuses people otherwise
[18:09:02 CEST] <lilibox> yes, it's Autodesk's chaos
[18:19:47 CEST] <ploop> I'm converting videos with the "-target ntsc-dvd" option, but every dvd creating utility I've tried still tries to reconvert them again which destroys the quality. are there any utilities that won't do such a thing if the video is dvd-compliant?
[18:32:21 CEST] <c_14> dvdauthor should do it, no?
[18:35:04 CEST] <grrk-bzzt> Hello
[18:36:18 CEST] <grrk-bzzt> I'm trying to convert one lossless video file to another lossless video codec without color loss
[18:36:35 CEST] <c_14> use the same pixel format
[18:36:52 CEST] <grrk-bzzt> But the pixel format is pal8
[18:37:31 CEST] <c_14> or something that can accurately transform like zscale or that new filter, whatever it was called
[18:38:15 CEST] <c_14> The pixel format of what? The input? Output?
[18:39:18 CEST] <grrk-bzzt> The input file uses the zmbv video codec and the pal8 pixel format
[18:40:34 CEST] <c_14> Extract a frame from the input (using -c copy or something) convert it to the other codec/pixel format and convert it back and check if they're the same
[18:40:54 CEST] <c_14> Doesn't have to be a frame, can be a section or something as well
[18:41:38 CEST] <grrk-bzzt> I'm using ffmpeg -i a.avi -i b.avi -filter_complex "blend=all_mode=difference" so far to see if there's any differences
[18:41:48 CEST] <c_14> See if ffmpeg does the right thing by default, or if you need to specify a different destination pixel format (higher bit depth or something) or use zscale or the colorspace filter
[18:42:19 CEST] <c_14> use -fflags +bitexact and just check the sha hash, or use the framecrc muxer or something
[18:42:30 CEST] <grrk-bzzt> Alright
[18:42:40 CEST] <c_14> there's also the hash/framehash muxers and md5/framemd5
[18:43:45 CEST] <grrk-bzzt> Thank you for your advices c_14
[19:12:51 CEST] <grrk-bzzt> Seems impossible to obtain the same hash
[19:13:10 CEST] <c_14> With what command[s]?
[19:13:38 CEST] <grrk-bzzt> c_14, I'm using -flags +bitexact -sws_flags +accurate_rnd+bitexact -fflags +bitexact
[19:14:12 CEST] <c_14> what output codec/pixel format?
[19:15:02 CEST] <grrk-bzzt> I've tried rawvideo output codec, and so far I've tried with rgb24, pal8 and yuv420p
[19:24:07 CEST] <c_14> zmbv (pal8) -> ffv1 (rgba) -> zmbv (pal8)
[19:24:09 CEST] <c_14> seems to work here
[19:24:15 CEST] <c_14> Checked with -f framemd5 and -f md5
[19:25:00 CEST] <grrk-bzzt> OK, I'll check that c_14
[19:25:18 CEST] <c_14> -f hash -hash sha512 also matches
[19:25:30 CEST] <grrk-bzzt> Did your commande included the same -flags -sws_flags and -fflags I've used?
[19:25:38 CEST] <c_14> I didn't use the sws_flags
[19:25:55 CEST] <c_14> they probably shouldn't hurt though
[19:26:07 CEST] <grrk-bzzt> just ffmpeg -i a.avi -c:v ffv1 out.avi ?
[19:28:14 CEST] <c_14> -pix_fmt bgra
[19:28:23 CEST] <c_14> it chooses bgr0 by default
[19:28:34 CEST] <c_14> which might and might not work since pal8 has alpha and bgr0 doesn't
[19:28:39 CEST] <c_14> depends on if your source has alpha
[19:28:41 CEST] <grrk-bzzt> c_14, you've said rgba earlier
[19:28:52 CEST] <grrk-bzzt> No, my source doesn't have alpha
[19:28:55 CEST] <c_14> I meant bgra
[19:29:04 CEST] <c_14> ffv1 doesn't even support rgba
[19:29:08 CEST] <grrk-bzzt> Alright
[19:29:26 CEST] <grrk-bzzt> So it'll ignore alpha info?
[19:31:11 CEST] <c_14> if you use bgr0, yes
[19:31:39 CEST] <c_14> Hashes seem to still match on the video I tested on (which doesn't have alpha) so it should be fine
[19:38:26 CEST] <grrk-bzzt> Hum, it doesn't here
[19:42:46 CEST] <grrk-bzzt> ffv1 (bgra) -> zmbv (pal8) doesn't work, and it actually degrades the video quality
[19:43:05 CEST] <grrk-bzzt> Hum, I'll try with the *flags
[19:44:05 CEST] <c_14> https://pb.c-14.de/t/kng.1rTFdm <- if you run that, does it output a diff?
[19:44:17 CEST] <c_14> (I'm trying to check if it's something with your video or with your ffmpeg version)
[19:47:52 CEST] <grrk-bzzt> c_14, it's doesn't work: Requested output format 'hash' is not a suitable output format in: Invalid argument
[19:48:04 CEST] <c_14> Your version of ffmpeg is probably rather old
[19:48:07 CEST] <c_14> What version are you running?
[19:48:22 CEST] <grrk-bzzt> ffmpeg version 3.0.2
[19:48:37 CEST] <c_14> Ah, that particular change might be only in git?
[19:49:26 CEST] <c_14> try that https://pb.c-14.de/t/kng.TMm73v
[19:49:42 CEST] <c_14> md5 should be good enough
[19:51:33 CEST] <grrk-bzzt> Alright? What should I see?
[19:51:52 CEST] <c_14> if you don't see anything it's good.
[19:51:57 CEST] <c_14> If it didn't match, you'd see a diff at the end
[19:52:03 CEST] <grrk-bzzt> So it's good
[19:52:09 CEST] <c_14> you can check in/out and in~ and out~ manually if you want to make sure
[19:52:17 CEST] <grrk-bzzt> I did
[19:52:19 CEST] <c_14> So it's something with your video probably
[19:52:26 CEST] <grrk-bzzt> alright
[19:52:32 CEST] <grrk-bzzt> I can send you the sample
[19:52:35 CEST] <c_14> maybe it does have alpha?
[19:52:46 CEST] <grrk-bzzt> Just a moment
[20:02:11 CEST] <c_14> The output looks like it has had dithering applied to it
[20:02:25 CEST] <grrk-bzzt> Indeed
[20:07:16 CEST] <c_14> the ffv1 output isn't dithered though
[20:22:04 CEST] <grrk-bzzt> sorry c_14 I had a network problem
[20:22:19 CEST] <grrk-bzzt> I did not receive anything past "the ffv1 output isn't dithered though"
[20:22:32 CEST] <c_14> That was the last message I sent.
[20:25:05 CEST] <c_14> I can't seem to turn off the dithering during the zbmv conversion, even -sws_dither none doesn't help
[20:25:44 CEST] <c_14> *zmbv
[20:26:43 CEST] <grrk-bzzt> Is it a ffmpeg bug?
[20:27:58 CEST] <c_14> Either a bug, or a deficiency in the zmbv encoder
[20:30:39 CEST] <grrk-bzzt> Alright
[20:30:50 CEST] <grrk-bzzt> Good to know i've spotted something then
[20:31:59 CEST] <grrk-bzzt> To be honnest, I wanted to run some test to see which lossless video codecs to see which one offered the best compression ratio for retrogaming recording
[20:32:16 CEST] <grrk-bzzt> But it seems more complicated than that
[20:34:54 CEST] <c_14> Well, if it doesn't have to be 100% perfectly lossless (minor loss due to chroma conversion/dithering) it's a lot easier
[20:36:12 CEST] <JEEB> well, ffv1 and libx264 both support RGB coding
[20:36:29 CEST] <JEEB> so if you want 1:1 lossless to the original capture that's possible
[20:42:46 CEST] <Prelude2004c> hey guys.. good day
[20:43:55 CEST] <Prelude2004c> i am using ffmpeg to convert somethign into HLS.. and i am using nfs mount to basically drop the files in a remote server.. i may be having some troulbe with that nfs mount for some reason ( not sure why ).. i would like to use webdav with some queue system.. anyone have any idea what i can use to send the files via webdav and then if lets say it doens't respond, it can queue files for say 30 seconds.. and at some point when the destinat
[20:43:55 CEST] <Prelude2004c> ion comes bcak online it dumps them all at the same time ?
[20:53:40 CEST] <BtbN> why not just run ffmpeg on the target machine?
[20:55:46 CEST] <DHE> if it's a live stream then you have a user experience problem anyway. otherwise I would just save locally and rsync either periodically or when it's done
[21:24:01 CEST] <grrk-bzzt> <JEEB> so if you want 1:1 lossless to the original capture that's possible ’ How?
[21:26:30 CEST] <JEEB> grrk-bzzt: pick ffv1 or libx264_rgb and encode with it. ffv1 is lossless by definition, libx264 has to have -q:v 0 for lossless
[21:28:21 CEST] <grrk-bzzt> Which pix_fmt should I use?
[21:29:34 CEST] <JEEB> grrk-bzzt: by default you should get the input pix_fmt
[21:29:41 CEST] <JEEB> see ffmpeg cli's output
[21:43:20 CEST] <erraunt> Hi. I was reading some of the jpeg2000 code and TODO with a seen references to ISO papers, but the document costs 175 CHF, are there any other ways to read it?
[21:48:47 CEST] <JEEB> google is generally your friend with such things
[22:12:13 CEST] <Prelude2004c> i can't sync.. its a live stream
[22:12:16 CEST] <Prelude2004c> has to be done in realtime
[22:13:16 CEST] <furq> what's wrong with nfs
[22:14:10 CEST] <Fyr> guys, is -filter:v a new option?
[22:14:17 CEST] <furq> no
[22:14:18 CEST] <Fyr> while -vf - old?
[22:14:21 CEST] <furq> oh
[22:14:27 CEST] <furq> well yeah it's what you should use in place of -vf
[22:14:45 CEST] <furq> i assume the old switches will be deprecated at some point but probably not any time soon
[22:15:04 CEST] <Fyr> so, filter:v, so as c:a and other syntax with ":" was introduced recently?
[22:15:11 CEST] <furq> comparatively recently
[22:15:16 CEST] <Fyr> ok
[22:17:51 CEST] Action: penguin42 is a bit confused - I'm converting from an rtsp stream off a cctv camera that I believe is h264 to mp4 (using -preset veryfast -f segment) and I'm seeing with perf top   that ffmpeg is spending a lot of time in idct - why? If it's just repackaging why does it need to idct it?
[22:18:17 CEST] <furq> because you're reencoding it
[22:18:36 CEST] <furq> use -c:v copy if you just want to remux it
[22:18:41 CEST] <furq> and also -preset is meaningless if you're remuxing
[22:19:25 CEST] Action: penguin42 tries -cv copy  -  - would it be otherwise the same line?
[22:19:46 CEST] <furq> that depends on what the line is
[22:19:48 CEST] <furq> but probably
[22:20:53 CEST] <penguin42> http://pastebin.com/VTjWSH5A is my current line - lets try that option
[22:24:45 CEST] <penguin42> ooh, promising
[22:34:47 CEST] <penguin42> furq: OK, that looks pretty good, except for some reason it's changed where segment has decided to segment the video
[22:35:01 CEST] <furq> segments have to start on a keyframe
[22:35:26 CEST] <furq> obviously if you're just remuxing then you have no control over keyframe placement
[22:36:00 CEST] <penguin42> furq: Right, my problem here is the opposite - it seems to be segmenting every second, where as with my world without the -c:v copy  it was segmenting about once every minute
[22:36:55 CEST] <furq> you can set -segment_time to override that
[22:37:00 CEST] <penguin42> ah thanks
[22:54:09 CEST] <penguin42> furq: Thank you - that seems to be doing the trick; and I now have two pretty much idle cores :-)
[22:54:40 CEST] <penguin42> poor little ULV celerons dont have very fast brains
[23:52:55 CEST] <tlvb> hello, I'm examining a script for converting a list of mts files into another format. The script creates a fifo and cats the mts files into that, in the background, the video camera may create multiple mts files even when filming a single event, and apparently cat-ing them together is entirely reasonable, however, I wonder if feeding this fifo file to ffmpeg, as the script does, is any different from using
[23:52:58 CEST] <tlvb> the concat demuxer?
[23:56:29 CEST] <tlvb> the original script can be found here: http://ubuntuforums.org/showthread.php?t=1630612 , and mine, although changed to convert to a different format as well, here: http://pastebin.com/JyYWZUdW
[00:00:00 CEST] --- Mon May 23 2016