[Ffmpeg-devel-irc] ffmpeg.log.20150710

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


[00:00:17 CEST] <klaxa> what settings did you use?
[00:02:21 CEST] <techtopia> ffmpeg -y -i test.ts -sws_flags spline -sn -vf yadif=0:0,crop=1912:1076:8:2 -s 1280x720 -c:v libx264 -preset slow -level 4.1 -crf 21 -r 25 -x264opts colormatrix=bt709 -map 0:0 -map 0:1 -acodec copy test_x264.mkv
[00:02:22 CEST] <techtopia> and
[00:02:29 CEST] <techtopia> ffmpeg -y -i test.ts -sws_flags spline -sn -vf yadif=0:0,crop=1912:1076:8:2 -s 1280x720 -c:v libx265 -preset slow -level 4.1 -crf 21 -r 25 -x264opts colormatrix=bt709 -map 0:0 -map 0:1 -acodec copy test_x265.mkv
[00:04:15 CEST] <klaxa> the thing is, crf are not equal between x264 and x265, passing -x264opts to libx265 seems weird, and presets also do not exactly do the same things in x264 and x265
[00:05:31 CEST] <techtopia> ahh yeah good point heh
[00:05:40 CEST] <techtopia> i just modded my normal encode line to do a quick test
[00:05:51 CEST] <techtopia> will do another test
[00:06:11 CEST] <techtopia> and need to find some documentation on the crf diffrences
[00:07:01 CEST] <c_14> techtopia: 8fps on libx265 for slow is pretty good (depending on the input of course)
[00:07:14 CEST] <c_14> A pure black input gave me 30fps and mandelbrot is giving me 4ish
[00:07:45 CEST] <techtopia> this was low motion reality tv
[00:07:51 CEST] <techtopia> mainly just people sat around talking
[00:08:12 CEST] <c_14> Also, don't manually set the level (unless you need to force it lower than x264/x265 sets it)
[00:08:32 CEST] <c_14> It'll pick the level itself
[00:08:37 CEST] <techtopia> ok :) will omit that too
[00:09:06 CEST] <c_14> You can probably also get rid of -r 25 (unless your input isn't 25 fps and you want to force it to 25 fps)
[00:10:21 CEST] <c_14> Oh, and you don't need -sn (since you're mapping explicitly)
[00:11:03 CEST] <techtopia> :D glad i posted my lines
[00:11:13 CEST] <techtopia> lots of fixes i can do
[00:53:24 CEST] <clevelandrocks> I'm trying to decode a a .mp4 into a .tiff sequence without altering/converting its colorspace.  I run this command: http://pastebin.com/HbnPhEi8 and .tiff files come out BUT the .tiff files that ffmpeg produces from this command are not vieable by anything I can find.
[00:54:16 CEST] <clevelandrocks> *not viewable
[00:57:51 CEST] <clevelandrocks> when I try to open the resulting .tiff files , programs report the .tiff files as being  "corrupted or too large"
[00:57:51 CEST] <pzich> do you need yuvj422p? that might be causing it
[00:58:06 CEST] <clevelandrocks> I have tried other color formats too
[00:58:15 CEST] <clevelandrocks> yuv420p produces the same result
[00:59:05 CEST] <pzich> does it work with other image formats, or exported as a movie?
[00:59:14 CEST] <clevelandrocks> the humorous part is that ffmpeg says this: "Incompatible pixel format 'yuvj422p' for codec 'tiff', auto-selecting format 'yuv422p' "
[00:59:44 CEST] <clevelandrocks> if i choose rgb24 the .tiff files that come out are viewable
[01:00:12 CEST] <clevelandrocks> but I don't want to change damage/change the color of the source
[01:02:47 CEST] <clevelandrocks> so when the 'auto-selected' yuv422p is used, the .tiff files that result are still no good
[01:03:09 CEST] <clevelandrocks> but when 'rgb32' is used, the .tiff files that result are at least openable
[01:03:23 CEST] <clevelandrocks> *'rgb24'
[01:11:06 CEST] <Kaedenn> Can I embed XMP data in .avi files?
[01:17:51 CEST] <Kaedenn> Actually, how would I go about adding a license to an AVI file?
[01:19:44 CEST] <pzich> with metadata? http://stackoverflow.com/questions/9464617/retrieving-and-saving-media-metadata-using-ffmpeg
[01:24:59 CEST] <clevelandrocks> @pzich : does that make sense?  any ideas on how to do this?
[01:25:39 CEST] <pzich> nope, I don't think images are generally stored in yuv formats though, so maybe you need to use an rgb one?
[01:26:44 CEST] <clevelandrocks> is there a way to convert from YUV to RGB that is lossless?
[01:27:04 CEST] <CoJaBo_> clevelandrocks: Only if you somehow have infinite precision..
[01:27:30 CEST] <CoJaBo_> So heres another fun and probably nigh impossible task I've been handed- is there any way to determine if a video file is from a camera vs. other source?
[01:28:02 CEST] <CoJaBo_> It seems like metadata could be useful in that, but I'm guessing it's gonna vary far too much to be useful...
[01:28:02 CEST] <pzich> it probably depends on if the camera adds metadata to the file
[01:28:11 CEST] <pzich> yep, probably too varied
[01:32:04 CEST] <CoJaBo_> gah.
[01:34:17 CEST] <CoJaBo_> ..oddly, it doesn't appear ANY of the files have any kind of metadata at all. :/
[01:38:35 CEST] <clevelandrocks> CoJaBo_: the source has a pixel format of yuv420p using smpte2048 color ranges.  thus I fear doing any pixel format conversion squashing those color ranges.  is this fear unfounded?
[01:53:55 CEST] <clevelandrocks> what's even stranger is that the "unreadable" .tiff files (the ones output in yuv420p colorspace) can be fed into ffmpeg to create a viewable .yuv file
[02:29:37 CEST] <DHE> Do you think I could do multiple transcodes of the same video at different resolutions/bitrates from the same first pass encoding stats? (libx264) or would that be a bad idea
[02:35:04 CEST] <durandal_1707> It's possible but slow
[02:39:07 CEST] <DHE> slow?
[05:32:48 CEST] <agorecki> Hi guys. How can I take three to five videos and combine them into a single video?
[05:40:25 CEST] <DHE> https://trac.ffmpeg.org/wiki/Concatenate
[09:05:08 CEST] <Taniey> ffmpeg filter tblend ,how can I make video have a fade in or out effect?
[09:07:20 CEST] <Taniey> anybody hear?
[09:18:05 CEST] <JEEBsv> where was the if() thing in filter chains documented again? somehow it evades my google-fu
[09:21:25 CEST] <JEEBsv> I guess true/false value, value if true, value if false
[09:36:58 CEST] <JEEBsv> ok, I'm trying to crop only if iw is WIDTH and ow is HEIGHT
[09:37:31 CEST] <JEEBsv> crop=out_w=iw:out_h=if(if(eq(iw,WIDTH),eq(ih,HEIGHT),0),ih-32,ih):y=if(if(eq(iw,WIDTH),eq(ih,HEIGHT),0),32,0):keep_aspect=1
[09:38:00 CEST] <JEEBsv> s/ow/ih/
[09:38:22 CEST] <JEEBsv> but it seems to barf at the if within an if
[09:39:25 CEST] <JEEBsv> or wait... it might just be a shell parsing issue...
[09:40:42 CEST] <JEEBsv> yup
[09:41:07 CEST] <JEEBsv> I had it surrounded by ''s but I still had to escape all of the ,s
[09:41:12 CEST] <JEEBsv> (within the ifs
[09:47:27 CEST] <durandal_1707> Taniey: fade filter?
[09:48:11 CEST] <Taniey> blend filter
[09:49:24 CEST] <Taniey> or tblend filter
[09:52:22 CEST] <Taniey> I use this commend " tblend=all_expr='A*(10-T)/10+B*0' " but the video change into green
[09:57:36 CEST] <Taniey> <durandal_1707>:how can i improve my expression ?
[09:59:27 CEST] <durandal_1707> Taniey: add format=gbrp, before tblend...
[10:00:45 CEST] <durandal_1707> you want to fade to black?
[10:01:05 CEST] <Taniey> en
[10:01:29 CEST] <Taniey> yes
[10:01:32 CEST] <durandal_1707> another filter to do this is fade filter
[10:03:52 CEST] <Taniey> en ,thanks
[10:08:58 CEST] <reza> quit
[11:36:13 CEST] <zhanshan> hi
[11:36:27 CEST] <zhanshan> what's the different between -c:a pcm_s24le and -acodec?
[11:36:31 CEST] <zhanshan> can I use both?
[11:37:46 CEST] <zhanshan> I mean can I use either -c:a  pcm_s24le OR -acodec pcm_s24le?
[11:46:34 CEST] <BtbN> acodec is the deprecated syntax
[12:40:33 CEST] <benbro1> can ffmpeg convert pcap file captured from RTP stream to audio/video?
[12:45:32 CEST] <hardikj> hi guys, I just installed ffmpeg for mac(yos) and on running it I get avfoundation not found, what am i missing?
[12:45:44 CEST] <hardikj> when I install it using brew it works fine
[12:46:03 CEST] <hardikj> but I downloaded it from here https://www.ffmpeg.org/download.html#build-mac
[12:46:19 CEST] <hardikj> from here http://ffmpegmac.net/ actually
[13:17:18 CEST] <svvitch> I transcode UDP streams to RTMP with this params "-loglevel verbose -i "udp://238.0.0.7:32000?timeout=90000000&fifo_size=100&overrun_nonfatal=1" -progress "http://192.168.1.50:8024/live/on_tc_progress?live_id=14570b4e-c0ce-11e1-ae5c-0025902b2f0c&app=admin-vip" -flags +global_header -map 0:v -c:v flv -qscale 1 -map 0:a -c:a aac -b:a 187k -strict -2 -ar 44100 -ac 1 -f flv "rtmp://192.168.1.133/admin-vip/14570b4e-c0ce-11e1-ae5c-0025902b2f0c""
[13:19:07 CEST] <svvitch> and every ffmpeg process continuously increase memory usage until for ex. 10 ffmpeg processes use 32 GB of RAM
[13:20:28 CEST] <svvitch> is there a way to deal with this ?
[13:41:20 CEST] <retard> you could always get more ram
[13:44:34 CEST] <svvitch> retard: seriously
[13:50:54 CEST] <DHE> sounds like you don't have enough CPU power to do the transcode in realtime and the UDP receiver is buffering all to hell
[13:51:31 CEST] <svvitch> load average is 7.29 for 16 core cpu
[13:51:40 CEST] <DHE> hmm.. doesn't quite explain 32 GIGS though..
[13:51:56 CEST] <svvitch> I think is enough :)
[13:52:36 CEST] <svvitch> DHE: yes, I tried a lot of params :( nothing helps
[13:52:38 CEST] <DHE> I'm doing something similar but x264 and output to a named pipe...
[13:53:52 CEST] <svvitch> DHE: x264 increase cpu usage
[13:54:21 CEST] <DHE> yaeh, I don't know what the requirements of .flv are or what sort of resolution you're using or what sort of CPU...
[13:55:14 CEST] <svvitch> the input stream is in H263, audio AAC so it's ok for flv
[13:55:28 CEST] <DHE> I just doubechecked the buffer sizes for UDP so that's not it...
[13:58:32 CEST] <svvitch> I decrease the fifo_size, but the ram usage anyway begin to grow, I don't know how much, now I'm watching
[14:01:59 CEST] <DHE> personally I'd use the default or a little bit more..
[15:06:40 CEST] <Vonor> hi, trying to build ffmpeg 2.7.1 on a gentoo system. freetype is installed but the configure script complains about freetype2 not found. according to this page a temp fix is to create a subdirectory and link the freetype.h. but it didn't work for me: https://lists.ffmpeg.org/pipermail/ffmpeg-devel/2013-November/151406.html
[15:09:04 CEST] <Vonor> checked the config.log and see that it seems to have issues compiling the freetype test code: http://pastebin.com/TzTUxLig
[15:43:54 CEST] <relaxed> Vonor: are you using an ebuild?
[17:10:07 CEST] <well0ne> Hello.
[17:11:03 CEST] <well0ne> I'm using a bunch of mpegts files (h264+mp3+same res.) to stream to an rtmp stream. with mpegts container and the same codecs, i'm able to stream all files throu , without reencoding thats really nice.
[17:11:33 CEST] <well0ne> i'm searching for an php/c programmer which knows the ffmpeg code. i want to emulate that function purely in php.
[17:11:41 CEST] <well0ne> can someone help? i would pay
[17:17:50 CEST] <well0ne> i'm not a specialist at this, but i think i need an librtmp implementation in php, and a implementation for reading out files
[17:19:30 CEST] <durandal_1707> why?
[17:20:09 CEST] <well0ne> i want to stream directly with php
[17:31:51 CEST] <rcombs> maybe don't do that
[17:33:14 CEST] <Eduardo_1> i don't think its a good idea, it will be porting ffmpeg and all libs to php
[17:46:17 CEST] <Daan_> Dear, I'm searching for a ffmpeg command which can do this: 1 img at the background, 1 image as overlay and on the background a music track, who can help me? :) Greetz, Daan_
[17:55:00 CEST] Last message repeated 2 time(s).
[18:19:29 CEST] <onefix> Has anyone tried doing a distributed encoding project for FFMPEG?  I was thinking that with all of these new "cheap" devices coming on the market (like the $9 "CHIP"), you could use like 5 of these to make a fast, cluster for encoding video...
[18:24:01 CEST] <onefix> I also came up with a really simple way of making it work...essentially all you need is a NFS share that is the same on all hosts, put a file in there with a job file and then the simplest way to do it would be to split the process among the devices in the cluster using timecodes ... all of the coding could probably be done in a simple scripting language like perl....
[18:25:05 CEST] <BtbN> It's not realy worth the effort
[18:25:18 CEST] <BtbN> usualy you distribute it by putting multiple files on multiple machines
[18:25:49 CEST] <BtbN> Encoding just isn't that challenging of a task that it's worth to distribute the encoding of a single stream
[18:26:54 CEST] <onefix> Really?  But, being able to encode a 2 hour video in 30 minutes versus 2 hours would be a great achievement...
[18:27:51 CEST] <Vonor> relaxed, nope. trying to build myself
[18:28:48 CEST] <onefix> Especially since the hardware encoders for H.264 are not nearly as good as x264...
[18:29:19 CEST] <Vonor> relaxed, the ebuild version (same as the source i downloaded), works fine. both with freetype and x256 - both fail when manually configuring the source
[18:35:22 CEST] <onefix> Found it.  It's called DVE and it gives an ~2x speed increase with 3 machines
[18:36:31 CEST] <BtbN> on a somewhat modern CPU you can encode a 2 hours video in way less than 2 hours.
[18:39:55 CEST] <onefix> BtbN: It really depends on what you are encoding, the size of the original file, and the presets you are using
[18:40:09 CEST] <Vonor> could it be the freetype version that causes the issue?
[18:40:22 CEST] <onefix> If you are targeting size and quality, you're going to pay in terms of encoding speed
[18:40:29 CEST] <Vonor> using freetype 2.6
[18:40:37 CEST] <BtbN> Of course you can crank the settings up untill it's slower than realtime.
[18:40:40 CEST] <BtbN> Or you can use h265
[18:41:23 CEST] <onefix> I would, but h264 is still the only format supported on most settop boxes ... rumor is the Roku 4 will have H.265 support
[18:42:03 CEST] <BtbN> h265 was an example of how to make it slower-than-realtime.
[18:42:29 CEST] <Daan_>  Dear, I'm searching for a ffmpeg command which can do this: 1 img at the background, 1 image as overlay and on the background a music track, who can help me? :) Greetz, Daan_
[18:51:00 CEST] Last message repeated 1 time(s).
[19:01:00 CEST] Last message repeated 4 time(s).
[20:40:14 CEST] <Kaedenn> I have a wav file I'm adding to my project but I want to insert it at some offset not at the beginning of the video.
[20:40:26 CEST] <Kaedenn> What argument do I use for that?
[20:41:07 CEST] <c_14> The easiest thing is to probably concatenate n seconds of silence to the front of the audio file.
[20:41:22 CEST] <klaxa> there is also -itsoffset
[20:41:36 CEST] <Kaedenn> I'm trying to align a visualization with the music it's visualizing and it's off by like 2 tenths of a second.
[20:41:45 CEST] <c_14> Yeah, but I'm not sure there are many players that will be happy when the audio starts late.
[20:41:58 CEST] <c_14> Oh, you should probably be able to use itsoffset for that
[20:42:03 CEST] <klaxa> ah
[21:37:12 CEST] <haasn> Can libavcodec be used to display images *as* they're loading? (eg. say they're loading from a network source)
[22:04:17 CEST] <durandal_1707> haasn: generally no, but perhaps you could loop or something
[22:05:09 CEST] <durandal_1707> there is draw_horiz_band or something like that though
[22:50:01 CEST] <pothibo> Hi, I was wondering if it was possible to read a stream, continuously, and the at the instance notice, start recording that stream so that the recording has no glitch (waiting for keyframe, etc.) Does that ring a bell?
[22:51:06 CEST] <pothibo> Because if I only use the command line, there is a warm-up period where all the libs get loaded into ram and then waiting on the first keyframe of the stream
[00:00:00 CEST] --- Sat Jul 11 2015


More information about the Ffmpeg-devel-irc mailing list