[Ffmpeg-devel-irc] ffmpeg.log.20120925

burek burek021 at gmail.com
Wed Sep 26 02:05:01 CEST 2012


[05:06] <funyun> hi. what do i do if i don't want to encode video from a source file? just audio? i tried "-map 0:1". but i get this error "Number of stream maps must match number of output streams"
[06:03] <ChannelZ> funyun: try -nv
[10:45] <denny--> Morning.  When the docs say timestamps can be given in hh:mm:ss[.xxx] format, is .xxx fractional seconds or frames?
[10:45] <ChannelZ> seconds I believe
[10:46] <ChannelZ> else it'd be :ff
[10:46] <denny--> okay, that's what I had assumed given the decimal point
[10:46] Action: denny-- nods
[10:46] <denny--> it just suddenly occurred to me that if I'd assumed wrong. that might explain my inaccurate cuts  :)
[10:46] <ChannelZ> Yeah sans the colons, it's always seconds elsewhere
[11:15] <denny--> if I want to change my process so that it does a reencode without losing quality, what options do I want instead of "-vcodec copy -acodec copy"?
[11:16] <denny--> I have -sameq in the example someone has sent me here, but that's not in the docs, so I'm guessing it's obselete or something
[11:16] <JEEB> sameq is not "same quality"
[11:16] <JEEB> just to note
[11:16] <JEEB> it's "copy quants"
[11:16] <denny--> right  :)
[11:16] <JEEB> it was renamed to be less derpy name-wise
[11:16] <JEEB> but the old setting still works
[11:16] <denny--> my ignorance knows no bounds.  So that's the same_quant thing now?  yeah
[11:16] <JEEB> yup, the old name still works for some reason
[11:16] <denny--> okay, so ignoring that, what options should I be reading about?
[11:17] <JEEB> do you mean lossless encoding or "visually lossless to you" btw?
[11:17] <denny--> preferably lossless - I'm working for a big-name broadcaster
[11:18] <denny--> they'll probably be upset if I mangle their video file  :)
[11:18] <JEEB> does it have to be openable via anything "pro"?
[11:18] <denny--> that's half the reason I was trying to use the copy codecs, that plus time is a factor, but at this point they're willing to sacrifice time to gain accuracy on these cuts
[11:19] <denny--> it has to be imported into various media management tools, so probably yeah
[11:20] <JEEB> very few things unfortunately take in lossless stuff, although I guess if you could get those people to install ut video decoders it could work :S I made an ut video encoder so you could use -vcodec utvideo for that, probably will want to output avi
[11:21] <JEEB> (ut video has decoder/encoder components for both DS/VFW (windows) and QT (mac)
[11:31] <denny--> they've been having some trouble importing the content into something called Carbon Coder after I process it sometimes, we're not sure yet if that's my fault or not.
[11:32] <denny--> I'm way out of my depth with all the video side of this, I'm just trying to automate a workflow - I write Perl for a living  :)
[11:32] Action: denny-- experiments with -vf select to see how long it takes
[11:57] <c10ud> Hello, i was looking into ffserver and filters and i read it's not possible to pass a filter through ffserver. I wondered if it was a quite easy task adding support for filters but after digging a bit into ffmpeg_opt code it is still unclear to me how filter options are applied to the stream. I guess it's the starting point for such implementation (parsing in ffserver and passing to ffmpeg shouldn't be hard from what i've seen). Hints?
[11:58] <ubitux> c10ud: please join the -devel channel to discuss this
[11:58] <ubitux> you'll likely get more help
[11:58] <c10ud> ubitux, ok thanks
[12:33] <denny--> heh
[12:33] <denny--> well, that almost melted this crappy laptop  :D
[12:34] <denny--> also the resulting file seems to have b0rked timing data - Windows thinks it's 0:00:00 long and WMP won't skip ahead in it.  Looks like it selected the whole file though, which isn't what I wanted.
[12:34] <denny--> overall, not a great success  :)
[12:36] <burek> just checking things.. :)
[12:37] <burek> yup, it works now
[12:39] <denny--> :)
[12:44] <burek> denny--, how exactly do you cut your streams
[12:44] <burek> do you have any pastebin or something
[12:45] <burek> c10ud, could you use: ffmpeg -i INPUT -vf filter_name=filter_options http://localhost:8090
[12:46] <burek> i.e. to use ffmpeg and do the filtering stuff and then pass all that to ffserver
[12:47] <c10ud> burek, i don't think that works (?)
[12:47] <burek> c10ud why not, did you try it?
[12:48] <c10ud> iirc, yes, but i'll try it again
[12:48] <c10ud> with the fresh compiled ffmpeg
[12:48] <denny--> http://pastebin.com/cGq2BHD3 is the cutting process - it extracts everything except the censored bits, then stitches them back together
[12:48] <denny--> then rewraps it to fix the metadata
[12:49] <denny--> but those individual sections, the -edit**.ts files, are not quite right - the start/end points are slightly wrong
[12:49] <denny--> so the stitch and join is fairly irrelevant
[12:50] <burek> denny--, what is the point of -g option
[12:50] <burek> if you are just copying your streams?
[12:51] <burek> also, if you use -ss and -t using -vcodec copy
[12:51] <burek> it will not cut your video just anywhere you want
[12:51] <denny--> oh, is that in there?  something I tried withotu really understanding it  :)  it's not there any more
[12:51] <denny--> okay, why not, and more importantly, how can I fix it?
[12:51] <c10ud> burek, does not work, i remembered correctly
[12:52] <burek> well, if you take a zip file and cut it at preciselly 100 bytes, why doesn't winzip recognize it anymore :)
[12:52] <burek> denny--, you can cut on some boundaries of encoded chunks/blocks when you are using a copy method
[12:53] <burek> if you want to cut it precisely on x:y:z.qqq then you need to re-encode your video, I guess
[12:53] Action: denny-- nods
[12:53] <denny--> so I was just tryign to do that this morning, but I seem to be getting the whole file instead
[12:53] <denny--> with broken metadata
[12:53] <denny--> let me patch that cdoe back in and show you the command,, maybe you can tell me what's wrong with it
[12:54] <burek> ok
[12:54] <burek> c10ud, what exactly doesn't work
[12:54] <burek> does it spit out any errors
[12:54] <burek> or it segfaults
[12:54] <burek> or what
[12:54] <denny--> DEBUG - FFMPEG: C:/ffmpeg.exe -loglevel error -i C:/PerlData/Temp/T5159900177A-t emp1.ts -vf select='gte(n,0)*lte(n,24000)' -y C:/PerlData/Temp/T5159900177A-edit 01.ts
[12:55] <c10ud> burek, i think they're just ignored since if you give the output to ffserver the filters are set to null
[12:55] <c10ud> i will try to fix this
[12:55] <burek> denny--, you didn't specify neither -vcodec nor -acodec, which means ffmpeg will choose some defaults
[12:55] <burek> which you most probably won't like :)
[12:55] <denny--> so that ran for about an hour, made my laptop very warm :) and I ended up with a file that seemed to be the whole file although I couldn't tell for sure, as the metadata was broken and it thought it was 0 seconds long, so WMP wouldn't let me skip through it
[12:55] <denny--> it didn't look terrible actually - right size at least - but yeah, broken in some way
[12:55] <burek> c10ud if that is so, you can always pipe 2 ffmpegs
[12:56] <c10ud> burek, or fix it :p
[12:56] <burek> ffmpeg -i input -vf ... -f nut - | ffmpeg -f nut -i - http://localhost:8090
[12:56] <burek> something like that
[12:56] <denny--> I'm taking in a broadcast quality HD mp4 file - what codecs shojld I be using to keep it as nice as possible?
[12:56] <burek> c10ud, well yes, but if you have time to wait :)
[12:57] <denny--> actually I'm converting it to a .ts file before I start cutting - so that the cat method of stitching it back together will work
[12:57] <burek> denny--, hold on, what is your logical goal
[12:57] <denny--> heh
[12:57] <burek> what are you trying to accomplish
[12:58] <denny--> I have to remove some censored parts from the video stream (captured from a live show) before it can be rebroadcast
[12:58] <burek> like logos and stuff?
[12:59] <denny--> (I'm also removing advert breaks, and generating XML to mark where new ads should be reinserted, but that bit works okay as doesn't have such high accurancy requirements)
[12:59] <denny--> no, like people swearing
[12:59] <denny--> I need to remove from frame X to frame Y, in a few places
[12:59] <burek> then you need to do a video editing
[12:59] <burek> which requires re-encoding
[12:59] <burek> so no -vcodec copy
[13:00] <burek> if you need accuracy, you'll need to re-encode I'm afraid
[13:00] <denny--> I'm currently doing that by removing everything up to frame X, and keeping everything after frame Y, and then stitching those two halves back together
[13:00] <denny--> *nod* Yeah, so it seems
[13:00] <denny--> so what would be good codecs to use?
[13:00] <denny--> and can you see why the command above is generating a whole file instead of just the bit I want?
[13:00] <denny--> and why the metadata is broken?
[13:01] <burek> I would probably create a bash script, to decode the input into raw video/audio, cut off parts I don't want, joining parts that are ok, all in one go, and then just encode again, not to loose the quality in multiple re-encodings
[13:01] <denny--> I think someone here suggested somethign similar, yeah
[13:01] <denny--> so is the codec called 'raw' for that?
[13:02] <burek> the "command above" does not have any -t or -ss option, so..
[13:02] <burek> it's pretty much obvious why
[13:02] <denny--> it has a vf select
[13:02] <burek> oh, let me see
[13:02] <denny--> does that not say "I only want frames X to Y" ?
[13:03] <burek> "rawvideo"
[13:03] <burek> read this while I check select filter: http://ffmpeg.org/faq.html#Concatenating-using-raw-audio-and-video
[13:03] <denny--> thanks
[13:07] <burek> maybe you need to escape those commas
[13:07] <burek> the example of the select filter says: -vf select='gt(scene\,0.4)'
[13:07] <denny--> no, it blew up with them escaped
[13:07] <denny--> yeah, I did it that way first
[13:08] <denny--> it went boom  :)  loads of errors about too many params and stuff
[13:08] <denny--> I think that's for shell escaping - I'm running through Perl, which handles that for me
[13:08] <burek> oh ok
[13:11] <burek> hmh everything looks alright
[13:11] <burek> which version of ffmpeg are you using
[13:11] <denny--> ffmpeg version N-37063-g14d94a1 Copyright (c) 2000-2012 the FFmpeg developers   built on Jan 23 2012 17:42:45 with gcc 4.6.2
[13:12] <denny--> might lose my connection in a second, they're playing with our network ports
[13:12] <burek> yup, you might update it :)
[13:12] <burek> ok
[13:13] <denny--> ffmpeg version N-37063-g14d94a1 Copyright (c) 2000-2012 the FFmpeg developers   built on Jan 23 2012 17:42:45 with gcc 4.6.2
[13:13] <denny--> I have that one here too
[13:13] <denny--> is that any good, or should I download a new one?
[13:14] <denny--> sorry, wrong paste
[13:14] Action: denny-- hits windows with a stick
[13:14] <denny--> ffmpeg version N-43060-ga85b4a5 Copyright (c) 2000-2012 the FFmpeg developers   built on Jul 30 2012 13:18:07 with gcc 4.7.1 (GCC)
[13:14] <denny--> that one
[13:14] <burek> the 2nd one is newer obviously :)
[13:15] <denny--> not sure why I stopped using that one and switched to the other one
[13:15] <denny--> to match the production server possibly
[13:15] <denny--> but I can probably update that without anyone complaining
[13:16] <burek> it's never too much of testing :)
[13:22] Action: denny-- updates to the latest zeranoe build
[13:49] <arpu> http://ffmpeg.org/ down?
[13:50] <spaam> oh noes. ffmpeg is dying
[13:51] <spaam> and yes. cant connect to it
[16:05] <natrixnatrix89> Hi guys.. If I have a raw video with BGRA32 colorspace.. Is there any way I could convert it, using ffmpeg? Because in  'ffmpeg -pix_fmts | grep bgra' I only see bgra, bgra64be and bgra64le.. Where can I get bgra32?
[16:05] <natrixnatrix89> I tried playing, using pix-fmt bgra.. but all I got was black screen..
[16:07] <natrixnatrix89> basically what I'm trying to do is to read output from dump-gnash.. I set the right width and height.. but I can't seem to get it working..
[16:21] <burek> natrixnatrix89, any logs using pastebin?
[16:26] <natrixnatrix89> http://pastebin.com/vJ3ATiHZ
[16:27] <natrixnatrix89> http://pastebin.com/Yp0xJv9W
[16:29] <natrixnatrix89> maybe o have to cut off the beginning of that output file, because it may be blank..
[16:29] <natrixnatrix89> and try to use only the last 409 frames
[16:29] <natrixnatrix89> but I still don't know how to find and set the appropriate pix_fmt
[16:30] <burek> natrixnatrix89, do you read messages that ffmpeg shows? :)
[16:31] <burek> like for example: Option -pix_fmt is deprecated, use -pixel_format.
[16:31] <natrixnatrix89> lol
[16:31] <natrixnatrix89> that is just a warning
[16:31] <natrixnatrix89> that's not related to this question
[16:31] <natrixnatrix89> it doesn't matter if I used -pixel_format
[16:31] <burek> well, where is the rest of the log then
[16:32] <burek> btw, bgra is (i think) bgra32
[16:32] <burek> in pix_fmts it shows as 32bit
[16:32] <natrixnatrix89> The problem is that it works fine. but the output is just lack screen..
[16:32] <natrixnatrix89> oh.. so I had chosen the right one already..
[16:35] <natrixnatrix89> ok.. but if you want ill show you another example where I use -pixel_format and -video size.. I used these, because ffmpeg instead of ffplay accepts those just fine..
[16:37] <natrixnatrix89> http://pastebin.com/Hv5i6ipc
[16:50] <burek> natrixnatrix89, could you just type ffmpeg -i test2.raw
[16:50] <burek> or ffmpeg -vcodec rawvideo -i test2.raw
[16:52] <natrixnatrix89> http://pastebin.com/2q8Bh1aJ
[16:52] <natrixnatrix89> could not find codec parameters..
[16:52] <burek> and the 2nd example?
[16:53] <natrixnatrix89> totally the same http://pastebin.com/E6wq6Wpa
[16:53] <natrixnatrix89> I think the stream contains no headers about size, pixel format, etc..
[16:54] <burek> it shouldn't use image2 demuxer
[16:54] <burek> try this
[16:54] <burek> ffmpeg -f yuv4mpegpipe -vcodec rawvideo -i test2.raw
[16:56] <natrixnatrix89> hmm. test2.raw: Operation not permitted
[16:57] <burek> well you need to figure out what format does that thingie write
[16:57] <burek> since .raw can be pretty much anything
[16:57] <burek> if you have any docs or something
[16:57] <burek> try to figure out what is the format/container that it uses to write into the file
[16:57] <natrixnatrix89> no docs.. but I was trying to follow this tutorial
[16:57] <natrixnatrix89> http://techblog.floorplanner.com/post/20528549445/server-side-png-rendering-of-swf-images-using-gnash
[16:58] <natrixnatrix89> which doesn't say about the format much either..
[16:59] <burek> can it output to avi or something
[16:59] <burek> According to output, we have 8-bit BGRA raw video. What does it mean? It is just a sequence of one-byte values for each of four channels (Blue, Green, Red, Alpha) for each pixel for each frame. If we open that file with mplayer, we’ll see that our image flickers.
[17:02] <natrixnatrix89> I tried playing on mplayer.. didn't really work..
[17:02] <natrixnatrix89> but ffplay played just blank screen.. and if I played with pix_fmt to others it started to vary.. mostly black and white stripes
[17:06] <natrixnatrix89> also I don't understand, how does he conclude that bgra32 is actually bgra 8bit..
[17:06] <burek> natrixnatrix89
[17:06] <burek> ffmpeg -f rawvideo -vcodec rawvideo -i test2.raw
[17:06] <burek> or
[17:07] <burek> ffmpeg -f rawvideo -i test2.raw
[17:07] <burek> or
[17:07] <burek> ffmpeg -f rawvideo -pix_fmt bgra  -i test2.raw
[17:07] <natrixnatrix89> I think I should note -s 300x250 for the input too
[17:08] <burek> yes
[17:08] <burek> and -r 100
[17:10] <natrixnatrix89> wow.. it worked.. But I'm still wondering.. why did I get just black screen when I used exactly the same params for ffplay
[17:11] <burek> what params
[17:11] <burek> ffplay -f rawvideo test2.raw ?
[17:12] <natrixnatrix89> yeah.. and also pix_fmt and -s
[17:12] <natrixnatrix89> although it didn't accept -r
[18:30] <panchtatvam> don't libavcodec get installed automatically after installing ffmpeg in fedora
[18:30] <panchtatvam> my vlc compiling is unable to detect libavcodec
[18:31] <panchtatvam> is there a separate pkg for devel files
[18:31] <panchtatvam> ?
[18:31] <panchtatvam> of ffmpeg
[18:31] <panchtatvam> ?
[18:31] <aleray> hi, I have a set of two m4v files created using handbrake. It is two chapters of a dvd. I'd like to combine them into one m4v without reencoding. I've been looking at the FAQ but couldn't get it to work. Could you tell me how to proceed? thanks
[18:31] <burek> pastebin your vlc's config.log
[18:31] <burek> panchtatvam
[18:34] <panchtatvam> vlc config.log for libavcodec not found error
[18:34] <panchtatvam> http://pastebin.com/0zjVWvwz
[18:36] <panchtatvam> also , ffmpeg config.log -> http://pastebin.com/5cBWk02L
[18:36] <burek> did you compile ffmpeg yourself?
[18:36] <panchtatvam> yes
[18:36] <burek> did you run ldconfig
[18:36] <panchtatvam> and it compiled successfully
[18:36] <panchtatvam> no
[18:36] <panchtatvam> ./configure
[18:36] <burek> make install && ldconfig
[18:36] <panchtatvam> make
[18:36] <panchtatvam> make install
[18:36] <burek> type ldconfig
[18:36] <panchtatvam> ok
[18:36] <burek> and try configuring your vlc again
[18:37] <panchtatvam> trying
[18:37] <panchtatvam> :)
[18:41] <aleray> any idea? :)
[18:43] <panchtatvam> @burek configured ffmpeg and then vlc using ldconfig and then make and sudo make install but still same error "configure: error: No package 'libavcodec' found "
[18:45] <burek> panchtatvam, try #videolan
[18:46] <burek> aleray, no idea :)
[18:46] <aleray> burek, :/ ok some more googling for me
[18:48] <panchtatvam> @burek ok :)
[19:10] <breik0> Hello!! Has someone managed to compile ffmpeg for iOS 6 ?
[19:11] <breik0> ;)
[19:42] <panchtatvam> is there no libavcodec-devel package for fedora 17 ?
[19:43] <panchtatvam> ffmpeg doean't provide the devel libraries for libavcodec :( and I'm unable to compile vlc without libavcodec-devel.
[19:43] <panchtatvam> any suggestions
[19:43] <panchtatvam> ?
[20:08] <JCDenton> http://pastie.org/private/1tzhawwa9evjachdnoh6g Any clue why this video can't be read? The input video won't play in VLC, it reads the time then stops playing immediately, but it does play in mplayer. The only thing I've read on the error is that it might be a DRM protected file, which I'm not sure why it would be
[20:10] <JCDenton> "-f mpeg" seems to make it work. http://ffmpeg.org/trac/ffmpeg/ticket/241
[21:31] <Mavrik> um, what's the 2.6+ linux static build site?
[21:35] <relaxed> Mavrik: http://goo.gl/DPrRY
[21:35] <Mavrik> thank you :)
[21:36] <relaxed> It's also on the ffmpeg download page for future reference.
[22:40] <Kirill> hi! is there any way to improve ffmpeg's mpegts detection? I've got an mpeg-ts with two streams, one audio, one data, but ffmpeg parses about 10mb of data (many packets) before it starts decoding. am I doing something wrong?
[22:47] <Kirill> (anyone around?)
[00:00] --- Wed Sep 26 2012


More information about the Ffmpeg-devel-irc mailing list