[Ffmpeg-devel-irc] ffmpeg.log.20131223

burek burek021 at gmail.com
Tue Dec 24 02:05:01 CET 2013

[01:17] <klaxa> oh by the way, parsing failed because the filename had some special characters apparently
[01:17] <klaxa> i extracted the subs and called the file subs.ass and that worked
[01:17] <klaxa> rather undesirable
[01:18] <sacarasc> Like spaces?
[01:21] <klaxa> >Applying option vf (set video filters) with argument subtitles=filename=[WhyNot] SaintYoung Men The Movie [BD 720p AAC][12DEEE77].ass.
[01:21] <klaxa> >[Parsed_subtitles_0 @ 0x33e2440] Setting 'filename' to value ''
[01:21] <klaxa> i escaped it properly in the shell
[11:18] <JEEB> hmm, it must be flying right over my head right now, but how did I set width to automatically make the picture 1:1 SAR when I have height set to 720?
[11:20] <JEEB> sar*iw/(1080/720)
[11:20] <JEEB> this does it but isn't exactly elegant :D
[11:21] <viric> sar*iw/(width/height) ?
[11:21] <viric> hm
[11:21] <viric> nah, forget it. I'd have to think that. But a formula of that kind is the way,  I think
[11:21] <JEEB> scale=sar*iw/(ih/oh):min(ih\,720)
[11:22] <JEEB> this seems to do it :)
[11:22] <JEEB> this also works for things that are smaller than 720
[11:24] <JEEB> (you probably want to also add the chroma subsampling thing to this, too) :D
[11:38] <viric> JEEB: and what about encoders requiring multiples of 8, 16, ... in width or height?
[11:38] <viric> I've hit that annoyance sometimes
[11:39] <JEEB> viric, you just use the same way as you do with the chroma subsampling example in the scale filter's docs
[11:39] <JEEB> for chroma subsampling I just ended up with scale=trunc(sar*iw/(ih/oh)/hsub)*hsub:trunc(min(ih\,720)/vsub)*vsub
[11:40] <JEEB> truncate with division by the limitation, and then multiply by that limitation again
[11:45] <viric> o
[11:45] <viric> k
[11:48] <clever> JEEB: making some decent progress on the rpi decoding and rendering
[12:06] <juke> hi
[12:07] <juke> is there a way to launch a script when an image is extract from video ?
[12:57] <DeadSix27> is it possible to create a slideshow of a SINGLE image, with background image in ffmpeg?
[12:57] <DeadSix27> (isnt really a slideshow anymore)
[12:58] <viric> I can't understand what you ask for
[12:59] <DeadSix27> Is it possible to create 1fps video, with only an image and music as source?
[13:00] <DeadSix27> e.g -i <path to image> -i <path to audio> -c:v libx264 -r 1 -c:a mp3 out.mp4
[13:00] <DeadSix27> would that work?
[13:01] <viric> you may need to use -shortest
[13:01] <relaxed> ffmpeg -i audio -loop 1 -i image -shortest -c:v libx264 -c:a copy out.mp4
[13:01] <viric> to stop encoding when the shortest (V or A) stops. The image will be in an endless loop, so...
[13:01] <DeadSix27> ah right.
[13:02] <DeadSix27> would it use the image dimensions?
[13:02] <DeadSix27> or do i have to define them manually?
[13:02] <viric> it will use those.
[13:03] <DeadSix27> shortest makes it 2s long.
[13:05] <relaxed> DeadSix27: -bf 0
[13:06] <relaxed> or grep the duration, omit -shortest and use -t $duration
[13:06] <DeadSix27> yes i just did the -t solution
[13:06] <viric> why 2s long
[13:06] <viric> ?
[13:07] <relaxed> b frames from libx264
[13:08] <viric> ok. why shortest doesn't make it as long as the audio?
[13:08] <DeadSix27> dont ask me
[13:09] <DeadSix27> hm
[13:09] <DeadSix27> using -bf 0, plays the audio fine, and length is right as well, but no video is showing
[13:11] <relaxed> try with -x264-params bframes=0 instead of -bf 0
[13:14] <DeadSix27> that makes the video size somehow work i mean prior to that, player just didnt "get the size" and it was like audio only
[13:14] <DeadSix27> now its getting the video size, but its still just black.
[13:14] <DeadSix27> sec.
[13:16] <DeadSix27> relaxed: http://pastie.org/private/ws5h4yemqvg5l72ke6bg
[13:17] <relaxed> you need -loop 1 before the input you wish to loop
[13:17] <relaxed> otherwise you're asking for one frame
[13:17] <DeadSix27> ah thought it does that automatically.
[13:18] <relaxed> and you don't need 0 bframes if you're using -t $audio_duration
[13:18] <DeadSix27> right. leftover
[13:18] <DeadSix27> (tried with, without, together)
[13:19] <DeadSix27> btw is it an issue of the experimental aac encoder, that the quality sounds always way worse compared to libvo_aacenc?
[13:20] <DeadSix27> compared to using*
[13:20] <relaxed> libfdk_aac is considered the best that ffmpeg supports.
[13:20] <DeadSix27> think that isnt compiled in zeranoes builds (or even linux only, forgot)
[13:20] <viric> isn't it incompatible with gpl?
[13:21] <viric> DeadSix27: you need a special build taking out all gpl pieces
[13:21] <relaxed> yes, but that doesn't stop people from building it locally
[13:21] <DeadSix27> :p
[13:21] <viric> right, if you mean to distribute it.
[13:21] <DeadSix27> well i dont really want to go through the compiling ffmpeg
[13:21] <relaxed> or just use neroaacenc and stream copy the audio with ffmpeg
[13:22] <DeadSix27> would be to heavy on the end, the main reason i use ffmpeg, is because its all in one.
[13:23] <viric> leave a message to the aac encoder people
[13:23] <viric> "this is very uncomfortable"
[13:23] <DeadSix27> on the other hand, i always wanted to try to compile ffmpeg my own
[13:24] <DeadSix27> is it much work to compile it with the mentioned library support?
[13:24] <DeadSix27> (compared to doing it without)
[13:24] <relaxed> compiling anything on windows is a chore, in my opinion
[13:24] <DeadSix27> yep, but got used to that much.
[13:25] <relaxed> however, once the tools are in place it's not that difficult
[13:25] <DeadSix27> but that ofc, is the main reason i usually just use zeranoes builds
[13:26] <DeadSix27> anyway, for later searching, what do i have to look out for?
[13:26] <DeadSix27> ffmpeg without gpl?
[13:26] <viric> for me, compiling anything on windows is very hard
[13:27] <relaxed> DeadSix27: if you're not distributing ffmpeg you don't have to worry about the gpl or any of that.
[13:28] <relaxed> I believe zeranoe's website has links and maybe a guide on how to compile ffmpeg on windows.
[13:28] <relaxed> also, https://trac.ffmpeg.org/wiki/CompilationGuide
[13:29] <DeadSix27> relaxed: what i meant by that, is that i thought its some kind of "option" when compiling
[13:29] <DeadSix27> "--with-gpl" = disables that speicifc aac encoder
[13:29] <DeadSix27> like that.
[13:30] <DeadSix27> theres ffmpeg on android o.O
[13:30] <relaxed> --enable-libfdk-aac requires --enable-nonfree
[13:30] <DeadSix27> ah ok
[13:30] <DeadSix27> thats what i meant.
[13:30] <viric> DeadSix27: why do you encode aac?
[13:30] <DeadSix27> required of players.
[13:30] <DeadSix27> they dont support ffmpeg/ac3
[13:30] <DeadSix27> errr
[13:30] <DeadSix27> mp3*
[13:30] <relaxed> most video players on android use ffmpeg's libs
[13:31] <relaxed> including my favorite, mx player
[13:31] <DeadSix27> but ye, where i can, i just copy the audio
[13:31] <DeadSix27> to prevent Quality loss
[13:31] <DeadSix27> relaxed: i usally use Moboplayer
[13:32] <tholin> I have a flv file with broken pts. Is there some way to remux it to fix the timestamps?
[13:34] <jnvsor> How exactly are the PTS broken? Both tracks? Scaled?
[13:35] <tholin> it jumps back and forward in time
[13:36] <jnvsor> Audio as well as video?
[13:36] <tholin> I think so
[13:41] <jnvsor> aresample looks like a good lead, but I'm not sure how to rebuild PTS from scratch or if it's even possible
[13:42] <viric> you can use setpts and asetpts in filters
[13:56] <tholin> what I'm really trying to do is extract an flv from a pcap dump using a buggy tool
[13:57] <tholin> I'm trying to fix it but the incremental timestamps adds up too fast and gets larger than the next absolute timestamp
[14:05] <DeadSix27> relaxed: Do you know where exactly this is located: https://trac.ffmpeg.org/wiki/MingwCompilationGuide#Git
[14:05] <DeadSix27> relaxed: see: "fix the HOME variable, so that it will point to $MINGW_PATH/msys/1.0/home/$USER"
[14:27] <DeadSix27> relaxed: nvm.
[14:38] <DeadSix27> relaxed: btw im not rly thaat skilled in this, but can you compile from linux to windows?
[14:38] <DeadSix27> that would make this whole thing way easier, since i have a couple linux servers
[14:39] <JEEB> yes
[14:39] <JEEB> use zeranoe's mingw-w64 toolchain script to build you a nice mingw-w64 toolchain for either 32bit or 64bit binaries, or both
[14:40] <JEEB> http://ffmpeg.zeranoe.com/blog/?p=242
[14:40] <DeadSix27> interesting
[14:41] <JEEB> basically see that you have the dependencies, and then let it roll for some time... then add the bin directories within PATH so you can use them freely
[14:42] <JEEB> and then you can configure with cross-prefix and friends
[14:42] <JEEB> for example with 32bit binaries, --cross-prefix=i686-w64-mingw32- --arch=i686 --target-os=mingw32
[14:43] <DeadSix27> well not so far yet.
[14:43] <JEEB> (yes, the dash after mingw32 is deliberate, since you have i686-w64-mingw32-gcc for example)
[14:43] <DeadSix27> i dont even met the requirements atm
[14:43] <JEEB> :)
[14:43] <JEEB> well, shouldn't take too long to get those installed
[14:43] <DeadSix27> -> cvs yasm git flex bison
[14:43] <DeadSix27> was needed.
[14:43] <DeadSix27> (Didnt know git wasnt installed on this server)
[14:43] <DeadSix27> (funny)
[14:44] <JEEB> but yeah, just noted you an example of the three settings you have to set when cross-compiling :)
[14:45] <DeadSix27> was similiar to what i had before:  --enable-cross-compile --target-os=mingw32 --arch=x86
[14:46] <DeadSix27> i like my style of time-wasting.. it began with a 10s compiling of a music piece
[14:46] <DeadSix27> creating*
[14:46] <DeadSix27> and ended in compiling ffmpeg myself for windows
[14:46] <DeadSix27> just how the heck did i get there.
[15:05] <saste> DeadSix27, install the static build and you're done
[15:10] <DeadSix27> ?
[15:43] <jarr0dsz> hi everyone, can anyone tell me how to make an image from a .flv ? just 1 single image in size 400x400 for example
[15:43] <jarr0dsz> i cannot find any info on how to create single image from movie only convert movie fully to images
[15:43] <jarr0dsz> dsfadsa
[15:43] <jarr0dsz> fasd
[15:43] <jarr0dsz> fsad
[15:43] <jarr0dsz> fasd
[15:43] <jarr0dsz> see no text this comes true?
[16:13] <DeadSix27> JEEB: when compiling how do i tell it, where libfdk_aac is?
[16:13] <JEEB> LD_LIBRARY_PATH or --extra-cflags="-I/your/windows/prefix/include" --extra-ldflags="-L/your/windows/prefix/lib"
[16:14] <JEEB> uhh, not LD_LIBRARY_PATH
[16:14] <JEEB> I meant :P
[16:14] <JEEB> one of those two alternatives
[16:16] <DeadSix27> JEEB: thanks  worked
[17:06] <RioKuro> Hey guys, I've got a problem. I'm trying to build HD Broadcast complient mxf files from dpx sequences. The dpx files have BT709 primaries but use full range levels. I'd like to scale them to studio levels but it's just not working. I've tried using the scale filter with in_range and out_range and I've also tried to use the -color_range option but still nothing. hope you guys can help.
[17:07] <RioKuro> oh i'm using ffmpeg 2.1.1 btw
[17:59] <PowerCC> hello
[18:02] <PowerCC> in order to hardcode (burn-in) subs if it's not srt it must be converted to .ass still?
[18:03] <PowerCC> libass
[18:03] <klaxa> i don't think so
[18:03] <klaxa> wait...
[18:03] <klaxa> if it's not srt t must be converted to ass?
[18:04] <klaxa> hmm... the docs only mention .srt and .ass
[18:08] <PowerCC> for example dvdsubs
[18:08] <PowerCC> or picture based
[18:09] <PowerCC> overlay video filter
[18:10] <PowerCC> i am interested in doing stuff on the fly (transcode)
[18:13] <PowerCC> something like
[18:13] <PowerCC> ffmpeg.exe -i path-to-movie\movie.mkv -filter_complex "[0:0][0:4]overlay[v]" -vcodec mpeg2video -b 16384k -maxrate 30000k -bufsize 4096k -ab 448k -ar 48000 -ac 6 -acodec ac3 -copyts -map [v] -map 0:1 -f mpegts -
[18:55] <pablocastellanos> Hi, how I can duplicate a stream and transcode it but putting it at a specific stream position?
[19:19] <jnvsor> When I use a scale filter sqscaler complains about a deprecated pixel format - What can I do to remove this warning?
[20:11] <jnvsor> Trying to get rid of the swscaler warnings about pixel format here, any ideas? http://pastebin.com/BNXehJ7K
[20:19] <stieno> How can one establish inter stream synchronisation?
[20:20] <stieno> Using rtp, rtcp video stream?
[20:25] <brontosaurusrex> is there an explanation how 10bit > 8bit and vice versa works by default in ffmpeg? (dithering or what?)
[21:10] <_fil_> Hi, I need to convert mp4 to mpeg in order to create a DVD with iDVD. How do I specify that I don't want to lose quality?
[21:14] <sacarasc> _fil_: When encoding to a lossy format, you will always lose quality.
[21:23] <Looney> quick question and I will be out of your hair
[21:24] <Looney> what is the difference between neon and no neon support
[21:24] <Looney> https://github.com/yixia/FFmpeg-Android/blob/master/FFmpeg-Android.sh
[21:24] <viric> neon are simd instructions for ARM
[21:24] <Looney> ok
[21:25] <viric> neon code is likely to process things faster on an ARM that understands neon
[21:25] <Looney> so, its like hardware acceleration
[21:25] <Looney> ok, and what is vfp
[21:26] <klaxa> something with floating point
[21:26] <Looney> only asking as I need to build ffmpeg with ssl enabled for android
[21:26] <klaxa> http://en.wikipedia.org/wiki/ARM_architecture#Floating-point_.28VFP.29
[21:26] <Looney> and I want to make sure my ssl build is compatible
[21:26] <viric> vfp instructions are instructions for an ARM typical FPU
[21:26] <Looney> ok
[21:26] <viric> there are different versions even.
[21:27] <Looney> so, if I build openssl with arm6 and arm7 I am good to go ? ?
[21:27] <viric> without any vfp, floating point code may be written with soft fp, no hw fpu
[21:27] <viric> ssl won't use vfp
[21:27] <Looney> I also hope so
[21:27] <viric> floating point encryption? :)
[21:28] <Looney> because those computation won't be much fuss
[21:28] <Looney> no, https stream  :-)
[21:30] <Looney> gotta get going for now
[21:30] <Looney> will bug you mateys more morrow  ;-)
[00:00] --- Tue Dec 24 2013

More information about the Ffmpeg-devel-irc mailing list