[Ffmpeg-devel-irc] ffmpeg.log.20140603

burek burek021 at gmail.com
Wed Jun 4 02:05:01 CEST 2014


[00:03] <Zanzibar_1> what is best h265 encoder
[00:06] <c_14> Pretty sure you're looking for x265. Not like there's that many to choose from.
[00:06] <bencoh> there are a few
[00:07] <bencoh> (3~4 opensource projects afair)
[00:08] <c_14> Right, forgot about that. Pretty sure they're all called x265 though...
[00:09] <c_14> Nah, one of them is called f265.
[00:12] <c_14> Oh man. Not sure which of them is the 'best' and I'm not even sure there is an objective best, but this is the one I've used when I played around with HEVC: http://x265.org/
[00:20] <Zanzibar_1> where do i get the exe file?
[00:22] <klaxa> i don't think you do at all
[00:22] <klaxa> compile from source or...
[00:22] <klaxa> no idea what the other options are
[00:25] <mehhthere> does anyone have experience building rtmpdump for RHEL/CentOS?
[03:31] <spectralsun> ffmpeg -i concat:"files|files2" isnt working for me
[03:31] <spectralsun> saying cannot find the whole thing as though it were looking for a single file
[03:31] <spectralsun> version 2.2.2
[03:35] <spectralsun> llogan: https://dpaste.de/bfTM
[03:39] <spectralsun> oh woops
[03:39] <spectralsun> middle file looks to be missing
[03:43] <llogan> spectralsun: the concat protocol only works with some formats.
[03:45] <llogan> the demuxer might be better for you
[03:45] <llogan> https://trac.ffmpeg.org/wiki/How%20to%20concatenate%20%28join,%20merge%29%20media%20files
[03:55] <tlacatlc6> is there a way to strip off non-audio parts? This "ffmpeg -i 00000.m2ts -c:a pcm_s16le -map 0:1 -f wav -" makes flac encoder not happy.
[03:58] <llogan> tlacatlc6: ffmpeg -i input -map 0:a:0 -f wav -
[03:58] <llogan> but ffmpeg has a native flac encoder too
[03:59] <tlacatlc6> thx, i'll try that. i know, i'm just trying to get it working with the external flac encoder.
[04:03] <tlacatlc6> didn't work same error! :(
[04:18] <vl4kn0> Hi, is it possible to use frame metadata in the expression evaluation language?
[04:27] <tlacatlc6> llogan, http://pastebin.com/i3GSVYyc
[04:34] <llogan> tlacatlc6: add --ignore-chunk-sizes as an option for flac
[04:37] <tlacatlc6> that worked thanks, i never saw that on the help output of flac.
[06:09] <tlacatlc6> btw, is it possible to use ffmpeg to encode audio to truehd?
[06:17] <sacarasc> Pretty sure it can't.
[06:20] <tlacatlc6> oh, i mistook that 'E' as encoding but i now see it says muxing supported.
[06:40] <theekoz> anyone know how to go from 16bit audio to 24/32 bit audio in ffmpeg
[06:40] <theekoz> reading through the documentation without luck but thought someone might know offhand
[06:41] <theekoz> -af s24 seem right?
[06:45] <theekoz> ac3 btw
[06:48] <tlacatlc6> theekoz, maybe something like -c:a pcm_s24le.
[06:49] <theekoz> will that work for AC-3 you think
[06:49] <theekoz> I'll check it out
[06:49] <theekoz> thanks
[10:06] <yrc> Hi! I have a DVD that begins to be scratched, but I can still salvage it with dvdbackup (I get the VIDEO_TS structure on the hard disk). I want to create a x264/mkv video with chapters. I managed without problems, but I get interlacing artifacts (the DVD is interlaced PAL). How would I go for removing these artifacts?
[10:07] <yrc> I understand from the documentation that deinterlacing is better not done because it lessens the quality. So how do I instruct ffmpeg to take interlacing into account when encoding the x264 video?
[10:07] <sacarasc> Using a crappy deinterlacing video filter with ffmpeg, or doing it properly using something like avisynth.
[10:09] <yrc> Hi sacarasc, thank you for your answer. I thought it was better to not deinterlace... Isnt it so?
[10:09] <sacarasc> You're going to have to deinterlace somewhere along the line, unless you're playing on a device that can show it properly.
[10:11] <yrc> sacarasc: My device is a BD/DVD/media player, and it can handle interlacing perfectly fine (at least, nothing bad shows when I view an interlaced DVD). In this case, I suppose it is best to leave the interlacing in order to keep quality.
[10:15] <yrc> Hum... So was I wrong after all? Is it better to deinterlace? In this case, I may try ffmpegs yadif; Ive read it is good enough...
[10:16] <sacarasc> I've never encoded stuff interlaced, and I usually play on a computer, so I like to deinterlace slowly when encoding.
[10:20] <yrc> Ok. Thanks for your help anyway :-)
[11:21] <Mavrik> hmm
[13:31] <SpecialEd> Hello, I got ffmpeg 2.2.git on a ubuntu 14.04 server and I'm trying to transcode a flash video file that uses AVC/AAC for video/audio codecs over to x264/aac mp4 using the command: ffmpeg -i somefile.flv -acodec libfaac -vcodec libx264 somefile.mp4
[13:32] <SpecialEd> the audio layer seems to work well but the video layer fails to transcode and ffmpeg provides the following output:
[13:32] <SpecialEd> [flv @ 0x2fb8c00] decoding for stream 0 failed
[13:32] <SpecialEd> [flv @ 0x2fb8c00] Could not find codec parameters for stream 0 (Video: h264, 585 kb/s): unspecified size
[13:32] <SpecialEd> Consider increasing the value for the 'analyzeduration' and 'probesize' options
[13:32] <SpecialEd> I've tried using different codecs as well as setting the flags: -analyzeduration 10000000 -probesize 100
[13:33] <SpecialEd> the source flv file does play with video with VLC so its not a corrupted file
[13:33] <SpecialEd> if anyone can point me in the right direction for a resolution path I would be very greatful :)
[13:50] <c_14> SpecialEd: try adding -s wxh as an input option. or try increasing analyzeduration and probesize to 2G
[13:52] <SpecialEd> ok c_14 trying now, bb in 2 mins :)
[13:53] <SpecialEd> ok actually I also have tied -s wxh
[13:54] <SpecialEd> so im gonna try 2G for ad and ps
[13:55] <SpecialEd> unfortunately no luck :(
[14:11] <c_14> Can ffplay play the file?
[14:15] <SpecialEd> hmm
[14:15] <SpecialEd> let me see, i honestly have never used ffplay
[14:19] <SpecialEd> yes it can play
[14:19] <SpecialEd> however I can't confirm if it plays the audio as I dont have speakers hooked up to my local ubuntu server
[14:19] <SpecialEd> but it definitely plays the video just as VLC does
[16:14] <DelphiWorld> hi guys;)
[16:14] <DelphiWorld> c_14: dear, what's a switable Subtitle format to be integrated in mpeg4
[16:15] <DelphiWorld> or rater streamed;)
[16:16] <c_14> mp4 offically only supports mpeg-4 timed text subtitles.
[16:16] <DelphiWorld> c_14: ok, is converting dvbsub  to this timed text subtitle pocible?
[16:16] <DelphiWorld> c_14: also, if its pocible, i wish to use a format that should be both mpeg4 & mpeg2ts compat
[16:17] <termos> it seems that i can run yadif on both interlaced and non-interlaced input, and it will do nothing if the input is not interlaced. Is this correct?
[16:17] <termos> in that case I won't have to run the idet filter on my video, which would be nice.
[16:18] <c_14> termos: theoretically yes, if you set the flags to the yadif filter correctly.
[16:18] <c_14> pretty sure dvbsub is a picture subtitle format while timed text is text, the only way to convert from dvbsub to timed text that I know of would be ocr.
[16:19] <DelphiWorld> c_14: so i shouldn't use mpeg4 so
[16:21] <termos> c_14: okey, I have it set to just the default 0:0:0 now though
[16:22] <c_14> termos: you're going to need deint=1, you should also be able to change the third 0 to a 1
[16:25] <c_14> DelphiWorld: matroska tends to support just about everything, so that's usually a good choice if it's supported by the clients
[16:25] <DelphiWorld> c_14: i belieuv is supported. but here come the question
[16:25] <DelphiWorld> if i use mkv
[16:25] <DelphiWorld> 1. what streaming PROTO to use?
[16:26] <DelphiWorld> 2. if i use MKV i'lle need to REENCODE  to get HLS
[16:28] <c_14> Can you give me a quick rundown of what you're trying to do?
[16:29] <DelphiWorld> c_14: IPTV platform with android & iOS client with pocibility of views over web
[16:30] <c_14> Nah, more specifically. What were you planning on using the mp4 for since hls streams are mpeg-ts anyway.
[16:31] <DelphiWorld> c_14: mp4 was for droid over rtsp, just asked for mp4
[16:31] <DelphiWorld> or flv
[16:31] <DelphiWorld> flv dont i'm sure
[16:35] <sine_> will -target pal-dvd also resize to the dvd format ?
[16:36] <sine_> scale resoloution
[16:36] <sine_> -o
[16:36] <c_14> I think you can send mpeg-ts over rtsp, but I can't find anything to back that up. You could always try throwing the dvbsub stream into the mp4 and see what happens. Might just error out though.
[16:37] <DelphiWorld> c_14: no encoder...
[16:38] <c_14> sine_: it resizes on my end, so probably yes.
[16:38] <c_14> DelphiWorld: with -c:s copy ?
[16:38] <DelphiWorld> oh no
[16:39] <DelphiWorld> c_14: let me give you the cmd
[16:39] <DelphiWorld> c_14: ffmpeg -re -i udp://@239.100.1.8:1234 -map 0:0 -map 0:8 -map 0:2 -c:v libx264 -vb 512k -x264opts colorprim=bt709:transfer=bt709:colormatrix=bt709:fullrange=off -c:a libfdk_aac -profile:a aac_he_v2 -b:a 32k -ac 2 -ar 44100 -f mp4 euronews.mp4
[16:39] <sine_> c_14: thanks. anyone know how i can specify a certain number of frames so i can test 10 seconds worth
[16:39] <DelphiWorld> where 0.8 is the subtitle
[16:40] <c_14> sine_: -t 10
[16:40] <sine_> -t 10 is 10 frames yes
[16:40] <sine_> so after the encoder options -t 100 would be 4 seconds of pal 25fps
[16:40] <c_14> sine_: that should be 10 seconds, not 10 frames
[16:40] <sine_> oh ok
[16:40] <DelphiWorld> c_14: finish with sine_ ;)
[16:40] <sine_> cool that will do
[16:40] <sine_> im out for now
[16:41] <c_14> DelphiWorld: try adding -c:s copy
[16:41] <DelphiWorld> c_14: so no need for a -map?
[16:42] <c_14> DelphiWorld: No, you still need that. just add -c:s copy to the output parameters
[16:45] <sine_> how can i set the starting frame and the amount of frames
[16:46] <DelphiWorld> c_14: lol, no luck:P
[16:46] <DelphiWorld> -f mpegts worked
[16:47] <DelphiWorld> is saving into file, i'lle ffprobe it
[16:47] <sine_> -ss -f i got it
[16:50] <DelphiWorld> ok c_14 here come the cmd that work
[16:51] <DelphiWorld> ffmpeg -re -i http://192.168.0.35:4264/bysid/1031 -map 0:0 -map 0:1 -map 0:6 -map 0:7 -c:v libx264 -vb 512k -x264opts colorprim=bt709:transfer=bt709:colormatrix=bt709:fullrange=off -c:a libfdk_aac -profile:a aac_he_v2 -b:a 32k -ac 2 -ar 44100 -f mpegts -c:s copy euronews.ts
[16:53] <c_14> hmm, seems mp4 really doesn't like dvbsub subtitles.
[16:53] <DelphiWorld> c_14: i think mp4 should be ignored, dont you?
[16:53] <c_14> hmm?
[16:54] <DelphiWorld> c_14: i think i will ignore using mpeg4
[16:54] <c_14> ah, ok
[16:56] <DelphiWorld> i thought mpeg4 is widely deployed and could be playable for most devices
[16:58] <c_14> ye, but it doesn't support many audio/video/subtitle formats.
[16:58] <DelphiWorld> then suck
[17:07] <theekoz> Is it possible that FFMPEG has a default audio stream delay in the AC3 codec (you can typically set the delay in the AC3).. causing some issues on export.
[17:09] <theekoz> trying to export AC3 audio via FFMPEG/FFMBC .. then import as ts-extra audio in OBE due to needing captions.
[17:09] <theekoz> always getting a 10-11 second audio stream delay and don't know why
[17:29] <Fjorgynn> I've got a question about ffmpeg. If I use libmp3lame and the -V 0 option (highest quality) and get ~139 kbps and my original file is 256 kbps CBR. What does that mean?
[19:49] <relaxed> Fjorgynn: it uses a variable bitrate instead of a constant bitrate, which means it only uses the bitrate needed.
[19:57] <max246> hello
[19:58] <max246> I was trying to record with ffmpeg (avconv )   from a webcam in fullhd, but it looks likle that the quality is not very good
[19:58] <max246> anybody here is kind of the king of ffmpeg and can help me ? :D
[19:58] <llogan> ffmpeg is not avconv
[19:59] <max246> oh I see
[19:59] <max246> so what is avconv ? :d
[19:59] <llogan> something from a fork of FFmpeg
[19:59] <max246> because on debian when I type ffmpeg, i HAve been asked to type avconv
[19:59] <max246> which one is more updated? ffmpeg I supposed
[19:59] <llogan> you're using the counterfeit "ffmpeg" from a fork
[19:59] <llogan> http://blog.pkh.me/p/13-the-ffmpeg-libav-situation.html
[20:00] <llogan> so you'll either have to use real ffmpeg if you want help here, or get help from libav (the fork)
[20:00] <max246> oh thanks
[20:01] <max246> will you still help me if I give you avconv stuff?
[20:01] <llogan> no
[20:01] <max246> I mean I am just recording ...
[20:01] <llogan> i don't use avconv
[20:01] <max246> I think the commands are very similar
[20:01] <llogan> it's easy to download and use ffmpeg
[20:01] <llogan> http://ffmpeg.org/download.html#LinuxBuilds
[20:01] <max246> mhh ok
[20:01] <llogan> download, extract, execute
[20:02] <max246> ok
[20:02] <max246> but after would you able to help me ? :D
[20:02] <max246> I wouldnt spend 30 minutes to compile then...
[20:03] <llogan> yes. i can help if you're using ffmpeg from FFmpeg
[20:07] <max246> would the static version work for me?
[20:07] <llogan> yes
[20:07] <max246> good
[20:07] <max246> so I have it
[20:07] <max246> I want to record from a webcam full hd, logitech c920
[20:09] <llogan> can you show the complete output of: ./ffmpeg -f v4l2 -list_formats all -i /dev/video0
[20:10] <max246> h264 is 1920x1080
[20:10] <max246> so that is good
[20:12] <max246> I would like to record at 1920x1080 with the codec h265
[20:12] <max246> * h264
[20:13] <max246> I think the sintay is something like ./ffmpeg -f h264 -s 1920x1080 -i -/dev/video0 test.avi
[20:13] <max246> isnti it?
[20:13] <llogan> can you show the complete output of: ./ffmpeg -f v4l2 -list_formats all -i /dev/video0
[20:13] <llogan> use a pastebin site
[20:14] <max246> http://pastebin.com/fEVU6Tut
[20:23] <llogan> max246: ./ffmpeg -f v4l2 -i /dev/video0 -c:v libx264 -crf 18 -pix_fmt yuv420p output.mp4
[20:24] <llogan> it should automatically choose the "best" stream for input, AFAIK
[20:24] <max246> ok could you explain me what crf is?
[20:24] <max246> is the quality ?
[20:25] <llogan> constant rate factor. think of it as an option for libx264 that is as close as you can get to "give me a certain, constant level of quality".
[20:25] <llogan> ~18 is roughly visually lossless generally
[20:26] <max246> mh... its like pixel
[20:26] <max246> and the resolution is 1280
[20:27] <max246> oh wait
[20:28] <max246> ok
[20:28] <max246> it worked, let me try to put something moving in front of it
[20:29] <max246> btw, is this the best quality? if I get the raw input do I get the best quality?
[20:30] <llogan> yes, i assume so
[20:32] <llogan> if your cpu is too slow to encode at real time you can stream copy one of the available compressed formats produced by the camera, but quality might not be as good
[20:32] <max246> mhh ok
[20:32] <max246> well it-s a new i3
[20:32] <max246> not new.. but still new generation
[20:32] <max246> model name	: Intel(R) Core(TM) i3-4130 CPU @ 3.40GHz
[20:33] <max246> I think I need to get the raw from the webcam, then I can render the video with the slow flag
[20:34] <max246> to get a raw video with the higher quality, do you know these parameters?
[20:34] <llogan> if the camera outputs shit quality, then that's what you have to work with
[20:35] <max246> well.. it is not very shit
[20:35] <llogan> i wonder what the native frame size is and if any of the available frame sizes are the camera just upscaling. not sure if they do that or not but it would seem like a good thing to offer in terms of marketing misleading
[20:36] <max246> but I would prefer to get the raw video
[20:36] <max246> I can compress it later
[20:36] <llogan> i'm assuming that's what my command is doing for you
[20:37] <llogan> i mean it's getting the "rawvideo" input and encoding it to H.264
[20:38] <max246> yeah but I have to ran 4 cameras at the same time, to I prefer dont waste or get a lag because I am compressing on the fly
[20:40] <llogan> you'll have to experiment to see if you can re-encode on the fly at the frame sizes you want, or add -preset ultrafast, or see of lossless mode is faster, or just stream copy
[20:40] <llogan> or choose a smaller -video_size
[20:41] <max246> I need to get at least 720p
[20:41] <max246> conversion can take more time
[20:41] <max246> I need to record 4-5 minutes from 4 cameras
[20:41] <max246> if the compresion will take 10 minutes, I dont care :D
[20:45] <max246> with this I should just get the raw video ./ffmpeg -f video4linux2 -input_format mjpeg -i /dev/video0 -c:v copy output.mkv
[20:45] <max246> or I am doing it wrong?
[20:45] <llogan> mjpeg is compressed
[20:46] <max246> oh
[20:46] <max246> how do I get  the uncompressed?
[20:46] <max246> sorry.. I am kind of new on this world..
[20:46] <max246> so I am asking so many silly questions
[20:47] <llogan> it will probably choose uncompressed by default, or add -input_format yuyv422 if you're paranoid
[20:48] <max246> eheh
[20:48] <max246> thanks
[20:49] <llogan> the compressed formats may not look too different than the raw
[20:49] <max246> like what?
[20:51] <llogan> your camera offers yuyv422, h264, and mjpeg. with my camera, for example, the quality difference between "raw" and "mjpeg" is not huge, as far as I can tell. meaning they both look like shit.
[20:51] <max246> ahah ok
[20:52] <max246> thats far enough
[21:15] <vl4kn0> Hi champs, I'm looking for a way to use frame metadata (lavfi.cropdetect.x) in the ffmpeg eval language. Is it possible? Or is there any other way to make use of cropdetect in the crop filter? Thanks.
[21:21] <Mavrik> as far as I see it's not really exposed.
[21:27] <vl4kn0> And how would you proceed to use the value from cropdetect in the crop filter? Not in the commandline but programatically. I currently have set up a filtergraph but since the filters are linked together there is no easy way to do that. Except maybe use multiple filtergraphs and adjust the values with each frame.
[21:30] <Mavrik> vl4kn0, um, look at the source of cropdetect filter.
[21:31] <Mavrik> it's a dict as extradata on frame.
[21:37] <user3> is this a lossless conversion, and back: "ffmpeg -i 1.mp4 -qscale:v 1 tmp.mpg" "ffmpeg -i tmp.mpg -crf 63 out.mp4" the resulting out.mp4 seems of lesser quality than the original 1.mp4???;
[21:38] <sacarasc> user3: No, mpeg is lossy, also crf 63 is exceptionally lossy.
[21:38] <sacarasc> -crf 0, with x264, is lossless.
[21:39] <user3> oops, I thought crf 0 was the lossiest and 63 the least lossy
[21:40] <user3> what about -qscale:v 1, is this the best I can do to keep the original quality?
[21:41] <sacarasc> Why not just go -i 1.mp4 -c:v libx264 -crf 0 out.mp4? The first step isn't really needed.
[21:41] <user3> actually, I want to concatenate files, and that was just a trivial case to try to understand how to do it
[21:42] <Mavrik> use concat muxers / filters
[21:42] <llogan> libx264 ignores qscale, IIRC
[21:42] <Mavrik> instead of reencoding several times, you're losing alot of quality that way.
[21:48] <user3> so this may still be a bit lossy but better than my previous example: "ffmpeg -i 1.mp4  -c:v libx264 -crf 0 tmp.mpg" "ffmpeg -i tmp.mpg -crf 0 out.mp4"
[22:22] <maksimkaaa> I am trying to use mmcat.sh to concat two videos (mp4), I keep getting this error: http://pastebin.com/pmKcjCc1 can anybody advice?
[22:38] <c_14> maksimkaaa: try adding -pix_fmt yuv420p to line 86
[22:39] <c_14> But you might consider using one of the other methods, it's easier to control, easier to customize, and easier to debug.
[22:39] <maksimkaaa> c_14: will check that, moment
[22:42] <maksimkaaa> c_14: what are the easier methods ?
[22:43] <c_14> If the files all have the same codec, just use the concat demuxer. If they don't the concat filter.
[22:45] <maksimkaaa> c_14: good point.. btw changing the line 86 I got an output but it was the first video only.. not sure why..
[22:47] <c_14> eh, add the same thing on line 105
[22:51] <maksimkaaa> c_14: worked :)
[22:59] <spectralsun> can you un-do slow motion?
[22:59] <spectralsun> "-vf 'setpts=%i*PTS'" % SLOMO_FACTOR,
[22:59] <spectralsun> that is how im doing the slowmo^
[23:00] <c_14> -vf 'setpts=(1/%i)*PTS'
[23:00] <c_14> should do it
[23:00] <c_14> not sure if setpts supports division though...
[23:03] <c_14> But basically, big numbers make slower, small numbers make faster.
[23:10] <spectralsun> c_14: thanks I will give that a try :)
[23:17] <Navion> Does anyone know where I can get libavformat.55.dylib for an Intel Mac?
[23:17] <c_14> Navion: compile it?
[23:18] <Navion> I suppose. I was hoping to find it as a dmg but the latest I can find id 52.
[23:28] <spectralsun> c_14: that seemed to work :)
[23:29] <spectralsun> SLOMO_FACTOR is 4
[23:29] <spectralsun> so i put in directyl 0.24
[23:29] <spectralsun> 0.25 *
[00:00] --- Wed Jun  4 2014


More information about the Ffmpeg-devel-irc mailing list