[Ffmpeg-devel-irc] ffmpeg.log.20150414

[burek]

[00:00:55 CEST] <gallardo> c_14: how about dirac?
[00:01:13 CEST] <gallardo> I tried lossless h264 but colors differ a bit
[00:01:33 CEST] <c_14> What pixel format did you use?
[00:01:44 CEST] <gallardo> idk lol
[00:02:01 CEST] <c_14> Maybe try with yuv444p
[00:02:03 CEST] <gallardo> yuv420p
[00:04:12 CEST] <gallardo> c_14: use for the source or for the encoding
[00:04:20 CEST] <c_14> for the encoding
[00:05:15 CEST] <gallardo> c_14, like this?: ffmpeg -i source.y4m -pixel_format yuv444p -c:v libx264 -preset veryslow -qp 0 -y test.mkv
[00:06:12 CEST] <c_14> -pix_fmt, but other than that looks ok. Though if you want a fast lossless encoding, you might not want -preset veryslow
[00:09:08 CEST] <gallardo> yes I was just testing lol
[00:09:21 CEST] <gallardo> I see that my problem comes from the source
[00:09:38 CEST] <gallardo> I mean, from the vncrec -movie outpet
[00:09:42 CEST] <gallardo> output*
[00:31:49 CEST] <Airwave> Hi. I'm trying to decode a g2m file using ffmpeg. It's shown as "Stream #0:2(nob): Video: none (G2M5 / 0x354D3247), 2560x1440, 229 kb/s, 1k tbr, 1k tbn, 1k tbc". From what I see, ffmpeg should have support for G2M5, but it still says "No decoder for stream #0:2, filtering impossible".
[00:33:29 CEST] <klaxa> !pastebin Airwave
[00:33:35 CEST] <klaxa> err
[00:33:47 CEST] <Airwave> Sure, just a sec.
[00:34:26 CEST] <Airwave> fflogger: Well, it's just one line:
[00:34:40 CEST] <Airwave> ffmpeg -i foo.g2m bar.mkv
[00:34:40 CEST] <klaxa> the complete output is hardly just one line
[00:34:48 CEST] <Airwave> Ah, you want the output too of course.
[00:34:52 CEST] <Airwave> Sec.
[00:35:20 CEST] <jxself> It is helpful to know details, such as the FFmpeg version and configuration. This is why the complete output is desired. ;)
[00:36:22 CEST] <Airwave> I understand.
[00:36:32 CEST] <Prelude2004c> hey guys.. quick question.. i have a transport stream with 2 audios ( audio 1 = 5.1 and audio 2 = descriptive video ) .. i have tried in ffmpeg to do -c:a:1 and -c:a:2 and on the -c:a:1 i would output one to stereo 2 channel and another to 5.1.. how does one do that ?
[00:36:40 CEST] <Prelude2004c> some reason its not really happening
[00:36:48 CEST] <Airwave> klaxa: http://fpaste.org/210523/
[00:38:10 CEST] <jxself> My first thought: the 2.4 series may be too old.
[00:38:22 CEST] Action: DragonsLord is away: AWAY
[00:38:42 CEST] <DragonsLord> Good night & sweet dreams :)
[00:39:38 CEST] <klaxa> Airwave: can you run ffmpeg -codecs and see if g2m is there as decodable?
[00:39:51 CEST] <klaxa> i ran ffmpeg -codecs | grep -i g2m
[00:40:32 CEST] <Airwave> klaxa: Whoops, I ran it that time with the distro version instead of the locally built newest version.
[00:40:44 CEST] <Airwave> Just a sec and I'll run it again, and also give you the codec info.
[00:42:44 CEST] <Prelude2004c> http://pastebin.com/gy8nWpVV
[00:42:54 CEST] <Prelude2004c> can someone have a look at this and give me a pointer on what is wrong with it
[00:42:55 CEST] <Airwave> klaxa: http://fpaste.org/210525/96496514/
[00:43:10 CEST] <Airwave> Sorry about ffmpeg's build info spam. Not sure why it's giving so much duplicate info.
[00:43:38 CEST] <Airwave> klaxa: D.V.L. g2m                  Go2Meeting
[00:43:46 CEST] <klaxa> hmm weird
[00:44:12 CEST] <klaxa> maybe try: ffmpeg -c:v g2m -i foo.g2m bar.mkv
[00:45:57 CEST] <Airwave> klaxa: I tried that earlier. I gives a lot of the following, and produces an audio-only file:
[00:46:01 CEST] <Airwave> [g2m @ 0xc243c0] Wrong magic 47324D35
[00:46:04 CEST] <Airwave> Error while decoding stream #0:2: Invalid data found when processing input
[00:47:45 CEST] <klaxa> either this specific codec is not supported (yet) or your file is corrupt
[00:49:35 CEST] <Airwave> klaxa: It was supposed to be added in https://ffmpeg.org/pipermail/ffmpeg-cvslog/2014-October/082499.html, apparently.
[00:52:36 CEST] <klaxa> oh, if you can, file a bugreport and upload a sample
[00:53:10 CEST] <Airwave> I'm afraid I can't distribute this file.
[00:54:01 CEST] <klaxa> hmm... that's a problem, heh
[00:54:20 CEST] <jxself> It becomes difficult to investigate things.
[00:55:52 CEST] <Airwave> Yeah. Sorry.
[01:14:56 CEST] <koz_> What audio format is less lossy - MP3 or AIFF? I'm asking here because you guys are likely to know.
[01:18:03 CEST] <koz_> I've started this collection of resources: https://notabug.org/koz.ross/awesome-gamedev I welcome contributions of everything!
[01:20:22 CEST] <koz_> Whoops, wrong channel.
[01:36:19 CEST] <lordkrondor> well AIFF is uncompressed, so there you go
[01:44:12 CEST] <gallardo> I just can't understand how come libx264 can encode losslessly SO FAST compared to lossless dirac or lossless vp8/9.
[01:44:23 CEST] <gallardo> Is there a reason for thatL
[01:44:24 CEST] <gallardo> ?
[04:13:56 CEST] <xxzz> gallardo: libx265 is much better
[04:14:24 CEST] <gallardo> xxzz: how about licences?
[04:18:37 CEST] <xxzz> gallardo: its available in pkgsrc so I suppose it can be used
[04:18:57 CEST] <gallardo> xxzz: what do you think of vp9?
[04:19:24 CEST] <xxzz> I can convert libx265 to mp4 in Linux but can't do that in OpenBSD
[04:19:42 CEST] <xxzz> can play HEVC with ffplay on OpenBSD
[04:20:04 CEST] <xxzz> have not tried VP9 google item
[04:20:41 CEST] <xxzz> openbsd ffmpeg is not built against libx265
[04:21:19 CEST] <xxzz> it was there in freebsd but frogs killed it
[04:25:22 CEST] <gallardo> uh
[09:39:15 CEST] <keren> I have a few videos possible mov and mp4, and I want to reduce the size of the files to the maximum, I am ok comprimising the quality to upto 80%, I could go all the way to 144px
[09:39:26 CEST] <keren> what would be the best output format ?
[09:40:07 CEST] <keren> Like a 800 mb video could go all the way to 80mb
[10:23:01 CEST] <capradma_> Hi! Since libavformat has a Jackmp depency, I'm not able to deploy release version of my software on macos
[10:29:11 CEST] <BtbN> Compile it without jackmp?
[12:30:43 CEST] <capradmar> BtbN: How?
[12:45:41 CEST] <capradmar> I wonder why no mention of Jackmp exists in ffmpeg source code: https://github.com/FFmpeg/FFmpeg/search?utf8=%E2%9C%93&q=jackmp
[12:49:45 CEST] <c_14> capradmar: afaik jack is only an optional indev
[12:50:38 CEST] <c_14> https://ffmpeg.org/ffmpeg-devices.html#jack
[13:17:41 CEST] <capradmar> c_14: Why is it by default when installed with brew?
[13:18:08 CEST] <c_14> Ask whoever manages the brew file.
[13:20:47 CEST] <capradmar> I created a PR: https://github.com/Homebrew/homebrew/pull/38633
[13:21:17 CEST] <capradmar> compiling without jack seems to solve the problem
[13:28:00 CEST] <JD___> Hi. I'm having some problems with some files that are denied by a quality control software called Baton QC. This is the error: "ref_idx_l0 value was out of range. Found 5. Allowed 0 to 3 Inclusively".
[13:28:52 CEST] <JD___> I'm using a Git build from February and the libx264 encoder. Does anyone know how to limit the ref_idx values to within the allowed ranges?
[14:28:42 CEST] <Anoia> JD___: from what I can see, the allowed range in the spec is at least 0 to 15
[14:31:00 CEST] <Anoia> (note that the size of the ref_idx_l0 array is 0 to 3)
[14:32:02 CEST] <Anoia> the actual limit of the values in it are defined by oither fields int he header
[16:04:00 CEST] <JD___> Anola: This target for the files is for hardware decoding equipment which might have limitations beyond the H.264 specification.
[16:06:30 CEST] <JD___> Are you saying that if I'm going to limit the range to 0->3, there are other header fields that needs to be adjusted? Which are those?
[16:15:23 CEST] <aleb> Hi, if I specify -vcodec libx264 -preset veryfast, where is the "veryfast" preset defined?
[16:18:27 CEST] <JEEBsv> aleb: in libx264 itself
[16:18:44 CEST] <JEEBsv> libavcodec should just pass that string to libx264
[16:39:26 CEST] <Anoia> hmmm
[16:39:53 CEST] <Anoia> does zeranoe usually do Windows builds for the releases?
[17:26:25 CEST] <dericed> Hi all, as some of you know, MediaArea applied to http://www.preforma-project.eu/ to work on FFV1, Matroska, and LPCM. Happy to say the proposal was successful and over the next 22 months we'll be working on designing conformance checkers, implementation checkers, and metadata wranglers for these formats. Part of the project includes efforts to make progress at further standardizing FFV1 and Matroska throu
[17:26:31 CEST] <dericed> gh IETF. I can post info to ffmpeg-devel soon, but wanted to post this good news to the irc first. Dave
[18:42:45 CEST] <sybariten> evening
[18:43:34 CEST] <sybariten> i have "mts" files that i want to use in the editing suite Sony Vegas. They are 1080i and i want to get rid of the interlacing so i have been trying to reencvode them with ffmpeg. But the result files arent popular with Vegas, which chokes.
[18:43:59 CEST] <sybariten> Now, i've done some test encodes with more or less the same commadn lnie, on MJPEG files and those outputs are OK with vegas
[18:44:39 CEST] <sybariten> how do i pinpoint what the difference is , that makes vegas complain? I should mention that the mts files have sound, whereas the MJPEG files do not
[19:20:13 CEST] <c_14> ffprobe the input file, the output file, and the files vegas accepts
[19:43:17 CEST] <sybariten> c_14: yeah i think i'm on to something now, thanks. Tried a different command line.....
[20:15:58 CEST] <sybariten> can someone explain the -c: concept....      more specifically i have this line   ffmpeg.exe -i ./00028.mts -y -map 0:0 -c:0 copy -map 0:1 -c:1 aac -b:1 128k -ac:1 2 -cutoff 190 00 -strict experimental -sn -s hd720 $(date +%Y%m%d-%H%M%S)_output.MP4
[20:16:57 CEST] <sybariten> i'm guessing video stream is just copied
[20:19:37 CEST] <klaxa> the -c concept is that you use -c (or -codec) to specify a codec for a stream
[20:19:59 CEST] <klaxa> you can also use the letters a, v, or s to specify a type rather than a stream-number
[20:20:11 CEST] <klaxa> for example -c:v sets the videostream codec
[20:20:17 CEST] <klaxa> a is for audio and s for subtitle
[20:20:28 CEST] <__jack__> or none to use the same codec for all stream (sample: -c copy)
[20:21:11 CEST] <klaxa> right, that's also possible
[20:21:41 CEST] <sybariten> okay. I'm guessing -sn in my example here means, avoid any subtitles
[20:22:26 CEST] <klaxa> yes
[20:22:38 CEST] <sybariten> but, i also have an -s hd720 which doesnt really do anything, presumably since i already had -c:0 copy . What's the prefered way to resize this stream down to 720? Well actually i also want to deinterlace
[20:23:14 CEST] <sybariten> i'm guessing i want to use yadif and scale in tandem.. but how
[20:23:23 CEST] <kepstin-laptop> sybariten: if you want to change the video, then you have to reencode.
[20:23:24 CEST] <__jack__> can you resize without reencode ?
[20:23:34 CEST] <sybariten> kepstin-laptop: thats OK with me
[20:24:10 CEST] <kepstin-laptop> sybariten: in that cause you'd use e.g. '-vf yadif,scale=hd720' then '-c:v libx264' (or something else), with appropriate encoder quality settings.
[20:25:00 CEST] <sybariten> aha, and what about -vcodec  ?
[20:25:25 CEST] <kepstin-laptop> right now you're reencoding that to aac, dunno if that's what you want.
[20:25:44 CEST] <__jack__> -vcodec: same as -c:v, more ugly (for me)
[20:26:26 CEST] <kepstin-laptop> oh, for some reason i read vcodec = acodec, lol
[20:27:06 CEST] <sybariten> yeah, i do not have vcodec in the lines i've pasted here, but i've seen it in other lines ive used. I was wondering how it relates to -c:v
[20:27:17 CEST] <klaxa> it's the same
[20:27:39 CEST] <sybariten> ok...  i think i'd prefer to do this with as little shorthand as possible
[20:27:42 CEST] <klaxa> using -c looks cleaner (to me) though
[20:28:20 CEST] <kepstin-laptop> the non-shorthand version of "-c:v" is actually "-codec:v" :)
[20:30:12 CEST] <sybariten> okay, thats cool....     and -vf becomes -filter:v ?  Also, what about the -map 0:0 i had in my example.... was it there just to make things more confusing?
[20:30:59 CEST] <kepstin-laptop> you can use map to select a subset of streams if your file has multiple
[20:31:18 CEST] <kepstin-laptop> e.g. if there are two audio tracks but you only want one, then you'd need to use -map to select the streams to keep
[20:31:31 CEST] <sybariten> ok
[20:31:34 CEST] <sybariten> overkill for this
[20:31:40 CEST] <kepstin-laptop> (if you don't use map, ffmpeg copies all streams, if you do use map, ffmpeg copies *only* the mapped streams)
[20:31:48 CEST] <sybariten> i see
[20:34:35 CEST] <sybariten> ok, so far i have   ffmpeg.exe -i ./00028.mts -filter:v yadif,scale=hd720 -c:1 aac -b:1 128k -ac:1 2 -strict experi mental $(date +%Y%m%d-%H%M%S)_output.MP4     but i'm guessing i need to do seomthign about the -c -b and -a settings there
[20:38:14 CEST] <kepstin-laptop> sybariten: since you're encoding to mp4, ffmpeg selects some default video codec and options for you, which are h264 (libx264) and a moderate -crf setting. It should just work like that, although you might want to tweak the settings more.
[20:39:59 CEST] <sybariten> yeah, hm, it lookied horriblyl broken
[20:40:01 CEST] <sybariten> in vegas
[20:40:44 CEST] <sybariten> gah... sigh... it actually looked pretty normal with VLC ... so the strange artefacts comes from vegas...
[20:40:55 CEST] <sybariten> this video project is killing me...  oh well
[20:41:20 CEST] <kepstin-laptop> did you accidentally get a build with a 10bit x264? that could cause fun issues.
[20:41:41 CEST] <sybariten> hm, dunno, can i try some simpler codec perhaps ?
[20:41:53 CEST] <kepstin-laptop> sybariten: pastebin your ffmpeg command output somewhere please.
[20:43:38 CEST] <sybariten> OK, this is what i used just now recently.  http://pastebin.com/Ugm6eB9S    as i said, the video opens in vegas, but it looks much worse there than in a normal video player. At least in its preview window. The video's all "jagged"... not the normal compression artefacts im used to seeing
[20:45:24 CEST] <kepstin-laptop> that's just the command line, I want to see the output from when it runs.
[20:45:28 CEST] <sybariten> oh
[20:47:34 CEST] <sybariten> http://pastebin.com/6YrCXXyZ
[20:55:13 CEST] <kepstin-laptop> hmm. that looks all about like I'd expect; the output video should be fine in most players
[20:55:20 CEST] <sybariten> I see, ok
[20:55:53 CEST] <sybariten> I am following a thread now on videoforum, about these issues (mts/avhcd video, NLE, ffmpeg) and will maybe pick up some tip there
[20:56:23 CEST] <sybariten> as it turns out, the avchd format really isnt edited very often
[21:03:30 CEST] <JD___> sybariten: Your goal is to deinterlace the 1080i master video file and import the result into Sony Vegas for editing?
[21:04:38 CEST] <sybariten> JD___: exactly
[21:06:06 CEST] <JD___> I'm not familiar with Vegas import formats, but I would personally have tried a mpeg ts with audio copied and mjpeg image codec at original dimensions
[21:07:19 CEST] <sybariten> i am actually trying out some mpeg (2?) solution as we speak
[21:07:33 CEST] <sybariten> ffmpeg.exe -i ./00028.MTS -filter:v yadif,scale=hd720 -vcodec mpeg2video -r 29.97 -qscale 4 -qm in 1 -intra -flags +ilme+ildct -top 1 -acodec ac3 -ab 384k -f mpegts test2.m2ts
[21:07:59 CEST] <JD___> ffmpeg -i 1080i.mts -vf yadif -c:a copy -c:v mjpeg 1080p.ts
[21:08:36 CEST] <JD___> Your command does a lot of work and going to cost you a lot of quality
[21:09:03 CEST] <sybariten> aha?
[21:09:11 CEST] <sybariten> OK i'll try yours
[21:09:26 CEST] <JD___> Try mine and see if Vegas imports it. File size will be huge, but it should be fast and keep quality acceptable
[21:09:48 CEST] <sybariten> one thing though.... i think that the bleeding file actually isnt 1920, really ... its a 1440 file with some PAR fuckery
[21:10:10 CEST] <sybariten> so i dunno if i'm gonna need an -aspect at some point. But i'll try
[21:10:29 CEST] <JD___> Shouldn't matter, you can change that when rendering in Vegas
[21:11:44 CEST] <sybariten> ok...
[21:12:39 CEST] <sybariten> yeah i dunno why the -r 29.97 option was suggested really... i feel that i wanna mess as little as possible with frame rate
[21:13:23 CEST] <sybariten> this is what ffmpeg -i  says in my case.    Stream #0:0[0x1011]: Video: h264 (High) (HDMV / 0x564D4448), yuv420p, 1440x1080 [SAR 4:3 DAR 16:9], 25 fps, 25 tbr, 90k tbn, 50 tbc
[21:13:53 CEST] <DragonsLord> Good evening!
[21:14:38 CEST] Action: DragonsLord is back (gone 20:36:16)
[21:17:15 CEST] <JD___> sybariten: did my command work? Is it encoding?
[21:23:58 CEST] <sybariten> nope
[21:24:25 CEST] <sybariten> vegas appearently didnt like mjpeg... ? strange, ive used that before. But not what i got from your command
[21:25:11 CEST] <sybariten> i just did this an mpeg2video one though, and rendered the vegas project out again... the outcome isnt stellar but quite OK
[21:43:21 CEST] <aleb> If I specify "-vcodec libx264 -preset veryslow", what bitrate will be used? How about if I specify "veryfast"? I see x264 does not include a bitrate value for these.. http://git.videolan.org/?p=x264.git;a=blob;f=common/common.c;h=de5d441b90793c47031cb6833b35e192aff0e39c;hb=HEAD#l223
[21:43:52 CEST] <aleb> JEEBsv, btw, thanks for your answer earlier :)
[21:44:46 CEST] <JEEBsv> aleb: libx264 defaults to its default - which is CRF 23
[21:44:56 CEST] <JEEBsv> that is not a specific bit rate but rather a rate factor
[21:55:29 CEST] <JD___> sybariten: Sorry about that. One last hail Mary is that Vegas uses VFW for decoding and that by installing something like ffdshow, you'll be able to import mjpeg and a lot more into Vegas.
[22:05:10 CEST] <sybariten> JD___: yeah... you know what i discovered though?  (this is kinda embarrassing)
[22:06:09 CEST] <sybariten> removing interlacing was just my own instinct from seeing the material with VLC, and it looked like shite with those jagged edges... turns out that if using the source files in vegas and just rendering to 1280x720p, it looks quite alright actually
[22:06:45 CEST] <sybariten> seems like my machine handles "HD" files in editing too so i might not need to rescale them beforehand
[22:16:43 CEST] <JD___> sybariten: user error = best error
[22:16:51 CEST] <sybariten> yeah
[22:17:07 CEST] <JD___> Easily corrected
[23:26:01 CEST] <DragonsLord> Good night, sleep fine!
[23:28:26 CEST] Action: DragonsLord is away: AWAY
[23:50:53 CEST] <kyleogrg> Hey.  I am converting 29.97i VOB files to 23.976p x264.  The VOBs were originally film, so I am doing detelecine.  However, as it encodes, I am getting occassional messages saying "Frame X at X is still interlaced."
[23:51:38 CEST] <kyleogrg> Is there a best way to detelecine?  I am doing fieldmatch and decimate.
[23:53:09 CEST] <c_14> Try adding a yadif between the deteceline and the decimate
[23:53:44 CEST] <lordkrondor> i've found that if you put the yadif between the fieldmatch & decimate, the video is faster than the audio. I usually have put the yadif after decimate.
[23:53:58 CEST] <lordkrondor> or done a pullup,idet,yadif combo
[23:54:14 CEST] <kyleogrg> what's idet?
[23:54:24 CEST] <lordkrondor> interlacing detection
[23:54:53 CEST] <kyleogrg> so what usually delivers the best results?
[23:54:57 CEST] <kyleogrg> for you
[23:55:30 CEST] <lordkrondor> i've used both, some SD stuff worked better with pullup, the HD stuff i've done looked better with fieldmatch
[23:55:37 CEST] <lordkrondor> i'd just play around and see how it works
[23:55:49 CEST] <kyleogrg> sure ok
[23:55:51 CEST] <kyleogrg> thanks
[00:00:00 CEST] --- Wed Apr 15 2015