[Ffmpeg-devel-irc] ffmpeg.log.20180216

[burek]

[00:23:56 CET] <kuahara> I am getting my ass handed to me with ffmpeg.  I am trying to run the following:   for "%%a" in ("dir %userprofile\video\input /b") do c:\datapoint\ffmpeg\bin\ffmpeg -i %%a -c:a copy -c:v copy -y "%userprofile%\desktop\video\output\"
[00:24:13 CET] <kuahara> basically just want to copy whatever video is in the input folder to .mp4 in the output folder
[00:24:38 CET] <kuahara> just copying the audio/video from any container type; don't really care what it is, to .mp4
[00:25:10 CET] <kuahara> When I run that command as is, there is no output on the cmd line.  Just drops down 2 lines as if I smacked the enter button twice.  What am I doing wrong?
[00:40:29 CET] <SortaCore> kuahara: for expects a folder, not the return of dir, from what I've seen
[00:41:06 CET] <SortaCore> https://ss64.com/nt/for2.html
[00:45:55 CET] <nicolas17> what is memory usage in x264 encoding proportional to? does it keep a whole GOP of frames in memory, for example?
[00:46:09 CET] <nicolas17> (such that larger GOP -> directly more memory usage)
[00:47:13 CET] <JEEB> it has a lookahead
[00:47:22 CET] <JEEB> and then the amount of the references
[00:47:41 CET] <nicolas17> ah rc-lookahead?
[00:47:46 CET] <JEEB> yes
[00:48:49 CET] <nicolas17> I was also calculating 1920*1080*3 bytes per frame but I guess that's wrong if the frames in memory are YUV and chroma-subsampled
[00:49:27 CET] <JEEB> well, you can encode 4:4:4 RGB if you want :D
[00:49:44 CET] <nicolas17> well you *could* :D
[00:50:07 CET] <JEEB> oh, and frame threads
[00:50:21 CET] <JEEB> since you're doing N pictures at the same time they must be in memory
[00:50:30 CET] <JEEB> (the input pictures)
[00:52:24 CET] <nicolas17> if x264 config line says "threads=1 lookahead_threads=1 sliced_threads=0" does that mean it's using one thread? or does it mean "multithreading = on"?
[00:53:08 CET] <JEEB> I think threads=1 is single threaded
[00:53:22 CET] <JEEB> and then there's I guess one thread for lookahead
[00:53:29 CET] <SortaCore> oh kuahara: yeah, you can use FOR like that, but you need FOR /F
[01:58:49 CET] <solidus-river> hey all, i'm trying to encode 1080p video frame by fraem in rgb format via ffmpeg and get no meaningfull errors however my call to avcodec_encode_video2 returns 0 and i get no packet back
[01:58:56 CET] <solidus-river> where do i start digging to figure out whats going wrong here?
[02:00:17 CET] <DHE> some codecs buffer frames. libx264 is notorious for buffering multiple seconds worth (assuming reasonable framerates) when using defaults.
[02:01:03 CET] <nicolas17> with a video several seconds long and "-preset slow" I have seen encode_video2 return no frames *ever*
[02:01:04 CET] <DHE> after a while it'll start doing one-in-one-out. when you're done feeding frames, feed it NULLs instead of frames and it'll keep spitting out the buffered frames as it finishes
[02:01:22 CET] <nicolas17> I got all the frames at once when I finished
[02:01:31 CET] <DHE> also avcodec_encode_video2 is the old API. if you're using newer ffmpeg versions consider using the new API
[02:08:29 CET] <solidus-river> DHE, i am using the latest version, where can i find docs on the new api? all the examples / docs i found use avcodec_encode_video2
[02:09:07 CET] <DHE> search the doxygen for avcodec_send_frame and avcodec_receive_packet   (when decoding, send_packet and receive_frame)
[02:09:17 CET] <solidus-river> i am using h264 so buffering might be a real possibility, thats annoying about the buffering
[02:09:21 CET] <DHE> match the ffmpeg version you're using though
[02:09:54 CET] <nicolas17> solidus-river: you can reduce buffering but reduce quality too
[02:09:56 CET] <solidus-river> DHE, thanks, also i'm not sure what FORMAT_YUV420 etc is about, I'm aiming to fid it rgb uint8 valuse, is there a better setting to use for that or what does that determine / where can i find more info on it
[02:10:12 CET] <DHE> solidus-river: short of multi-pass encoding, x264 does it for rate control decision making. you can av_dict_set(&encoderdict, "tune", "zerolatency", 0); if you really really want, but it's usually a big loss for quality
[02:11:10 CET] <DHE> the most common H.264 colourspace is yuv420. most hardware players (including quicktime) only support it.. so you usually get a warning if you're not using it and not setting bits to indicate you're sure about what you're up to
[02:13:10 CET] <solidus-river> ok, the example i shamelessly pulled uses yuv420 so i'm probably ok there, wanted to understand what that meant though
[02:13:15 CET] <solidus-river> is that the colorspace of each frame :?
[02:14:17 CET] <nicolas17> yes
[02:14:37 CET] <nicolas17> if you feed it something else (like RGB), ffmpeg will convert it for you
[02:16:36 CET] <DHE> ffmpeg yes. the API, no. you have to provide it the right colourspace, or use a converter.
[02:17:05 CET] <DHE> you could use libswresample or the libavfilter APIs
[02:17:05 CET] <nicolas17> I don't remember having to add a converter filter myself...
[02:18:29 CET] <nicolas17> oh my input was yuv already ^^'
[02:21:01 CET] <solidus-river> should i be using avcodec_alloc_context3 or is there a new api for that :|
[02:21:40 CET] <solidus-river> oy,n ow i'm worried these are all old api's is there a new api / latest and gretest ref for encodign a simple vid out there?
[02:23:08 CET] <solidus-river> oh well, i'll assume a context is a context
[02:23:31 CET] <DHE> yes a context is a context
[02:23:42 CET] <DHE> and as long as avcodec_encode_video2 is available it's safe to use
[02:24:08 CET] <nicolas17> I wrote an entire transcoding loop just because I wanted to mess with the PTSs
[02:24:56 CET] <nicolas17> then I found mkvmerge would let me do it from the command line -.-
[02:25:03 CET] <nicolas17> oh well it was educational
[02:25:16 CET] <DHE> indeed
[02:25:40 CET] <DHE> I do have my own API-based application as well. learned a lot about ffmpeg
[07:10:45 CET] <solidus-river> i'm writing out an mpeg1 video, how do i tell it to not use compression or use the nicest compression possible :? I'm using the avcodec api's but ausing sws to convert from rgb to yuv420
[07:11:33 CET] <nicolas17> I don't think mpeg1 has a lossless mode
[07:12:15 CET] <solidus-river> should i switch over to h.264 then?
[07:12:20 CET] <solidus-river> and just use film mode :?
[07:15:24 CET] <furq> film doesn't really have anything to do with compression, it just tweaks the detail retention bias
[07:15:47 CET] <furq> just use a low crf value or something
[07:16:26 CET] <noob> Hello. I just wanted to suggest allowing cryptocurrency donations. E.g. in Bitcoin Cash and Monero.
[07:16:26 CET] <furq> -qp 0 is lossless but you'll obviously end up with a massive file
[07:17:52 CET] <noob> Is this the right place to suggest crypto donations or should I go to ffmpeg-devel or join some mailing list?
[07:19:17 CET] <solidus-river> furq, thanks!
[12:08:11 CET] <relaxed> G/15.9G]
[13:19:52 CET] <CounterPillow> Cut a new release already you bums
[13:20:23 CET] <durandal_1707> Not gonna happen
[13:21:15 CET] <CounterPillow> Whose dick do I have to suck?
[13:21:59 CET] <durandal_1707> ask on devel, this is user help channel
[14:15:18 CET] <kir> Hi, I am using libavformat to write opus audio stream to a .webm file (audio only). I am passing each opus packet correctly (in my case 1 opus frame per packet). But when I call av_write_interleaved_frame() it is returning -22 value and throwing the message " Only VP8 or VP9 video and Vorbis or Opus audio and WebVTT subtitles are supported for WebM." I have configured the ffmpeg build with libopus only.
[14:18:16 CET] <AppAraat> hi, is it usually normal for Youtube to take an entire day to process a video like this? http://ix.io/KRh
[14:19:45 CET] <sfan5> >Format profile                           : High 4:4:4 Predictive at L4
[14:19:49 CET] <sfan5> >Chroma subsampling                       : 4:2:0
[14:20:05 CET] <sfan5> something is odd about your file
[14:20:22 CET] <sfan5> other than the 4:4:4 subsampling, files like this should process fine
[14:28:26 CET] <AppAraat> it's an output from Blender. I set h.264 encoding with lossless profile
[14:34:35 CET] <JEEB> sfan5: high 4:4:4: predictive generally means lossless
[14:34:46 CET] <JEEB> since it's the only profile that lets you do lossless coding :)
[14:34:59 CET] <sfan5> makes sense
[14:35:03 CET] <DHE> isn't there an RGB profile as well?
[14:35:58 CET] <JEEB> I'm pretty sure it's only included with that one?
[14:36:05 CET] <JEEB> although that could be because it's generally utilized together with lossless
[14:36:29 CET] <JEEB> although I guess it depends on the features you require since RGB/YCbCr is just a field in the metadata
[14:36:32 CET] <JEEB> it doesn't affect coding
[14:36:41 CET] <JEEB> (you could in theory encode 4:2:0 RGB lol)
[14:36:46 CET] <DHE> eww...
[14:42:22 CET] <pmjdebruijn> JEEB: artifacting would be much worse though :D
[14:45:56 CET] <io__> what codecs and settings to use to produce the best video quality to be uploaded to youtube. my input is an avi fullframes uncomressed. I used h264 with crf 0, the file works great on my pc, but on youtube, it just is not as sharp at all...
[14:49:39 CET] <doomedary> Hey
[14:50:33 CET] <doomedary> 6a97ba521 breaks AVIO_FLAG_DIRECT completely
[14:50:38 CET] <doomedary> can anyone verify this?
[14:51:11 CET] <doomedary> and how should I proceed? release/3.4 when reverting that commit makes my code that relies on AVIO_FLAG_DIRECT work again flawlessly
[14:52:31 CET] <doomedary> also, tbqh, that commit looks a bit retarded to me in any case
[14:56:24 CET] <doomedary> (symptom is, with AVIO_FLAG_DIRECT and this commit, av_write_frame and av_interleaved_write_frame will fail with EIO which is also rather odd, at least whenever a packet submitted does not fit the new packet_size requirements)
[14:56:48 CET] <doomedary> (which is funny because I had a MJPEG stream work or not depending on how small the compression gets a single frame)
[15:09:00 CET] <sine0> waddup beyotches
[15:18:10 CET] <sine0> guys I ask here, only for the knowledge and crossover, googling this will result me in a million blog posts and stupid shareware. Im looking for the best, opensource, free media streaming over a lan solution
[15:18:28 CET] <bindi> emby or plex
[15:18:33 CET] <bindi> well opensource, i guess emby
[15:18:43 CET] <sine0> is that on windows
[15:18:46 CET] <sine0> i will look
[15:19:17 CET] <bindi> you can run emby with mono
[15:20:07 CET] <bindi> but what do you actually need from the program?
[15:20:22 CET] <sine0> ok emby is incoming on my 2012 r2 server.
[15:20:39 CET] <bindi> you could just set up a samba share, maybe even a simple http server to serve the files and attach that to your favourite media player on the network
[15:20:42 CET] <sine0> I just need to be able to stream media in my house from my windows server, things i dump or download there.
[15:20:53 CET] <bindi> plex is the most polished one really
[15:20:57 CET] <bindi> i tried emby a few times but i keep goin back to plex
[15:24:49 CET] <sine0> oh wow its working, streaming in browser, local netflix style. thanks bro
[15:24:58 CET] <sine0> if i get issues i will try out plex
[15:29:20 CET] <Nacht> Hhmm. I take it HLS doesn't like it when you got a list of segments each starting with the same PTS ?
[15:42:42 CET] <gh0st3d> Hey guys, I'm trying to concatenate two mp4 files. I have it working with the -filter_complex method, but I think my understanding is the other methods would be faster? I may be wrong there. Anyways, when I try the other method, the video times get added, but the second clip doesn't play on the output video
[15:43:05 CET] <gh0st3d> Here's a paste with the command I'm running to merge the two files. Then I included the ffprobe with only the differences between the  files showing. https://apaste.info/HVkH
[15:43:34 CET] <gh0st3d> I'm pretty new to this so I'm not sure which if any of these differences would cause the concatenate issue
[15:44:30 CET] <sfan5> the pixel formats don't match
[15:44:40 CET] <sfan5> neither do the same rates of the audio
[15:45:00 CET] <sfan5> sample*
[15:45:14 CET] <gh0st3d> Cool, I should be able to figure that out then! Thank you!
[15:45:37 CET] <sfan5> oh and also the channel layout of the audio
[15:45:48 CET] <sfan5> you will have to re-encode this to merge it, basically
[15:46:24 CET] <gh0st3d> my understanding is if those settings all matched up in the first place, they wouldn't have to get re-encoded to merge. Is that correct?
[15:47:00 CET] <sfan5> if those settings match, doing "primitive" concat usually just works, yes
[15:47:02 CET] <gh0st3d> The second clip is generated programmatically, then they get merged. So I'm thinking if I get the settings right when making the 2nd clip the merge will be much faster
[15:47:38 CET] <gh0st3d> Gotcha. Thank you very much for the help
[15:56:36 CET] <Fyr> guys, are FFPROBE/FFMPEG able to retrieve language of streams from Bluray?
[15:57:04 CET] <Fyr> I did: ffprobe -i bluray:path, it has shown the streams.
[15:57:27 CET] <Fyr> but it didn't show languages of the streams.
[16:28:29 CET] <JEEB> Fyr: depends on if the thing called "bluray" (probably uses libbluray) can read the data from somewhere (I'm not sure if that's actually available)
[16:28:38 CET] <JEEB> the MPEG-TS itself has no language codes with blu-ray as far as I know
[16:28:45 CET] <JEEB> the playlists I think only have chapters
[16:28:55 CET] <JEEB> so it might just be that there is no language data anywhere
[16:29:12 CET] <Fyr> JEEB, players show languages.
[16:29:19 CET] <JEEB> yes but isn't that in the menus?
[16:29:22 CET] <alexpigment> isn't the language in the streams itself?
[16:29:23 CET] <Fyr> it means, there are language codes somewhere.
[16:29:29 CET] <JEEB> alexpigment: http://up-cat.net/p/9cff2231
[16:29:37 CET] <JEEB> this is one of the blu-rays I have
[16:29:45 CET] <JEEB> it might be in the playlist but I'm not sure :P
[16:29:47 CET] <alexpigment> well, lemme rephrase
[16:29:48 CET] <JEEB> I would have to check the spec
[16:29:54 CET] <alexpigment> you *can* set the stream language
[16:29:57 CET] <alexpigment> usually in audio
[16:30:14 CET] <JEEB> yea, not sure if BD spec prohibits you from setting the metadata
[16:30:18 CET] <JEEB> (in MPEG-TS)
[16:30:23 CET] <alexpigment> hmmm
[16:30:27 CET] <JEEB> or if they require the metadata to be somewhere else
[16:30:29 CET] <alexpigment> i can test this
[16:30:45 CET] <JEEB> if I get home I can poke at the spec for shits and giggles
[16:31:39 CET] <JEEB> most of the language stuff I've seen on blu-rays has been just in the menus (but my memory might just be bad)
[16:31:54 CET] <alexpigment> it could be there too - i'm not sure
[16:32:08 CET] <JEEB> and menus are usually random java crap and not standardized anyways
[16:32:10 CET] <alexpigment> but i'm testing the audio right now just for my own curiosity
[16:32:17 CET] <JEEB> so you can't exactly parse the menus
[16:32:24 CET] <alexpigment> true
[16:32:43 CET] <alexpigment> well, in the pro world anyway
[16:33:05 CET] <alexpigment> i use bdmv-style menus, because that's what all consumer software makes
[16:33:28 CET] <JEEB> yup
[16:33:31 CET] <JEEB> that was what I was talking about
[16:33:38 CET] <JEEB> it's not exactly metadata :)
[16:34:32 CET] <alexpigment> well, i just at least confirmed that the audio stream language is maintained on blu-ray
[16:34:36 CET] <alexpigment> fwiw
[16:35:00 CET] <alexpigment> i'm using ac3 though
[16:35:10 CET] <alexpigment> maybe the blu-ray pcm is different
[16:35:29 CET] <alexpigment> i could also test that, but i don't care enough about this to go through the additional effort ;)
[16:37:29 CET] <alexpigment> JEEB: speaking of specs
[16:37:42 CET] <alexpigment> are there publicly available specs on how to create BDMV menus?
[16:38:09 CET] <alexpigment> it seems like most people license their menu creation from sonic
[16:38:15 CET] <JEEB> no
[16:38:15 CET] <alexpigment> or whatever company that makes scenarist
[16:38:27 CET] <alexpigment> i didn't think so, but i figured i'd ask since it's been a few years
[16:38:48 CET] <alexpigment> unfortunately, AVCHD doesn't support java menus, so BDMV is the only way to get menus on them
[16:38:59 CET] <alexpigment> it would be nice to know how to make them without a third party tool
[16:39:04 CET] <alexpigment> oh well
[16:39:13 CET] <sfan5> "unforunately doesn't support java [...]"
[16:39:17 CET] <sfan5> never thought i'd hear that
[16:39:34 CET] <alexpigment> well, if you know how to make java menus but not bdmv, it's unforunate ;)
[16:46:56 CET] <Nacht> Ok. I'm having trouble to understand something. I take an .ogg file. Transcode it to acc mpegTS. The input file has a PTS of pkt_pts=-1 in it's first frame. Yet the mpegTS just keep starting with pkt_pts=126000
[16:47:16 CET] <Nacht> Trying to manipulate it with asetpts doesn't seem to have any logic to it either
[20:42:13 CET] <marcurling> Hello, which #chanel would you recommend to have counsel on a graphic board, please?
[20:44:12 CET] <sine0> #hardware
[20:46:50 CET] Action: marcurling thanks sine0
[21:01:48 CET] <alexpigment> marcurling: out of curiosity, what is it for?
[21:01:52 CET] <alexpigment> gpu encoding?
[21:01:55 CET] <alexpigment> playing games?
[21:02:04 CET] <alexpigment> multi-monitor setup?
[21:03:03 CET] <alexpigment> either way, it's a terrible time to buy a good one :(
[22:25:54 CET] <lyncher> hi. I'm trying to add a new audio stream to a mpegts with libavcodec but I'm having some issues with PTS/DTS
[22:26:24 CET] <lyncher> they are non monotonically increasing
[22:27:21 CET] <lyncher> how can I change the PTS from the new stream increase correctly?
[22:33:55 CET] <lyncher> is there any code sample that merges two input files?
[22:35:30 CET] <JEEB> you just need to synch their DTS and PTS before feeding to muxer
[22:35:45 CET] <JEEB> and if your AVPacket's timestamps are broken you have to fix them before muxing
[22:40:56 CET] <JEEB> also in unrelated things, anyone here played with cheap ISDB-T modulators?
[22:41:07 CET] <JEEB> I think I need one for writing a receiver driver :P
[22:41:54 CET] <lyncher> I've played with cards from TBS....
[22:41:58 CET] <lyncher> ISDB
[22:42:59 CET] <lyncher> https://www.tbsdtv.com/products/tbs6209-dvb-t2-c2-tc-isdbt-octatv-tuner.html
[22:43:25 CET] <JEEB> i have a tuner and an app that can use it :D
[22:43:45 CET] <JEEB> i want to write a driver for the linux "dvb" subsystem
[22:44:03 CET] <JEEB> but to see if i can actually set the freq
[22:44:07 CET] <JEEB> i need a test signal
[22:51:47 CET] <iive> JEEB, asking in #linuxtv might be good idea, they are writing kernel driver for dvb receiver
[22:53:11 CET] <iive> there mchehab is probably the maintainer that does the mergers
[22:53:43 CET] <JEEB> cheers
[00:00:00 CET] --- Sat Feb 17 2018