[Ffmpeg-devel-irc] ffmpeg.log.20160426

[burek]

[00:03:21 CEST] <prelude2004c_zzz> so can anyone tell me how i setup ffmpeg with -timeout -1
[00:03:42 CEST] <prelude2004c_zzz> doesn't seem to work and it keeps exiting when input is missing for a few seconds
[00:21:14 CEST] <Sirisian|Work> Does ffmpeg have issues when a device can't keep up with the reencoding? Looks like at 25 of 25 fps I don't see any issues, but if I start to stress the server by running the video on the device the encoding fps drops to 14-18 fps and after 20s to 5 minutes I see http://pastebin.com/1Q3VVyKy I plan to submit a bug, but the issue tracker is down it seems. I assume it's using a circular buffer someplace that might be running out
[00:21:14 CEST] <Sirisian|Work> of space and getting corrupted?
[00:26:14 CEST] <Sirisian|Work> Yeah this is very easy to reproduce on my end. I'm not sure the devs will feel the same way though.
[00:27:26 CEST] <__jack__> Sirisian|Work: not really an issue
[00:28:02 CEST] <__jack__> you can't hold stuff forever, it must goes out at least as fast as the input, on average
[00:28:09 CEST] <__jack__> else, it will fail
[00:29:02 CEST] <__jack__> you must either make an async encode (use the stream with fast encode, then do slower stuff, or add power
[00:29:43 CEST] <furq> you can set the buffer size for some protocols but it doesn't look like you can set it for http
[00:29:54 CEST] <furq> although either way it'll eventually fail
[00:30:44 CEST] <Sirisian|Work> Strange. It just failed after 15 minutes now even when encoding at 25 of 25 fps. Maybe that's not the issue.
[00:33:17 CEST] <Sirisian|Work> oh looks like maybe it was at 23 fps for a while.
[00:34:26 CEST] <spirou> hypothetically, would it be possible to analyze an existing music file to see what would be the lowest bitrate when making a mp3 without any quality loss?
[00:34:43 CEST] <furq> i doubt it
[00:36:38 CEST] <spirou> for the sake of argument, if you had a 64k/s mp3, then convert it to a .wav or .flac or something nondestructive, so a program could then find out that "oh! this song can be converted to 64k/s mp3 with no loss :-)"
[00:36:57 CEST] <furq> there would be loss though
[00:37:09 CEST] <spirou> anyway?
[00:38:34 CEST] <furq> you could encode it to 320kbps the second time and it would sound worse
[00:38:47 CEST] <furq> every time you do a lossy encode some of the original signal is lost
[00:39:00 CEST] <spirou> thats sad
[00:39:13 CEST] <spirou> it could be fun it such a thing was possible though
[00:39:28 CEST] <furq> that's why all of the reposted jpegs on your uncle's facebook wall have got so much noise around all of the text
[00:39:40 CEST] <spirou> and in the simples case a 100% quiet file would be compressed to 0kb/s automatically then :D
[00:40:12 CEST] <spirou> yeah thats true, same problem exist for jpg too I guess.
[00:41:30 CEST] <spirou> the new compression wastes bits when it tries hard to recreate the artifacts as good as possible thinking they are important
[00:42:09 CEST] <furq> yeah the encoder has no way of knowing the difference between signal and artifacts
[00:42:29 CEST] <furq> also fwiw if you are doing low-bitrate encodes then vorbis and opus are much better choices than mp3
[00:42:42 CEST] <furq> probably aac too but i avoid that if it's not in mp4 video
[00:42:47 CEST] <DHE> Sirisian|Work: if possible, lower the CPU requirements. for libx264 you can use a "fast" preset for example
[00:43:28 CEST] <spirou> furq: there is still problems with hardware that only accepts one format though.
[00:43:31 CEST] <DHE> second on vorbis or opus. mp3 is actually a pretty bad codec by modern standards
[00:43:45 CEST] <furq> mp3 is fine above 128kbps
[00:43:59 CEST] <furq> but it's never been able to compete with vorbis below that
[00:44:08 CEST] <DHE> spirou: what hardware? AAC, and/or Vorbis actually have fairly decent support in devices I find
[00:44:10 CEST] <furq> and opus is better than vorbis if you've got something which supports it
[00:44:20 CEST] <furq> and yeah vorbis is fairly well supported and aac even moreso
[00:44:42 CEST] <furq> granted i've never owned an mp3 player that i didn't flash rockbox onto, but they both supported vorbis out of the box
[00:45:00 CEST] <DHE> iirc vorbis was designed to be patent-free and public domain from day 1. I've seen video games use it for BGM, etc
[00:45:13 CEST] <spirou> for example, I have a mp4 film file here, that the TV don't accept. if I let ffmpeg put it in a mkv, then there is no sound :(
[00:45:20 CEST] <furq> that might be the principal use for vorbis tbh
[00:45:36 CEST] <furq> i've seen it used there more than anywhere else except maybe wikipedia
[00:45:37 CEST] <spirou> even if I let ffmpeg recompress the sound track to something else.
[00:45:58 CEST] <Sirisian|Work> DHE, yeah I saw those fast and ultrafast settings. Just tested at 30 fps encoding and the most it can do is 24 or 25 fps so definitely not going anywhere near 30. It's taking a long time to fail though sometimes.
[00:46:23 CEST] <furq> spirou: mp4 is usually aac audio, but you never know with cheap hardware players
[00:46:34 CEST] <Sirisian|Work> yeah there it failed after like 7 minutes. Kind of wish it had a way to recover.
[00:47:20 CEST] <furq> doubtless the manual just says "SUPPORT TO MP4 VIDEO EXCELLENCE QUALITY"
[00:47:51 CEST] <spirou> furq: yeah this tv is the cheaper kind, some years old. its manual say .mkv can have EAC3 / AC3 sound.
[00:48:16 CEST] <spirou> for mp4 it instead say it support "PCM/MP3"
[00:48:25 CEST] <furq> oh yeah i think you mentioned this the other day
[00:48:40 CEST] <furq> i don't think pcm in mp4 is even a thing
[00:48:44 CEST] <furq> or not an official thing, anyway
[00:49:01 CEST] <furq> doubtless some dreadful person has hacked it in
[00:49:15 CEST] <spirou> as long as I don't recompress the video, even if I remove the audio completely it don't like the file as a mp4. for mkv it don't do sound instead.
[00:51:13 CEST] <spirou> (and ywah I bet it support aac for mp4 anyway, I should test with some other film clip, up until now it have accepted anything mp4 I thrown at it)
[00:52:20 CEST] <spirou> furq: what is rockbox btw?
[00:52:52 CEST] <furq> http://www.rockbox.org/
[00:53:15 CEST] <spirou> not that I ever had any mp3 player that was easy flashable. the one I have is the small no-display kind.
[00:53:31 CEST] <kepstin> i'm still sad apple killed the ipod classic; an ipod classic with rockbox is an incredible music player :/
[00:53:32 CEST] <furq> everyone just uses their phone now anyway
[00:54:16 CEST] Action: kepstin lost his ipod classic on a bus last year, and replaced it with a 200gb sd card in his phone.
[00:55:21 CEST] <furq> i hope after you took the 200GB micro sd card out of the packet, you took a moment to marvel at its existence before putting it in your phone
[00:55:24 CEST] <spirou> I don't carry around a phone but I like my small mp3 player sometimes. I doubt there are any manufacturer name printed on it though :D
[00:56:10 CEST] <spirou> something I really would like to flash is the TV's software though!
[00:56:21 CEST] <furq> i just ended up sticking all my music on google play music
[00:56:39 CEST] <kepstin> there's probably some exploitable vulerabilities in your tv :/
[00:56:48 CEST] <furq> they seem to believe i acquired it legally, which is nice of them
[00:58:17 CEST] <spirou> simple things like, for some channel if you press the button to get program list then the tv hangs so you have to pull the power :(
[00:58:18 CEST] <furq> it's just a shame it transcodes everything to mp3 or else it'd be a nice free backup system
[00:59:23 CEST] <furq> i sure am glad i don't own a tv
[00:59:39 CEST] <spirou> hehehe
[00:59:42 CEST] <furq> i hope they still make completely dumb tvs by the time i have space for one
[01:00:21 CEST] <spirou> yeah it is better to buy a big computer monitor really.
[01:00:56 CEST] <kepstin> last time my friend bought a tv, it didn't support the video output from some ps1 games :/
[01:01:34 CEST] <TD-Linux> 240p, yeah I guess.
[01:01:47 CEST] <kepstin> yeah
[01:01:55 CEST] <furq> i would normally make a comment about how it does support the output from an emulator
[01:02:02 CEST] <furq> but psx emulators are still remarkably shit
[01:03:07 CEST] <kepstin> this guy ended up buying a 'framemeister' upscaler for it. But then he realized that most ps1 games can be played on a ps3.
[01:03:26 CEST] <spirou> hehehe
[01:03:52 CEST] <TD-Linux> the framemeister is wonderful just for the name alone
[01:03:56 CEST] <spirou> does the ps3 have some ps1-emulator built in? I don't think it is same kind of hardware...
[01:04:04 CEST] <kepstin> software emulation for ps1
[01:04:06 CEST] <TD-Linux> xiphmont recently mailed be a sony bvm though
[01:04:07 CEST] <spirou> ok
[01:04:17 CEST] <kepstin> aside from the first-generation ps3 models, some of which contained a ps2
[01:04:25 CEST] <kepstin> which of course contains a ps1
[01:04:32 CEST] <spirou> :D
[01:05:07 CEST] <kepstin> the framemeister is even better when you realize that it's japanese, and the creator originally misspelled the name as 'flamemeister'
[01:05:23 CEST] <furq> nice
[01:06:17 CEST] <furq> http://www.hyperactivebroadcast.com/public/upload_product_image/Sony_BVM_14F5E_Monitor_Hire.jpg
[01:06:21 CEST] <furq> that's a nice-looking microwave
[01:06:41 CEST] <spirou> About the tv's lack of compassion for this film, would it be that it is something wrong with the video stream that makes it gets lost or not able to use other streams too?
[01:06:48 CEST] <spirou> ffprobe say it is "Stream #0:0(und): Video: h264 (High) (avc1 / 0x31637661), yuv420p, 1280x720 [SAR 1:1 DAR 16:9], 2749 kb/s, 23.98 fps, 23.98 tbr, 24k tbn, 47.95 tbc (default)"
[01:07:11 CEST] <furq> looks fine to me
[01:07:16 CEST] <spirou> are there a way to get more information out of it?
[01:07:38 CEST] <TD-Linux> furq, this is mine http://people.xiph.org/~tdaede/bvm/20160421_0001.jpg
[01:07:54 CEST] <furq> that one doesn't even have a defrost button
[01:08:04 CEST] <TD-Linux> it doesn't have any buttons actually
[01:08:09 CEST] <TD-Linux> I can't get to the menu as a result
[01:08:28 CEST] Action: TD-Linux ordered a "remote" for it
[01:08:37 CEST] <TD-Linux> "remote" in quotes because it's a tethered RS-485 1U box
[01:08:46 CEST] <spirou> I'm kind of newbie when it comes to this stuff... I guess there is timecodes and ways it interleave the data from the streams and that, that is omporant maybe..
[01:08:58 CEST] <furq> http://ak1.ostkcdn.com/images/products/L1012155.jpg
[01:09:03 CEST] <furq> i've still got one of these kicking about
[01:09:18 CEST] <furq> except i somehow managed to scratch the antistatic coating
[01:11:08 CEST] <furq> just to clarify i mean the monitor and not the space motorcycle
[01:11:22 CEST] <spirou> TD-Linux: wonderful design! :D
[01:15:22 CEST] <spirou> I didn't know sony could make such a beautiful tv's. a deluxe model too with trinitron.
[01:25:17 CEST] <TD-Linux> spirou, if you think that's nice, check out the rear http://people.xiph.org/~tdaede/bvm/20160421_0003.jpg
[01:26:15 CEST] <spirou> is it some kind of cooling on the right?
[01:26:37 CEST] <TD-Linux> yes, for the power supplies.
[01:26:58 CEST] <TD-Linux> the inputs are all option cards that you install http://people.xiph.org/~tdaede/bvm/20160421_0006.jpg
[01:27:38 CEST] <spirou> lots of ins and outs! of a tv is always nice heh
[02:29:10 CEST] <spirou> How come, when you convert to mp3 by defaul it uses 128 kb/s but when checked with standard unix  "file"  command it thinks it's  64 kbps ?
[02:30:32 CEST] <spirou> but if I during the ffmpeg converting specify  -b:a 128k  then "file" correctly identify it as "128 kbps"
[02:33:02 CEST] <spirou> vlc also says that the using defaults mad version is 64 kbit/s
[02:33:07 CEST] <spirou> *made
[02:33:50 CEST] <DHE> could be VBR mode which targets 128k. first few chunks are silent so they compress really small. so a heuristic doesn't get it right
[02:33:56 CEST] <spirou> (while ffprobe say "Stream #0:0: Audio: mp3, 48000 Hz, stereo, s16p, 128 kb/s"  "encoder : Lavc57.16")
[02:34:45 CEST] <spirou> oh hmm...ok? (I guess I shouldn't let it use defaults anyway, but still a bit strange)
[02:36:48 CEST] <spirou> both files in my test went almost same size though 1920710 "64bit" vs 1920902 "128kbit"
[02:43:59 CEST] <spirou> Hmm yeah on the other tab VLC say it is 128 kb/s (+some actual statistics jumping up and down between 125 and 130)
[02:44:41 CEST] <Sirisian|Work> Not sure if anyone related to the site is here, but I'm trying to submit a bug report and when I registered an account it won't send me an email. It's been over an hour. Might look into that.
[02:47:36 CEST] <spirou> Sirisian|Work: it didn't got stuck in some mail filter or something? the confirm mail should be automatic sent one would think, nobody does such manually...
[02:48:05 CEST] <Sirisian|Work> Their site was down for most of the day today for me. Maybe something broke.
[02:52:13 CEST] <spirou> yeah perhaps something is still is off. I guess you should check tomorrow
[02:52:21 CEST] <spirou> it is the https://trac.ffmpeg.org/register right?
[03:04:59 CEST] <Sirisian|Work> yeah
[04:27:14 CEST] <prelude2004c_zzz> hey.. anyone know how i can increase PTS for video and audio without transcoding ?
[04:27:37 CEST] <prelude2004c_zzz> i have a video that i am segmenting with FFMpeg but... the pvr start time for recording is always -6
[04:27:40 CEST] <prelude2004c_zzz> i need to make it 1
[04:27:43 CEST] <prelude2004c_zzz> or 0
[04:27:58 CEST] <prelude2004c_zzz> i dont know how to increase pts forward without the actual transcoding thing
[04:47:19 CEST] <k_sze[work]> When using FFV1 level 1, what's the default `-context`? I don't think I can tell from the console output of ffmpeg.
[04:48:14 CEST] <k_sze[work]> It's also doesn't seem to be documented at either https://trac.ffmpeg.org/wiki/Encode/FFV1
[04:48:51 CEST] <kepstin> prelude2004c_zzz: you can try the -avoid_negative_ts output (muxer) option.
[04:50:06 CEST] <kepstin> prelude2004c_zzz: documented in https://ffmpeg.org/ffmpeg-all.html#Format-Options
[04:54:32 CEST] <prelude2004c_zzz> yup tried that
[04:55:12 CEST] <pandb> i modified the muxing.c example to continuously grab image data from an arbitrary window and encode it into a video. However, the resulting video always plays too fast.
[04:56:03 CEST] <pandb> i've set the time_base fields to (AVRational){1, 30} for the codec context and stream object for my video
[04:56:37 CEST] <pandb> and i increment the pts field for the AVPacket i use in my call to avcodec_encode_video2
[04:56:43 CEST] <pandb> i'm not sure what else to try
[04:57:56 CEST] <pandb> the functions for generating the image data and encoding the video are in separate threads, with the images placed on a queue and extracted from the queue in the video thread
[04:58:35 CEST] <pandb> if the queue is empty, then the previous image is used to encode a new frame for the video
[04:58:58 CEST] <pandb> i've experimented with putting the image-generating thread to sleep for various periods
[04:59:22 CEST] <pandb> i've been able to slow playback down to an extent, but it always plays too fast
[08:17:18 CEST] <k_sze[work]> When encoding as libx264, is there some restriction to the valid number of slices?
[08:47:55 CEST] <spiderkeys> I'm having trouble getting correct timing information from a raw h264 stream using custom AVIO contexts and can't really seem to figure out why. I've looked at an H264 analyzer and the SPS and PPS info is definitely there, but when I use find_stream_info, it only gets some of the stream info. Bitrate, timing info, all seems wrong. My format detection is at CMuxer::Update() if anyone can take a quick look
[08:47:57 CEST] <spiderkeys> https://github.com/OpenROV-Dev/geomuxpp/blob/inproc_rewrite/src/CMuxer.cpp
[08:49:30 CEST] <spiderkeys> Basically, filling a buffer with a ton of h264 frames, probing it to make sure I get the right format, then using find_stream_info to get the particulars. This is the output that I get: http://pastie.org/10813158
[08:51:50 CEST] <spiderkeys> I've tried all sorts of flags, fiddling with my camera settings, etc, but the timestamps are always equal to AV_NOPTS
[08:53:32 CEST] <spiderkeys> analyzer shows inclusion of VUI info: http://i.imgur.com/iwECnmZ.png
[08:54:47 CEST] <spiderkeys> cut off of my screenshot:  timing_info_present_flag :1    num_units_in_tick :6006    time_scale :180000    fixed_frame_rate_flag :0
[10:59:56 CEST] <navlelo_> I am trying to cache a rtsp stream and then serve the cached stream using ffserver... I am struggling a bit to find the right options for this though. Does anyone know how to do this?
[11:52:21 CEST] <otila> I use libx265 to encode movie, -x265-params crf=15:nr-inter=250:nr-intra=250  , x265 debug line shows correct CRF value, but it does not matter what crf I use, 15 or 30,.. it creates output file with the exact bitrate, about 60 Mbit/s
[11:52:31 CEST] <otila> could be also x265 bug
[13:44:41 CEST] <TenLeftFingers> Can anyone give me a command to do a conversion from mp3 to ADP4? Searching for ADP4 in the man pages turned up nothing but I do see it mentioned in some sources on git.
[14:07:32 CEST] <mad_ady> hey guys
[14:07:55 CEST] <mad_ady> @JEEB: remember I asked yesterday how I could trigger ffserver to start my ffmpeg stream when the client connects?
[14:08:11 CEST] <mad_ady> well, I wrote my own workaround: https://github.com/mad-ady/ffserver-trigger/blob/master/ffserver-trigger.pl
[14:08:18 CEST] <mad_ady> it's good enough for my needs
[14:09:24 CEST] <Wader8> hello
[14:10:30 CEST] <Wader8> is it possible to make FFMPEG show status console output for audio and video separately, the bitrate stuff, as it's quite confusing when it's showing bitrate for the whole file, many times I'd like to see just video bitrate
[14:40:14 CEST] <Wader8> here ticket https://trac.ffmpeg.org/ticket/5476#ticket
[14:45:15 CEST] <Infiltrator> Is there a way to insert a stream into a file?  Or do I just need to do -c copy and then delete the old file(s)?
[14:54:12 CEST] <Wader8> if you're dealing with MKVs you can try also combination with MKVToolNix
[14:54:26 CEST] <Wader8> there are ways so you can combine stuff
[14:55:02 CEST] <Wader8> you can most probably extract the raw video and audio into a file and use that to merge, and add others
[14:55:15 CEST] <Wader8> you need to research how to use -map option
[15:07:49 CEST] <Infiltrator> Sorry, I wasn't clear.  I want to be able append another stream to an existing file instead of creating a new output file and then deleting the old input file(s).
[15:11:46 CEST] <Wader8> I've been thinking about that but I didn't need it so I don't know exactly
[15:43:08 CEST] <TenLeftFingers> I've left a comment on the ADP4 support that was added here: https://github.com/Nevcairiel/FFmpeg/commit/a66243a2014c4ce689b5514f82effeecf6fc20eb
[15:43:46 CEST] <TenLeftFingers> Can anyone explain the issue I'm having?
[15:44:50 CEST] <furq> TenLeftFingers: the codec name is adpcm_ima_ws
[15:44:57 CEST] <furq> although it's decode-only on this build i've got here
[15:49:55 CEST] <Carlrobertoh> I'm screen casting my video for 10seconds, but the output video is only 4 seconds. What could be the issue =
[15:50:02 CEST] <Carlrobertoh> timestamps ?
[15:50:12 CEST] <Carlrobertoh> i've set it correctly, ithink
[16:03:00 CEST] <TenLeftFingers> furq: thank you. Can I ask what build you're on?
[16:03:18 CEST] <furq> i checked on 2.8.6
[16:03:37 CEST] <furq> it might be worth checking with 3.x
[16:04:12 CEST] <TenLeftFingers> furq: great, thank you so much. I was at a complete loss.
[16:14:37 CEST] <Carlrobertoh> anyone?
[16:48:19 CEST] <Prelude2004c> hey guys... anyone know why none of the closed caption data is not coming through. .i am using scodec copy
[16:48:33 CEST] <Prelude2004c> should ffmpeg not keep that in the stream and just pass it on to segment ?
[16:48:47 CEST] <DHE> depends. closed captions can also be embedded in the video stream itself
[16:48:54 CEST] <Prelude2004c> ${ffmpeg} -i "$stream" -copyts -copytb 0 -codec copy -scodec copy $mapping -avoid_negative_ts 1 -f mpegts -
[16:49:17 CEST] <DHE> use ffprobe on the stream. it should indicate a subtitle stream is present, or it may show "Closed Captions" on the video stream
[16:49:25 CEST] <Prelude2004c> i am taking the video... and stdin to hardware encoding ( nvidia ) .. and then outputting back to ffmpeg to join audio/video
[16:49:40 CEST] <Prelude2004c> ok checking
[16:51:38 CEST] <Prelude2004c> doens't show me anything in the stream but may be embeded in the video
[16:52:04 CEST] <Prelude2004c> actually yes it has it
[16:52:05 CEST] <Prelude2004c> Closed Captions,
[16:52:15 CEST] <DHE> yeah, nvenc won't process those
[16:52:32 CEST] <DHE> libx264 will if you're up-to-date enough to support the "-a53cc 1" option
[16:52:48 CEST] <Prelude2004c> hum.. do you think i should somehow extract it on copy to a file or something and then put it back together on the last ffmpeg session after the encoding
[16:53:04 CEST] <DHE> if you have a tool that will do that, sure that might be your best bet
[16:53:32 CEST] <Prelude2004c> a tool? i am thinking a script .. this is live content
[16:53:37 CEST] <Prelude2004c> has to do it in realtime on decode
[16:53:45 CEST] <Prelude2004c> just not sure how i would do that
[16:54:46 CEST] <Prelude2004c> maybe it also has omething to do witht he fact that we are at 60fps input and i am converting it to 30fps
[16:54:48 CEST] <Prelude2004c> on the output
[16:57:44 CEST] <Prelude2004c> any clues on what i could do about this ?
[17:55:46 CEST] <kthelgason> Hi
[17:56:16 CEST] <kthelgason> I'm building FFmpeg from master, and some actions that worked in 3.0 don't seem to anymore
[17:56:38 CEST] <kthelgason> Looking through the changelog I can't find any changes relevant to what I'm donig. Is it possible that master is in a borked state?
[18:03:39 CEST] <durandal_170> kthelgason: like what?
[18:05:37 CEST] <kthelgason> The command I'm running is ffmpeg -f avfoundation -pix_fmt nv12 -i "0:0" /tmp/out.mp4
[18:05:57 CEST] <kthelgason> this works fine with 3.0, and 3.0.1, but not master
[18:10:00 CEST] <jbreeden> Hello all; I'm new here, forgive me if I'm not following etiquette for asking a question. I'm trying to use FFmpeg to extract a KLVA data stream from MPEG-TS, and convert it to WebVTT. Is there a way to create a subtitle output from a 'data' input?
[18:12:12 CEST] <durandal_170> kthelgason: please report bug on Trac
[18:12:29 CEST] <kthelgason> durandal_170: Ok, I'll do that, thanks.
[18:22:16 CEST] <c_14> jbreeden: not without writing a decoder for the input codec
[18:44:52 CEST] <Prelude2004c> hey guys.. question.. " ERROR: nvEncodeAPI.h not found. " ... i copied the sdk stuff to /usr/local/include
[18:44:57 CEST] <Prelude2004c> and also /usr/include
[18:45:10 CEST] <Prelude2004c> ran ldconfig and everything.. not sure why its not seeing it
[18:45:30 CEST] <c_14> Is it /usr/include/foobar/nvEncodeAPI.h or /usr/include/nvEncodeAPI.h?
[18:45:33 CEST] <c_14> It has to be the second
[18:45:40 CEST] <c_14> Or you have to add /foobar/ to the include path
[18:51:01 CEST] <jbreeden> c_14: thanks, I've started down that path. The only way I could see to do it was to modify libavformat/mpegts.c to treat KLVA as a AVMEDIA_TYPE_SUBTITLE rather than AVMEDIA_TYPE_DATA. Is there a way to start with data, decode, then reencode as subtitle?
[18:52:48 CEST] <c_14> I'm pretty sure you have to add a decoder in libavcodec and register it etc and then it should work
[18:52:52 CEST] <sfan5> Prelude2004c: i build with --extra-cflags=-I/usr/include/nvidia-sdk since nvEncodeAPI.h is in /usr/include/nvidia-sdk
[18:52:55 CEST] <sfan5> that works finme
[18:55:06 CEST] <jbreeden> c_14: yes, but it appears that my decoder's AVCodec struct can either specify a type of AVMEDIA_TYPE_SUBTITLE or AVMEDIA_TYPE_DATA, I don't see the proper way to configure it so that FFmpeg knows that the decoder is starting with data and ending with subtitle
[18:56:31 CEST] <c_14> Just set it to subtitle for both input and output?
[18:58:12 CEST] <jbreeden> c_14: Yeah, that's what I have right now. It feels like a hack, and I was hoping there is a more elegant solution that I failed to see.
[18:58:49 CEST] <jbreeden> I didn't really want to mess with any existing code other than to make my plugin, you know?
[19:04:58 CEST] <Prelude2004c> sfan5 i dont have nvidia-sdk folder
[19:05:03 CEST] <Prelude2004c> i need the tools ?
[19:05:04 CEST] <Prelude2004c>  nvidia-cuda-toolkit
[19:08:11 CEST] <c_14> afaik you only need the header file
[19:12:26 CEST] <sneke> Is ffmpeg able to split a flac based on a cue file or possibly a list of break points??
[19:13:14 CEST] <furq> sneke: not to my knowledge
[19:13:29 CEST] <furq> if you need to do it on the command line then use cuetools and shntool
[19:14:05 CEST] <sneke> That is what I have used so far, but shntool is unable to split this 3.1GB file I have.
[19:14:10 CEST] <furq> fun
[19:14:12 CEST] <atomnuker> sneke: https://github.com/atomnuker/luasplit
[19:14:26 CEST] <atomnuker> a quick script I wrote just for that purpose
[19:14:30 CEST] <furq> if you're on windows then http://www.cuetools.net/wiki/CUETools is excellent
[19:14:53 CEST] <sneke> I'll try that out, atomnuker
[19:15:55 CEST] <furq> but yeah if you don't care about getting a non-conformant cue or about ar/ctdb verification then it's a bit unnecessary
[19:16:34 CEST] <sneke> I'm not entirely sure what you're saying
[19:16:58 CEST] <furq> cuetools (the windows one) will write a valid cuesheet for the split files
[19:17:22 CEST] <furq> and will attempt to verify that it's a good rip
[19:24:28 CEST] <sneke> The script doesn't work for me, atomnuker. It could't get the current dir, and when I fixed that, shnsplit complained about an unknown file format.
[19:24:31 CEST] <sneke> furq: ah
[19:25:28 CEST] <sneke> I'll try to split it manually with ffmpeg
[19:27:00 CEST] <Prelude2004c> furq .. you know this stuff.. nvencodeapi.h missng... how does one get nvidia-sdk ?
[19:27:09 CEST] <furq> i've never used nvenc
[19:27:31 CEST] <Prelude2004c> oh .. ic.. ok  well the sdk i installed but.. for some reason never created the /usr/include-nvidia-sdk
[19:27:33 CEST] <Prelude2004c> its odd
[19:27:34 CEST] <furq> check config.log to see which paths it's searching for
[19:27:53 CEST] <Prelude2004c> its searching for /usr/include
[19:27:59 CEST] <furq> 17:45:30 ( c_14) Is it /usr/include/foobar/nvEncodeAPI.h or /usr/include/nvEncodeAPI.h?
[19:28:00 CEST] <Prelude2004c> and the files are there.. that is why i am blown again
[19:28:02 CEST] <furq> 17:45:33 ( c_14) It has to be the second
[19:28:56 CEST] <furq> sneke: maybe decode the flac to wav and then try again with shntool?
[19:29:25 CEST] <furq> oh nvm it's a known bug with shntool
[19:31:02 CEST] <furq> it looks like the debian package has patches which fix processing large files
[19:31:22 CEST] <furq> if you're not on debian then you could maybe try to apply those to upstream
[19:32:31 CEST] <furq> http://http.debian.net/debian/pool/main/s/shntool/shntool_3.0.10-1.debian.tar.xz
[19:32:55 CEST] <sneke> That's the same version I use, currently
[19:33:07 CEST] <furq> on debian?
[19:33:55 CEST] <sneke> Arch Linux
[19:34:11 CEST] <furq> yeah the patches are from the debian maintainers, i don't think they made it upstream
[19:34:22 CEST] <sneke> ah
[19:36:18 CEST] <sneke> surprisingly small patches
[19:38:42 CEST] <sneke> I believe I know what to do, but in the meantime, is there a time flag for ffmpeg that cuts at a specified time, instead of a duration (specified by -t)?
[19:39:49 CEST] <furq> -to
[19:40:21 CEST] <sneke> lovely, thanks for the help
[20:02:40 CEST] <Pepe> Hello. I saw wish issue for widevine and its there for 2-3 yrs. Is there any progress on it? We would like UPC/Ziggo Horizon for XMBC, but we are waiting for ffmpeg
[20:04:58 CEST] <JEEB> widevine is a DRM thing...
[20:06:32 CEST] <Pepe> yeah I know.. :/
[20:07:38 CEST] <DHE> DRM vendors tend not to be too open with their specs and libraries
[20:09:31 CEST] <durandal_170> no DRM, thanks
[20:09:36 CEST] <Prelude2004c> unreal ...... this is silly
[20:10:06 CEST] <JEEB> if it's DASH-based it's at least standardized how the encryption is done. *but* the stuff that a DRM vendor puts in the area for "how to get the key" is completely up to the vendor and opaque
[20:10:20 CEST] <JEEB> so if you have interest in reverse engineering their current implementation, glhf
[20:10:53 CEST] <Prelude2004c> i have done this 10 times.... always worked by copying the sample files to /usr/include .. ( ERROR: nvEncodeAPI.h not found. ) .. ldconfig doesn't update it and it doesn't see it.. i think i am going to need a new pc from banging on keyboard soon
[20:10:54 CEST] <JEEB> but not sure if it has a place in FFmpeg, since in general we don't handle implementations of DRM schemes
[20:11:23 CEST] <Prelude2004c> the only reason i am doing this.. and mabye there is something else here. one of my ffmpeg version can see the closed captions.. the other one .. i  cannot
[20:11:44 CEST] <Prelude2004c> is that not weird ? one of the ffmpeg versions is newer adn CC is missing from the input... yet i run the other version and CC is present
[20:13:57 CEST] <DHE> I've looked at verimatrix encryption support in HLS...
[20:14:26 CEST] <DHE> I have documents of the ultra-high level of how it works, but that's all you get. the DRM layer is black voodoo
[20:14:28 CEST] <furq> Prelude2004c: ldconfig has nothing to do with finding a header
[20:15:08 CEST] <furq> pastebin the last ~50 lines of config.log i guess
[20:17:40 CEST] <Prelude2004c> i think its something opencl
[20:17:46 CEST] <Prelude2004c> looking at http://developer.download.nvidia.com/compute/cuda/6_5/rel/installers/cuda_6.5.14_linux_64.run
[20:19:07 CEST] <jbreeden> Prelude2004c: what is the full path to your nvEncodeAPI.h currently?
[20:20:36 CEST] <Prelude2004c> it's /usr/include/nvEncodeAPI.h
[20:21:14 CEST] <jbreeden> and you tried it in /usr/local/include and it didn't work either?
[20:21:35 CEST] <Prelude2004c> i may need 6.4.. i installed 7.5 before
[20:21:47 CEST] <Prelude2004c> yup /usr/local/include.. copied to there.. too
[20:21:59 CEST] <Prelude2004c> its odd.. give me a min.. let me see maybe its an opencl issue missing or something
[20:23:17 CEST] <Prelude2004c> it it be its not compiled with " /usr/local/cuda-7.5 " ?
[20:23:23 CEST] <Prelude2004c> i mean compatible
[20:23:41 CEST] <furq> does config.log not tell you why it failed
[20:25:16 CEST] <Prelude2004c> http://pastebin.com/PPKq07NY
[20:27:04 CEST] <jbreeden> http://askubuntu.com/questions/470796/fatal-error-sys-cdefs-h-no-such-file-or-directory
[20:35:35 CEST] <Prelude2004c> jbreeden... it got me closer :)
[20:35:55 CEST] <Prelude2004c> needed to put the ../include in the -I/usr/local/cuda-version/include
[20:36:04 CEST] <Prelude2004c> thank you
[20:36:30 CEST] <jbreeden> :D
[21:01:24 CEST] <Prelude2004c> one last thing.. because all that for nothing :P haha
[21:01:42 CEST] <Prelude2004c> source shows " Stream #0:16[0x1231]: Video: h264 (High) ([27][0][0][0] / 0x001B), yuv420p(tv, bt709), 1280x720 [SAR 1:1 DAR 16:9], Closed Captions, 59.94 fps, 59.94 tbr, 90k tbn, 119.88 tbc "
[21:01:52 CEST] <Prelude2004c> so closed caption exists.
[21:02:46 CEST] <Prelude2004c> ${ffmpeg} -i "$stream" -copyts -copytb 0 -codec copy -scodec copy $mapping -avoid_negative_ts 1 -f mpegts -  .... so the output shows :  Stream #0:0: Video: h264 ([27][0][0][0] / 0x001B), yuv420p, 1280x720 [SAR 1:1 DAR 16:9], q=2-31, 59.94 fps, 59.94 tbr, 90k tbn, 90k tbc
[21:02:54 CEST] <Prelude2004c> where did my closed caption go :(
[21:03:07 CEST] <Prelude2004c> -scodec copy is also there.. so :(
[21:07:47 CEST] <Prelude2004c> any idea ?
[21:11:25 CEST] <DHE> did I already explain that?
[21:11:45 CEST] <DHE> wait, you copied the video stream as well?
[21:23:28 CEST] <kepstin> Prelude2004c: what is $mapping?
[21:42:47 CEST] <Prelude2004c> oht aht
[21:42:49 CEST] <Prelude2004c> sec
[21:43:19 CEST] <Prelude2004c> "-map i:0x1811 -map i:0x1814 -map i:0x1815"
[21:43:25 CEST] <Prelude2004c> its basically the pids i am grabbing
[21:43:32 CEST] <Prelude2004c> the video pid , and the two audio pids
[21:50:18 CEST] <kepstin> so you're mapping the video and two audio, but not the subtitles
[21:50:22 CEST] <kepstin> so you get no subtitles
[21:50:44 CEST] <kepstin> if you use -map, you get only the stuff that you map
[21:51:34 CEST] <courrier> Hey guys, for some reason my video has a single black frame hidden between two normal frames, making a fast black flash when the video plays, how can I identify and remove that frame?
[21:52:16 CEST] <courrier> mpv that allows frame-by-frame playback backwards allowed me to see the actual frame but it doesn't tell the frame id
[21:57:02 CEST] <Prelude2004c> that makes sense but the input i see no PID for sub-tabiles
[21:57:07 CEST] <Prelude2004c> sub-titles
[21:57:11 CEST] <Prelude2004c> where would i find that pid
[21:58:24 CEST] <Prelude2004c> i am pretty sure the closed caption is in the video pid
[22:13:31 CEST] <Prelude2004c> any hints ?
[23:10:25 CEST] <patdavid> best option to transcode 4k mp4/mov from a drone to playback smoothly if space isn't an issue?
[23:10:40 CEST] <patdavid> lagarith? ffv1? huffyuv? maybe prores?
[23:11:36 CEST] <furq> i doubt those would play back more smoothly than h.264
[23:11:36 CEST] <kepstin> limit is probably io bandwidth with most lossy codecs
[23:12:02 CEST] <kepstin> with most lossless or near-lossless codecs*
[23:12:37 CEST] <patdavid> is there a better option for smoother playback that is less cpu intensive?
[23:14:04 CEST] <kepstin> for 4k video, if you have a hardware h264 decoder you can use, that's probably the best option.
[23:14:28 CEST] <patdavid> ok
[23:14:46 CEST] <DrSlony> patdavid try mpv
[23:14:56 CEST] <DrSlony> https://mpv.io/
[23:14:57 CEST] <patdavid> w/o a hardware decoder, what about ffv1?  it seemed pretty quick last time i checked
[23:15:21 CEST] <DrSlony> lossless bandwidth kills my hardware
[23:15:25 CEST] <TD-Linux> what device are you playing back on
[23:15:37 CEST] <flux> are there players that buffer decoded frames? that could smooth out between the impact of decoding p- and i-frames
[23:15:47 CEST] <patdavid> one of my wifes students desktop machines in a science lab
[23:15:53 CEST] <furq> patdavid: at 4k you're going to be running up against the limits of disk access speeds
[23:15:57 CEST] <flux> or do all of them do it :)
[23:15:59 CEST] <patdavid> furq, that's what i figured
[23:16:09 CEST] <furq> uncompressed at 24fps that's ~300mbps
[23:16:12 CEST] <flux> not with ssd :)
[23:16:13 CEST] <TD-Linux> patdavid, does it even have a 4k monitor? can you just drop the res to 1080p?
[23:16:30 CEST] <patdavid> no, they need the spatial resolution to identify wildlife in their frame
[23:16:35 CEST] <patdavid> :(
[23:16:43 CEST] <patdavid> i considered proxy video as an option
[23:17:02 CEST] <patdavid> at 1080p, then if they get a decent confidence that something might be there, seek in the original 4k footage for it
[23:17:25 CEST] <DrSlony> is the 4k interlaces?
[23:17:26 CEST] <TD-Linux> that might be reasonable. depending on your vo it might have trouble just displaying the 4k frame
[23:17:27 CEST] <patdavid> it's still on the table, just more complicated for natural science undergrads
[23:17:29 CEST] <DrSlony> *interlaced
[23:17:35 CEST] <patdavid> DrSlony, no
[23:17:38 CEST] <patdavid> progressive
[23:17:40 CEST] <patdavid> 25fps
[23:18:00 CEST] <flux> patdavid, so what is the bottleneck here, why isn't the original video smooth?
[23:18:05 CEST] <DrSlony> patdavid what if you copy the video, or a part of the video, to a RAM drive, does it still stutter on playback?
[23:18:21 CEST] <patdavid> flux, i'm guessing the h264 decoding on the cpu (+ ram is likely not much)
[23:18:31 CEST] <patdavid> DrSlony, haven't checked yet
[23:18:34 CEST] <furq> it might be worth trying a different player
[23:18:37 CEST] <patdavid> DrSlony, don't have any files handy at the moment
[23:18:40 CEST] <furq> but yeah if the bitrate is very high it could be that too
[23:18:42 CEST] <flux> patdavid, so it's quite an old pc then if it doesn't have a GPU that accelerates it?
[23:18:47 CEST] <DrSlony> if linux, try copying to /tmp or /dev/shm/
[23:18:54 CEST] <patdavid> flux, i am thinking that's the case
[23:19:17 CEST] <furq> what player is the machine using
[23:19:28 CEST] <patdavid> vlc i think
[23:19:32 CEST] <patdavid> not sure which build - could be old
[23:19:37 CEST] <TD-Linux> there are codecs that are less complex than H.264, but you might be hitting other limits (like memory bandwidth) on your potatoboxen that make a faster codec less useful
[23:19:44 CEST] <furq> maybe give mpv a try then, i know ffmpeg's h264 decoder is pretty well-optimised
[23:19:52 CEST] <patdavid> ok, worth a shot
[23:19:57 CEST] <patdavid> thank you all for the help
[23:19:59 CEST] <TD-Linux> vlc uses the same decoder so maybe
[23:20:13 CEST] <patdavid> might need to upgrade to a bigger potato
[23:20:18 CEST] <furq> yeah i suspected it might
[23:20:19 CEST] <flux> mpv is also a fast seeker and has builtin-support for looping short parts of video interactively
[23:20:29 CEST] <flux> it seems to me it would be a nice tool for finding parts from a video
[23:20:39 CEST] <furq> i guess mpv is more likely to have the newest lavc though
[23:20:51 CEST] <furq> especially if it's not *nix
[23:20:51 CEST] <patdavid> wonderinf it wouldn't make sense to dump minutes of vide to frames and loading up in blender for zoom/pan/etc
[23:21:05 CEST] <flux> I think decoding 4k video purely CPU-based is not going to happen with h264, though
[23:21:13 CEST] <TD-Linux> patdavid, you could try a high quality but lossy intra format, like motion jpeg or relatives
[23:21:31 CEST] <patdavid> TD-Linux, thx, i'll try it with some sample footage tomorrow
[23:21:44 CEST] <patdavid> flux, yeah, brought my macbook air to its knees...
[23:21:52 CEST] <patdavid> (c2013 machine)
[23:24:40 CEST] <flux> I have a couple? years old AMD 8350, it takes 260% cores to decode 4k video without hardware acceleration
[23:25:27 CEST] <furq> the profile will make a difference as well
[23:25:44 CEST] <furq> although if it's come off a gopro i can't imagine it's using anything too intensive
[00:00:00 CEST] --- Wed Apr 27 2016