[Ffmpeg-devel-irc] ffmpeg.log.20170211

[burek]

[00:48:36 CET] <tmatth> is it possible to control the rtsp buffer size with the ffmpeg CLI?
[00:50:59 CET] <atomnuker> tmatth: -help muxer=rtsp says there's -buffer_size
[00:51:47 CET] <tmatth> atomnuker: yeah, saw that but it doesn't seem to be getting picked up (i'm looking for client side buffering btw, from reading rtsp urls)
[00:52:07 CET] <tmatth> or maybe it is and it's just not mentioned in the verbose logs :/
[00:57:51 CET] <DHE> ran into a weird bug where the udp output parameter "-bitrate 10M" would result in an out-of-memory condition with certain other parameters...
[00:59:33 CET] <DHE> okay, yeah. if you mix -bitrate for output and a large analyzeduration  (default may be fine) you can quickly end up in an out-of-memory scenario. shrinking analyzeduration helps a lot
[01:04:33 CET] <IntruderSRB> hey guys, I know it's not a ffmpeg issue but this is a community I'm hoping to find someone with experience in HLS/FPS streaming with fMP4. To avoid spamming on the channel can anyone with such experience contact me in private (or here if it's allowed to discuss offtopic) for a quick consulting? Thanks! P.S. I've entered media streaming world few months ago so I plan to stick here and learn/share whenever I can :) Hope to meet ppl with
[01:04:33 CET] <IntruderSRB>  similar interest :)
[01:05:41 CET] <DHE> by FPS you mean video games?
[01:07:17 CET] <IntruderSRB> no, but it would be far more interesting than FairPlay Streaming :)
[01:07:55 CET] <DHE> DRM?
[01:08:19 CET] <IntruderSRB> yep
[01:08:20 CET] <DHE> oh I'll just google it myself instead
[01:08:39 CET] <IntruderSRB> Apple DRM system for HLS streams
[01:09:16 CET] <DHE> interesting. HLS spec (by apple) does have specifications for how to encrypt streams (which ffmpeg supports in full AES mode only) so I imagine they would have some compatibility going there
[01:09:44 CET] <JEEB> nah, fairplay is a "proper" DRM (although I haven't seen it mentioned in contracts yet)
[01:09:53 CET] <JEEB> so it does nonstandard stuff there
[01:10:03 CET] <DHE> so much for that idea
[01:10:05 CET] <IntruderSRB> FPS uses SAMPLE-AES encryption
[01:10:21 CET] <DHE> I have seen a vendor (won't name names) that is ffmpeg-compatible to a certain degree. and I'll stop here.
[01:10:22 CET] <IntruderSRB> but the whole idea of using fMP4 instead of TS is new (mid 2016)
[01:11:13 CET] <DHE> that's interesting mainly because it would help browsers that need HLS.js to convert TS to MP4 anyway, and because mpeg-ts is actually pretty bad for overhead
[01:11:18 CET] <JEEB> yeah, there's no fragmented isobmff in HLS support in libavformat yet
[01:11:39 CET] <IntruderSRB> @dhe exactly
[01:11:50 CET] <IntruderSRB> and additionally you would be able to use same fMP4 in DASH and HLS
[01:12:42 CET] <JEEB> yeah, that'd be useful
[01:13:43 CET] <IntruderSRB> @jeeb hm that's strange ... 'cause I tested today ffmpeg with HLS/fMP4 and it was ok? as long as I keep encryption out of the way... there's no sample-aes support yet
[01:14:09 CET] <JEEB> wait, was that added?
[01:15:00 CET] <JEEB> I don't see it @ http://git.videolan.org/?p=ffmpeg.git&a=search&h=HEAD&st=commit&s=hlsenc
[01:15:46 CET] <IntruderSRB> let me explain what I did and it worked - not sure what's going on underneath
[01:15:58 CET] <JEEB> if you generate the playlist yourself of course that will work :P
[01:16:07 CET] <JEEB> since avformat can mux fragmented ISOBMFF just fine
[01:17:20 CET] <IntruderSRB> ah ... well yea, I provided my streamer m3u8 and ffmpeg was ok with probing/downloading it into a file
[01:17:52 CET] <JEEB> well yes, ffmpeg's HLS *demuxer* doesn't expect the stuff to be MPEG-TS
[01:18:03 CET] <JEEB> it probes and it finds fragmented ISOBMFF => shit works
[01:18:08 CET] <DHE> call me curious. how would I produce such files with ffmpeg? (assuming I wrote my own .m3u8 writer)
[01:18:17 CET] <JEEB> see the HLS spec
[01:18:32 CET] <JEEB> https://tools.ietf.org/html/draft-pantos-http-live-streaming-20#section-3.3
[01:18:41 CET] <JEEB> that's the part about fragmented ISOBMFF
[01:19:09 CET] <DHE> right, I've seen that. I guess the question is how to make ffmpeg write such files...
[01:21:10 CET] <JEEB> see what movenc muxer flags dashenc uses
[01:21:25 CET] <JEEB> this lists all of them https://www.ffmpeg.org/ffmpeg-all.html#mov_002c-mp4_002c-ismv
[01:21:27 CET] <DHE> ... that works
[01:31:55 CET] <IntruderSRB> btw. has anyone of you worked with 'cbcs' mpeg-cenc encryption scheme?
[01:33:15 CET] <JEEB> someone implemented de- and encryption with CENC for ISOBMFF
[01:33:19 CET] <JEEB> but not sure which mode
[01:33:45 CET] <JEEB> (in other words, "you are welcome to read the code")
[01:37:34 CET] <IntruderSRB> saw in movenc there's support for 'cenc'
[01:38:06 CET] <JEEB> yes, that's the encryption part
[01:40:06 CET] <IntruderSRB> 'cbcs' should be pretty similar - except constantIV and pattern applied to subsample ... but there's a catch I can't figure out yet :)
[02:24:39 CET] <kota> If I'm trying to stream multiple files, concatenated via concat filter, in real time, what's the best way to do it? -re seems to only work for the first input file
[03:02:54 CET] <JC_Yang> I have some question about hardware decoding with ffmpeg, first, on linux, is VA-API the de facto standard API for hardware decoding? Are most linux distros and android using it as the standard API for hardware decoding? Do most distros/android require the hardware vendor provide a VA-API compatible driver to enable hardware decoding/acceleration?
[03:59:18 CET] <thebombzen> JC_Yang: no
[03:59:32 CET] <thebombzen> VDPAU exists for nvidia cards
[03:59:38 CET] <thebombzen> it depends on the hardware decoder
[03:59:44 CET] <thebombzen> most manufacturers have their own thing
[03:59:57 CET] <thebombzen> nVidia has VDPAU, Intel has VAAPI for example
[04:01:32 CET] <JC_Yang> but don't they have to comply to a standard user space API? for the sake of user library such as ffmpeg, for so many arm soc out there, which API do they typically adhere to?
[04:01:52 CET] <JC_Yang> is VA-API dominate in the world except nvidia?
[04:02:37 CET] <TD-Linux> not at all. android doesn't have va-api at all
[04:02:46 CET] <thebombzen> why would they "have to"
[04:05:21 CET] <JC_Yang> then how does a general pu user space library such as ffmpeg cooperate with so many different capable-hardware decoder? write hardware specific code for each piece of new chip? back to the dark age of the past?
[04:05:32 CET] <JC_Yang> general purpose... typo
[04:08:40 CET] <JC_Yang> I never consider a set of API as bad thing, though the linux kernel give up a consistent driver API, but it's exception, not the norm, isn't it?
[04:17:13 CET] <thebombzen> JC_Yang: how does a something like ffmpeg cooperate?
[04:17:23 CET] <thebombzen> well in general ffmpeg will software decode unless the user tells it to
[04:17:32 CET] <thebombzen> and the user has to tell it which hardware decoder to use
[04:25:32 CET] <thebombzen> also
[04:25:40 CET] <thebombzen> what's the point of lossy intra-only codecs
[04:26:27 CET] <thebombzen> like I don't understand why one would use prores over Motion JPEG 2000
[04:49:47 CET] <aokfire> Hi. I downloaded a stream using yt-dl, and after completion, ffmpeg says: "Fixing malformated aac bitstream". I notice it's related to ADTS. When it fixes it, ADTS is gone from the file.
[04:50:11 CET] <aokfire> I was just wondering if this is critical or necessary, since after muxing with MKVMerge ADTS is gone as well and the audio seems fine overall.
[04:50:13 CET] <aokfire> Thank you
[05:32:52 CET] <basiclaser> hi all
[05:33:33 CET] <basiclaser> im trying to convert all mp4 containers to mov, but my loop ( for f in *mp4; do ffmpeg -i "$f" -acodec copy -vcodec copy -f mov "$f".mov ; done ) stop after the first file
[05:34:50 CET] <basiclaser> ah i think i got it :D
[06:01:11 CET] <sware> anyone know what would cause h264 decoding to be jumping back and forth? It's not stuttering it's more like a motion or time problem?
[06:20:08 CET] <dongs> sware: seen that on shitty usa tv captures
[06:21:11 CET] <sware> I know the hardware is capable
[06:22:52 CET] <saaeb> hi there  .
[06:24:17 CET] <sware> I know it's a bug in libavformat somewhere involving time base
[06:28:19 CET] <saaeb> in older version of ffmpeg i just wrote ffmpeg -i input output.mp4 and input would be converted with losing quality
[06:28:29 CET] <saaeb> in ffmpeg3 it will lose quality
[06:42:45 CET] <saaeb> 👀
[06:46:12 CET] <dongs> nice
[06:46:15 CET] <dongs> keil just crashed hard
[06:52:23 CET] <dysfigured> any of you worked with gopro footage? i have a friend who has some sort of .gopro video file that they want stills from
[06:55:55 CET] <llamapixel> https://gopro.com/how-to-prepare-source-video-files-for-editing-windows-pc/
[07:00:48 CET] <dysfigured> i'm on linux, and i don't think they have access to the software
[07:01:06 CET] <furq> what does ffprobe say about the file
[07:01:58 CET] <dysfigured> no idea, i don't have it yet, i was wondering if anyone here had experience with it before i ask him to upload it to me and i find out i need their special snowflake software.. google isn't helping me much
[07:15:20 CET] <dysfigured> i'd be really surprised if gopro invented some special new container format. surely it's just an mp4 or something with a renamed file extension
[07:16:37 CET] <furq> well if it wasn't there'd probably be some information about it on the internet
[07:16:53 CET] <furq> if you can get the guy who has it to run ffprobe on it then that should tell you
[08:14:28 CET] <sware> I've read a few places that ffmpeg syncs to the audio rather than the video. Is this true and if so where exactly is it done?
[08:14:48 CET] <llamapixel> https://gopro.com/support/articles/what-video-format-do-hero3-cameras-record-in
[08:37:35 CET] <TommyC> Hello, I'm not sure if this is the correct channel but I'm just wondering how to build ffplay (i.e., what configure flags I would need to pass)?
[08:38:11 CET] <TommyC> I thought SDL was the only dependency for ffplay (excluding support for audio/video formats) which I do have installed and the configure script sees as well.
[08:38:36 CET] <furq> https://github.com/FFmpeg/FFmpeg/blob/master/configure#L3194
[08:38:39 CET] <furq> that should be all you need
[08:38:46 CET] <TommyC> furq: Thank you! I'll take a look.
[08:40:05 CET] <TommyC> Indeed that is all I needed. Thank you again.
[13:43:26 CET] <NyB> Hi folks! Not quite OT, but is there somewhere a back-of-the-envelope table with equivalent bitrates (for roughly the same quality) for some "common" HEVC and AVC options? Of course the video material & source matters, but maybe there's a general rule of thumb?
[15:21:31 CET] <MrSassyPants> Question, what's up with WebVTT subtitles in .mkv / .webm files? Are they supposed to work?
[15:25:07 CET] <c_14> MrSassyPants: simple ones should work
[15:25:22 CET] <MrSassyPants> Isn't webVTT quite simple to begin with?
[15:28:35 CET] <c_14> WebVTT can have stuff like external CSS etc
[15:29:15 CET] <MrSassyPants> well the thing is, I can extract it as SRT and use the subtitles like that, but muxed into the webm, the subtitles don't show
[15:29:41 CET] <MrSassyPants> So I'm thinking something's wrong with the muxer or the demuxer
[15:29:59 CET] <c_14> what player?
[15:30:10 CET] <MrSassyPants> tested mplayer & ffplay
[15:30:12 CET] <c_14> And try outputting raw webvtt and checking in a text editor?
[15:30:19 CET] <MrSassyPants> yes, did
[15:30:28 CET] <MrSassyPants> the .srt file that I extracted from the webm works
[15:30:43 CET] <MrSassyPants> both as a text file that I looked at and as a subtitle file for mplayer
[15:31:11 CET] <c_14> I meant -f webvtt
[15:36:49 CET] <c_14> So I just tested with some of my files, ffplay just won't play webvtt for whatever reason
[15:36:52 CET] <c_14> works with mpv though
[16:30:15 CET] <BennyHyphen> Hello #ffmpeg Problem solving time: I have a 30 hour video without sound and 4 hours worth of 5 minute music tracks. I would like to sprinkle the tracks over the video so the entire video has some audio. Is this possible without re-encoding the audio? Assume the codec for all the audio files is the same.
[16:32:49 CET] <DHE> might be possible with the concat format
[17:18:24 CET] <MrSassyPants> c_14, with -f webvtt, I get a working text file out of the (apparently non-subtitled) webm too
[17:24:11 CET] <c_14> MrSassyPants: like I said, ffplay won't play them for me either. Try a different player like mpv (or maybe vlc or something)
[17:24:55 CET] <MrSassyPants> c_14, lel, vlc can't do the subtitle nor the audio
[17:25:41 CET] <MrSassyPants> yeah, mpv can do it
[17:26:07 CET] <c_14> Probably a bug in ffplay then, not in the ffmpeg muxer
[17:29:19 CET] <MrSassyPants> well, mplayer too
[17:30:49 CET] <MrSassyPants> But: Another mkv (I have no idea what was used to mux it) with embedded WebVTT  works in mplayer, but not ffplay
[17:31:23 CET] <MrSassyPants> but with the ffmpeg-encoded webm, neither can play it!
[17:32:18 CET] <c_14> Then ffmpeg probably uses some superset of features that mplayer doesn't support
[17:34:25 CET] <MrSassyPants> if that was all, then ffplay should be capable of playing embedded webVTT
[17:35:03 CET] <MrSassyPants> I think something's screwed up with the webvtt muxing in ffmpeg, and it affects ffplay too
[17:35:17 CET] <c_14> I'm not sure why ffplay won't play it
[17:36:01 CET] <c_14> Have you tried extracting the webvtt subtitles from the working copy with -c:s copy -f webvtt out.vtt and then running ffmpeg -i out.vtt out2.vtt on that and diffing them to see what's different from the working and the non-working copy?
[17:36:16 CET] <c_14> (The one file you mentioned that worked in mplayer but not when converted)
[17:37:07 CET] <MrSassyPants> c_14, sec
[17:38:13 CET] <faLUCE> what does it happen if I execute sws_scale() from an image 320x240 to an image 640x480? Does it throw a warning at runtime?
[17:38:29 CET] <faLUCE> is it safe?
[17:38:56 CET] <MrSassyPants> c_14, "Could not write header for output file #0 (incorrect codec parameters ?): Invalid argument"
[17:39:22 CET] <MrSassyPants> c_14, when trying to extract in the first place with -f webvtt
[17:39:35 CET] <MrSassyPants> c_14, it's 1 line
[17:39:41 CET] <MrSassyPants> pastebin url would also be a line
[17:39:42 CET] <c_14> There should be more output
[17:39:47 CET] <MrSassyPants> yes there would be
[17:40:03 CET] <MrSassyPants> you want the whole, sec
[17:40:26 CET] <MrSassyPants> Ah, I think I got the issue
[17:40:45 CET] <MrSassyPants> it's not webvtt, and -c:s copy doesn't work naturally
[17:41:17 CET] <MrSassyPants> ok that might explain why mplayer plays that one :|
[17:41:36 CET] <MrSassyPants> it works as it should with -map 0:s:0
[17:41:50 CET] <MrSassyPants> but that's a moot point then
[17:42:33 CET] <MrSassyPants> sorry, I forgot the original wasn't actually using webvtt as its subtitle embed
[17:44:03 CET] <MrSassyPants> BUT
[17:45:31 CET] <MrSassyPants> the mapped, converted webvtt from the original file differ from the mapped, extracted vtt from the re-encoded file
[17:46:02 CET] <MrSassyPants> but it's mostly microsecond timing it seems
[17:46:29 CET] <MrSassyPants> otherwise both are equally legible
[17:47:15 CET] <t3v4> Hi guys :)
[17:47:22 CET] <t3v4> Is anybody here? :)
[17:47:34 CET] <t3v4> i have questions related to resampling
[17:47:38 CET] <t3v4> if anybody can help me :)
[17:48:17 CET] <t3v4> I am wondering if I can use resample to give me normalized output?
[17:49:19 CET] <t3v4> I am writing software for audio analysis and I am getting FLT samples which are greate for me, but interval is pretty high so I want to normalize sample values between -1.f and 1.
[17:49:30 CET] <t3v4> or any other smaller interval
[17:49:48 CET] <t3v4> from cpp code ofc
[18:32:53 CET] <flux> I have this 192.168.1.238 sending udp to 239.255.42.42:5004 on eth1 and I would like to have ffmpeg somehow access it. should ffmpeg -i udp://@192.168.1.1:5004?sources=239.255.42.42 foo.mkv do at least something? it's just waiting.
[18:33:18 CET] <flux> it also doesn't add itself to the multicast membership list, is there some way to do that?
[18:36:19 CET] <flux> ah, ffmpeg -i udp://@239.255.42.42:5004?localaddr=192.168.1.1 foo.mkv did something. thanks ;-)
[18:44:14 CET] <JEEB> yea
[18:44:19 CET] <JEEB> I was poking at that the other day
[18:54:19 CET] <flux> it seems though that the source data is highly likely rtp, but if I just replace 'udp' with 'rtp', it similarly ends up waiting..
[18:54:38 CET] <flux> there is no rtcp, though, so I wonder if that's the problem.
[18:55:39 CET] <flux> I'm trying to receive data from a HDMI over IP extender
[19:03:27 CET] <flux> oh, udp worked just fine, so no rtp here. -f mpegts fixed it (paired with -c:v copy)
[19:14:38 CET] <Chloe> how do I map the same input stream multiple times with different options?
[19:16:56 CET] <furq> -map 0:v -map 0:v -c:v:0 libx264 -c:v:1 ffv1
[19:22:58 CET] <Chloe> I mean different decoding options, sorry
[19:24:22 CET] <furq> i don't think there's any way to do that
[20:21:04 CET] <aerodavo_> i have a CLI question about capturing usb hdmi device using avfoundation in ffmpeg on mac
[20:21:44 CET] <aerodavo_> http://pastebin.com/5Nt4faJ6
[20:22:25 CET] <aerodavo_> my main concern is that the audio and video seem to be a little bit out of sync
[20:22:52 CET] <aerodavo_> is there a way to adjust the audio stream somehow so that it is in sync?
[20:25:50 CET] <aerodavo_> i also have a question about "dup=" and "drop=". they start out very low, but for longer recordings, the numbers for each have gotten up to around 200
[20:37:20 CET] <aerodavo_> should i be addressing this to a certain user somehow? any help is much appreciated, this is my first time on here :)
[20:38:38 CET] <JEEB> you should just wait and see if someone decides to help you. staying around is the biggest factor.
[20:39:05 CET] <furq> no, you should ask a question, wait 30 seconds, say "hello?" and then leave one second before someone replies
[20:39:08 CET] <furq> that's what i always do
[20:39:15 CET] <JEEB> also you can offset input timestamps for streams with -itsoffset
[20:39:34 CET] <JEEB> (do note that ffmpeg.c can do funky things with that)
[20:40:30 CET] <furq> there's also the adelay filter if you're reencoding
[20:40:45 CET] <furq> and you can do either of those in a second pass after the capture
[20:43:10 CET] <Chloe> furq: can I do this in a better way? https://www.irccloud.com/pastebin/V9EWmZsW/somethinglikethis.sh
[20:43:32 CET] <Chloe> preferably only reading the input file once
[20:45:48 CET] <furq> doesn't -txt_page take a list
[20:45:54 CET] <furq> although idk if it splits them
[20:46:41 CET] <Chloe> It doesnt split them
[20:46:52 CET] <Chloe> hence having to take it in multiple times
[20:47:51 CET] <aerodavo_> if i'm getting more and more dropped and duplicate frames, does that mean the audio sync is drifting, or can audio sync be maintained with dropped or duplicate frames?
[20:48:09 CET] <aerodavo_> thanks for the help!
[20:48:47 CET] <aerodavo_> it doesn't seem like it's getting worse, more like it's starting off a little out of sync
[20:48:52 CET] <aerodavo_> i'll try those two options
[20:52:34 CET] <aerodavo_> @JEEB, if i'm getting more and more dropped and duplicate frames, does that mean the audio sync is drifting, or can audio sync be maintained with dropped or duplicate frames?
[21:05:27 CET] <aerodavo_> when capturing a live input, if i'm getting more and more dropped and duplicate frames, must that mean that the audio and video are becoming more and more out of sync, or can audio sync be maintained with dropped and/or duplicate frames?
[21:05:45 CET] <aerodavo_> http://pastebin.com/5Nt4faJ6
[21:47:07 CET] <BennyHyphen> This command crashed my computer real hard. Not even the magic sysrq could resolve it: ffmpeg -i input.mkv -t 01:00:00.0 -acodec copy -vcodec copy first_hour.mkv
[21:47:22 CET] <BennyHyphen> Is it a bad idea to run a command like this?
[21:47:44 CET] <BennyHyphen> I actually don't understand why it is crashing.
[21:50:20 CET] <durandal_1707> what ffmpeg version?
[21:51:48 CET] <BennyHyphen> ffmpeg version 2.8.10-0ubuntu0.16.04.1
[21:55:43 CET] <durandal_1707> doesnt happen here with master
[21:55:51 CET] <durandal_1707> it errors out
[21:55:55 CET] <durandal_1707> instead
[21:57:10 CET] <BennyHyphen> What kind of error does it give?
[21:57:19 CET] <BennyHyphen> Can I report this as a bug?
[21:58:04 CET] <durandal_1707> invalid duration specification
[22:01:15 CET] <durandal_1707> you can report anything as bug
[23:34:41 CET] <IntruderSRB> anyone here had experience with 'cbcs' mpeg-cenc encryption? trying to implement it atm so I'd like to exchange a few issues :)
[00:00:00 CET] --- Sun Feb 12 2017