[Ffmpeg-devel-irc] ffmpeg.log.20180410

[burek]

[00:11:56 CEST] <pyBlob> kepstin: the input is a numpy array from a python program, after writing all data, the input stream of ffmpeg is closed via the close() method
[00:13:14 CEST] <pyBlob> kepstin: also the same behaviour occurs when first writing the data to  a file and then telling ffplay to load the file instead of stdin. also same behaviour happens when playing mp3/wav files
[00:14:28 CEST] <kepstin> well, "ffplay -autoexit file.mp3" works perfectly fine for me...
[00:15:22 CEST] <memo1> Hi, i need to save video from my cameras to the local disk and stream video on demand, how i do that with ffmpeg
[00:15:50 CEST] <kepstin> memo1: with difficulty, ffmpeg is not itself a streaming application.
[00:16:01 CEST] <kepstin> it can certainly do the "saving to local disk" part tho.
[00:16:29 CEST] <pyBlob> kepstin: https://pastebin.com/qGJMxJ3H might the "invalid return value 0 for stream protocol" have something to do with that issue?
[00:18:11 CEST] <pyBlob> it appears as soon as ffplay has played the whole file
[00:18:17 CEST] <kepstin> pyBlob: hmm, that's kind of strange. I haven't seen that before.
[00:18:56 CEST] <kepstin> are you running this on a headless (no display) system? You might try also adding the "-nodisp" option so it doesn't attempt to open the audio visualization.
[00:19:42 CEST] <memo1> kepstin: i see that ffserver is no longer supported, but it is useful nowdays?
[00:20:24 CEST] <kepstin> memo1: who knows. It's been removed from ffmpeg completely in recent versions.
[00:20:27 CEST] <memo1> kepstin: i want to use ffmpeg to build a nvr.  So the streaming is for my remote clients that wants to watch rel time videos.
[00:20:42 CEST] <memo1> kepstin: is possible?
[00:21:37 CEST] <pyBlob> kepstin: I'm running this directly from python, as well as from a shell on my desktop (so there is a display ;) ). Adding -nodisp gives the same behaviour
[00:22:05 CEST] <kepstin> memo1: at its simplest, you could just have ffmpeg record to an HLS stream on disk, serve that up with some web server, and have an html/js player frontend.
[00:23:15 CEST] <kepstin> pyBlob: you have me stumped. This should be working and I don't know why it's not for you.
[00:23:27 CEST] <memo1> kepstin: thank you
[00:24:05 CEST] <c_14> pyBlob: ffmpeg -i blah.mp3 -f alsa default ?
[00:26:19 CEST] <kepstin> pyBlob: or you could use e.g. mpv or something too. Or even just aplay for raw audio.
[00:26:55 CEST] <pyBlob> c_14: also prints the return value warning but then exists (though there is no nice visualization ... obviously)
[00:27:14 CEST] <c_14> sounds like something is borked either with the file or ffmpeg
[00:27:17 CEST] <c_14> or alsa I guess
[00:27:36 CEST] <c_14> Does this happen with all files?
[00:27:42 CEST] <c_14> Does speaker-test -c2 work?
[00:32:13 CEST] <memo1> kepstin: if i want to watch and save at the same time?, locally on my machine?
[00:34:55 CEST] <pyBlob> c_14: it seems to happen with all audio-only files. when there is a video track "-autoexit" works
[00:37:14 CEST] <furq> i wouldn't rely on ffplay for anything really
[00:37:27 CEST] <furq> you can pipe raw pcm to other stuff, like mpv
[00:37:35 CEST] <quint> I'm sure this is asked When listing the codecs on ffmpeg, what exactly do the "DEVASILS" indicators? And specifically, when one is strictly V L.. What does that mean exactly?
[00:37:47 CEST] <quint> Asked a lot **
[00:38:10 CEST] <c_14> quint: https://pb.c-14.de/t/kng.9kD0IL
[00:38:21 CEST] <furq> quint: http://vpaste.net/tcCym
[00:38:22 CEST] <c_14> it's listed at the top, right beneath the banner
[00:38:39 CEST] <furq> yeah what he said
[00:39:40 CEST] <quint> Yeah, I saw that. But what can I do with it exactly? I downloaded an AV1 sample encode. Is it possible to output it to yuv4mpegpipe?
[00:41:09 CEST] <c_14> V L.. means you can't do anything with it
[00:41:09 CEST] <quint> "..V.L. av1                  Alliance for Open Media AV1"
[00:41:12 CEST] <quint> Ah.
[00:41:29 CEST] <furq> you can build git head against libaom
[00:41:29 CEST] <c_14> av1 is currently only supported via libaom
[00:52:04 CEST] <ploprof> Hi ! I am getting inconsistent file sizes when I run this command https://pastebin.com/2J7Aj8MV
[00:52:14 CEST] <ploprof> every time I run it the output file has a different size
[00:52:26 CEST] <ploprof> does anyone know why this is ?
[01:14:30 CEST] <memo1> hi, ffmpeg support h265?
[01:21:25 CEST] <DHE> memo1: decoding, yes. encoding, yes with libx265
[01:33:27 CEST] <pyBlob> kepstin: c_14: the exit thing seems to be a problem with the arch linux build of ffmpeg, I just pulled and built the latest master from github and it works fine. thanks for the troubleshooting.
[01:41:07 CEST] <DHE> quint: if you want a static build of ffmpeg with av1 support on linux, I can make one...
[04:20:55 CEST] <testerbeta> Hi
[04:21:00 CEST] <testerbeta> how can i read what codecs
[04:21:03 CEST] <testerbeta> has a video?
[04:21:07 CEST] <testerbeta> with ffmpeg?
[04:21:20 CEST] <testerbeta> mediainfo only says me that has ne of the aacs avaible at ffmpeg
[04:22:42 CEST] <fella> testerbeta: how about 'ffprobe'?
[04:24:05 CEST] <testerbeta> can u explain me (i've serioustroubles with cinelerra and mp4)
[04:24:35 CEST] <testerbeta> and i  really want mp4 (i've an amd fxcpu,  if i use ogg, my renders are a real torture,  with x64  are fast)
[04:24:42 CEST] <testerbeta> but mp4 its crazy
[04:25:32 CEST] <fella> i'm not sure if i understand your problem at all
[04:35:46 CEST] <testerbeta> I've read it at ffprobe
[04:35:53 CEST] <testerbeta> but idk how made the SAME setup¿
[04:37:10 CEST] <testerbeta> https://p.teknik.io/BwKjL
[04:43:40 CEST] <ariyasu> if your having trouble decoding the video it's more than likely because it is 4:4:4
[04:43:54 CEST] <ariyasu> its .h264 video btw
[04:45:22 CEST] <ariyasu> ffmpeg -i awkif.mp4 -vcodec libx264 -pix_fmt yuv420p -preset slow -profile:v high -level 4.1 -acodec copy newfile.mp4
[04:45:42 CEST] <ariyasu> do that to encode it to x264 at 4:2:0
[04:46:05 CEST] <ariyasu> which you will be able to playback without any issues on pretty much anything
[05:43:34 CEST] <bumblehead> hello I hope it is not a bother to ask... I came here to know dark shikari's opinion of the AV1 video codec
[05:44:22 CEST] <bumblehead> and in general, #ffmpeg's opinion
[12:17:22 CEST] <jaggz> I'm trying to convert a bunch of photos into a video but my video's not showing anything at all.
[12:19:39 CEST] <jaggz> I'm using multiple -i's though.. hr
[12:22:39 CEST] <jaggz> ffmpeg -framerate 1 -i "$tmpf" -vf scale=1280:ih*720/iw "$ovid"
[12:23:12 CEST] <jaggz> tmpf is a file with lines like:   file 'images/foo.jpg'
[12:26:32 CEST] <jaggz> Error while decoding stream #0:0: Invalid data found when processing input
[12:26:52 CEST] <jaggz> ffmpeg -y -r 1/10 -f concat -safe 0  -i "$tmpf" -c:v libx264 -vf 'fps=25,format=yuv420p,scale=1280:ih*720/iw' "$ovid"
[12:37:11 CEST] <jaggz> option framerate not found too.. sigh
[12:38:56 CEST] <JEEB> that's an AVOption only available in demuxers or decoders that have that option
[12:39:01 CEST] <JEEB> it's not a global option like -r
[12:43:07 CEST] <jaggz> JEEB, hrm
[12:43:31 CEST] <JEEB> so most likely -f concat does not have an AVOption called 'framerate' :P
[12:44:31 CEST] <jaggz> hmm
[12:50:35 CEST] <jaggz> [swscaler @ 0x558173ffb7c0] deprecated pixel format used, make sure you did set range correctly
[12:50:40 CEST] <jaggz> one page says I can ignore that
[12:50:47 CEST] <jaggz> [out_0_0 @ 0x55817553a1c0] 100 buffers queued in out_0_0, something may be wrong.
[14:24:47 CEST] <Fyr> guys, is there any problem with AC-3 in MP4? I remux into MP4, PotPlayer doesn't see the AC-3 tracks in the file.
[14:25:44 CEST] <JEEB> not all demuxers support it, that's all. recent FFmpeg should mux it just fine
[14:25:56 CEST] <JEEB> see which demuxer PotPlayer is using?
[14:27:02 CEST] <Fyr> JEEB, FFMPEG
[14:27:33 CEST] <JEEB> then they built it without AC3 or something? or just using something really old
[14:27:57 CEST] <JEEB> because I'm pretty sure mpv which I build with recent FFmpeg is JustFine with AC3 (and E-AC3) in MP4
[14:29:11 CEST] <JEEB> and some other things like L-SMASH can demux/mux AC3 in MP4 as well
[14:29:16 CEST] <JEEB> if I recall correctly
[14:29:48 CEST] <JEEB> but yea, it is mostly a case of what implements it, and which features are enabled in FFmpeg if someone uses FFmpeg in a corporate'ish environment
[14:29:50 CEST] <Fyr> yeah, the problem was in the demuxer.
[14:38:03 CEST] <GamleGaz> Is it possible to have a custom video playback stream, where we manually add data?
[14:38:43 CEST] <BtbN> what?
[14:40:09 CEST] <GamleGaz> Ehm, we are building a bitTorrent video streaming client from scratch, and have a way of downloading the video-chunks in order
[14:40:50 CEST] <sfan5> >bittorrent streaming
[14:40:51 CEST] <sfan5> aaa
[14:40:56 CEST] <GamleGaz> I have found some ffmpeg stream options but they seem to be using rtsp streams
[14:42:49 CEST] <sfan5> to answer your question: i'm pretty sure you can use an avio context for that
[15:41:31 CEST] <wiak> soo anyone know how to embed vtt subs into webm?
[15:41:40 CEST] <wiak> or can point me to a guide?
[15:42:48 CEST] <kepstin> iirc webm doesn't permit embedded subtitle tracks
[15:43:07 CEST] <kepstin> for html5 video tags, you'll want to use separate subtitle files for now.
[15:43:30 CEST] <kepstin> (you can embed webvtt into matroska fine, of course)
[16:06:08 CEST] <wiak> good to know, kepstin thanks
[17:29:20 CEST] <Sanbatista> Test
[17:29:51 CEST] <DHE> Sanbatista: test passed
[17:32:39 CEST] <teratorn> hi all, when you seen an error like this: [Parsed_concat_6 @ 0xa28c040] Input link in5:v0 parameters (size 1920x1080, SAR 853:848) do not match the corresponding output link in0:v0 parameters (1920x1080, SAR 1:1)
[17:32:57 CEST] <teratorn> i.e multiple videos going in to a concat with different SAR values ( I think ), what is the options to fix it?
[17:33:16 CEST] <teratorn> I know I can throw a setsar filter in there, but I don't want videos to have messed up aspect ratios
[17:34:22 CEST] <DHE> the aspect ratios are already messed up. SAR 1:1 has square pixels, 853:848 is slightly not square.
[17:34:46 CEST] <teratorn> yes but I'm assuming it is "supposed" to be that way
[17:35:30 CEST] <teratorn> is there any way to change to 1:1 SAR (it would be a lossy operation on the frame, right?)
[17:35:34 CEST] <DHE> by default ffmpeg's rescaler doesn't affect the image size. if you took a 1920x1080 video and resized it to 640x480 it would still fill a 16:9 aspect screen. the SAR would be adjusted to make non-square pixels so that the resulting image is the same ratio
[17:35:41 CEST] <DHE> no, SAR is metadata. it can be changed safely
[17:36:05 CEST] <furq> if you think the sar is correct and you want to concat to a 1920x1080 video then you'd have to crop and pad
[17:36:34 CEST] <furq> but that's such a tiny difference that you might as well just change the sar to 1
[17:36:37 CEST] <DHE> I'm guessing that if the pixels are ALMOST square then it was probably an editing mistake somewhere and just setting the SAR to 1:1 would be fine..
[17:36:41 CEST] <furq> right
[17:37:03 CEST] <teratorn> sure but this is just the first error in a video processing pipleline that will handle all kinds of unforseen content
[17:38:02 CEST] <furq> corrected for sar you'd end up with 1908*1080
[17:38:15 CEST] <teratorn> DHE: so you're saying the rescaler maintains the DAR regardless of the dimensions selected?
[17:39:03 CEST] <teratorn> furq: we're happy with padding, so how did you come up with that?
[17:39:18 CEST] <furq> 1920 / (853/848)
[17:39:32 CEST] <teratorn> ah, right
[17:39:50 CEST] <furq> there's probably some way to figure it out automatically in the crop params
[17:40:13 CEST] <DHE> teratorn: by default, yes that's how it rescales. it prefers to not distort the user-seen image
[17:40:27 CEST] <furq> i would consider the theory that it's a mistake though
[17:40:29 CEST] <teratorn> DHE: makes sense, ty
[17:40:42 CEST] <furq> i'm not sure how you would get a 1080p file with that sar
[17:41:19 CEST] <teratorn> no idea where it came from.. it's just some mp4/x264 video
[17:41:50 CEST] <DHE> maybe someone did some minor cropping and bumped into the requirement that the encoded video width/height be a multiple of 16 (or other number) and now we're here.
[17:42:12 CEST] <teratorn> sounds plausible :)
[17:42:48 CEST] <furq> actually yeah sorry you should be scaling, not cropping
[17:43:07 CEST] <furq> scale=w=iw/sar
[17:43:50 CEST] <furq> that's a huge oversimplication but you get the idea
[17:43:53 CEST] <furq> fi
[17:48:00 CEST] <teratorn> I can scale to change the SAR to 1:1 ?
[17:50:01 CEST] <kepstin> yeah, downscale the image slightly in one direction (which one depends on the sar) to get it as close to 1:1 as possible - it might not be possible to get it exact due to rounding - then use setsar to force it to 1:1, then pad to your desired output size.
[17:50:08 CEST] <teratorn> to be clear, the DAR is w:h of the image frame in pixels?
[17:50:19 CEST] <teratorn> SAR is the virtual ratio of each pixels width:height
[17:50:30 CEST] <kepstin> teratorn: no, dar is the size that the video will be displayed by the player
[17:50:53 CEST] <kepstin> sar and dar are different ways of saying "here's how much you need to stretch the video during playback"
[17:50:57 CEST] <teratorn> so after factoring in the SAR?
[17:51:20 CEST] <kepstin> note that you can't change dar and sar independently, they are related to eachother via the video size
[17:51:20 CEST] <teratorn> kepstin: ok
[17:51:26 CEST] <furq> dar is (w / h) * sar
[17:51:54 CEST] <furq> non-square pixels were a thing for a while, hence the distinction
[17:52:03 CEST] <furq> they probably still are but less so nowadays
[17:52:32 CEST] <kepstin> if you use the scale filter, it preserves dar and changes sar. If you use the setsar filter, it preserves w/h and changes dar. If you use the setdar filter, it preserves w/h and changes sar.
[17:52:37 CEST] <Johnjay> >non-square pixels
[17:52:49 CEST] <teratorn> kepstin, furq, DHE: thx, I think I get what I need to do now! :)
[17:53:02 CEST] <kepstin> the crop filter preserves sar and changes dar :)
[17:53:11 CEST] <furq> i feel like force_original_aspect_ratio should be involved here somehow
[17:53:16 CEST] <furq> but i've never needed to touch it
[17:56:16 CEST] <teratorn> there should be a "MAKE THIS SHIT COMPATIBLE IN THE LEAST USER VISIBLE WAY" filter
[18:55:57 CEST] <Newb1112> Anyone here familiar with the nvenc encoder on Windows 10? I'm getting this error when trying to utilize it, "Cannot load nvEncodeAPI64.dll [h264_nvenc @ 0000024f550bdc00] The minimum required Nvidia driver for nvenc is 378.66 or newer" but the installed version of the nvidia drivers is 391.35.
[18:58:30 CEST] <Newb1112> I'm using the default Windows 10 x64 build of ffmpeg, version 3.4.2
[19:01:53 CEST] <sfan5> loading the wrong DLL? do you have older driver versions installed? (if that's even possible)
[20:26:54 CEST] <JohnDee> Hi guys ! I was wondering, is there any way to retrieve a mp4 video duration, without using these boxes : movie / movie header / duration ? It seems that I can't retrieve the duration of some videos with this method
[20:37:17 CEST] <ChocolateArmpits> JohnDee, did you try using ffprobe ? I think """ -show_entries format=duration """ should work
[20:38:18 CEST] <c_14> not on videos without or with an incorrect index
[20:38:24 CEST] <c_14> on those you'll have to find the duration yourself
[20:55:23 CEST] <JohnDee> Thanks a lot guys. Yes I know about the ffprobe option, but I'm trying other options :)
[20:58:04 CEST] <JohnDee> I'm on android so, I'd like to try with the metadatas
[22:01:07 CEST] <akkad> what licnese is ffmpeg under?
[22:01:46 CEST] <furq> lgpl2, gpl2 or gpl3
[22:01:53 CEST] <furq> depending on which libs you link it to
[22:02:24 CEST] <kepstin> akkad: the license of resulting binary depends on which options you enable at build, and some combinations of options give you unredistributable binaries (no license to distribute)
[22:05:23 CEST] <akkad> ahh. saw it linked in the MijiaCamera.exe for mi sphere camera app for windows
[22:05:38 CEST] <akkad> statically linked
[22:06:13 CEST] <akkad> f:\camera\mi_sphere_camera_windows\src\mijiacamera\ffmpeg\mp4join.cpp
[22:37:32 CEST] <fhi3u4> Quick question. Is SAMPLE-AES supported by ffmepg?
[22:38:56 CEST] <furq> https://github.com/FFmpeg/FFmpeg/blob/master/libavformat/hls.c#L1241-L1243
[22:38:58 CEST] <furq> doesn't look like it
[22:39:47 CEST] <fhi3u4> Ok thanks
[00:00:00 CEST] --- Wed Apr 11 2018