[Ffmpeg-devel-irc] ffmpeg.log.20170324

burek burek021 at gmail.com
Sat Mar 25 03:05:01 EET 2017


[01:03:31 CET] <hiru> hello everyone! is it normal having inaccurate cuts using -ss and -to parameters? I tried using a starting point but the output video starts a minute and half away from the starting point I declared..
[01:05:09 CET] <hiru> this is the line I used: ffmpeg -i <input> -ss <starting-point> -to <ending-point> -c:v libxh264 -preset medium -crf 17 <output>
[01:16:15 CET] <znf> Hi. Anyone has any idea why would a rtsp stream work just fine in VLC but not in ffmpeg? I get "Output file #0 does not contain any stream" when I try to transcode it
[01:17:19 CET] <furq> znf: try increasing probesize and/or analyzeduration
[01:17:22 CET] <znf> ffprobe finds the stream, tough it can't seem to determine the resolution, but ffplay doesn't work either (I get method not allowed)
[01:19:46 CET] <znf> I increased them to 999999
[01:19:53 CET] <znf> still nothing
[01:21:45 CET] <furq> the default for both is 5000000 so that's not a surprise
[01:21:58 CET] <furq> try 50M or something
[01:25:49 CET] <marbar> hey everyone. does anyone here have experience w/ hls on ios?
[01:25:58 CET] <marbar> i'm segmenting a source video with this command:
[01:26:04 CET] <marbar> ffmpeg -i 9 -map 0 -c:v libx264 -c:a aac -deinterlace -f ssegment -segment_list hls/9/list.m3u8 -segment_list_flags +live -segment_time 10 hls/9/%03d.ts
[01:26:12 CET] <marbar> and on ios (only) it stops after about a minute and a half
[01:26:19 CET] <marbar> is there some flag i can add to ensure playback on the ipad?
[01:30:24 CET] <znf> furq, nah, still getting 405 method not allowed
[01:31:43 CET] <dystopia_> marbar if your doing -c:a you don't need to specify the codec
[01:31:48 CET] <dystopia_> as your just remuxing from source
[01:32:00 CET] <dystopia_> as to your issue, i imagine it's an issue with ios
[01:32:30 CET] <xtina> hey guys. it's been a while since i've been in here, i'm the one working on livestreaming audio/video from my pi zero
[01:32:40 CET] <marbar> dysfigured, thanks, i'll change that. yes, it seems to be, as it plays fine everywhere else.
[01:32:42 CET] <dystopia_> as your live encoding, when you open the video on the ipad it see's it's currently like 90s long or whatever and quits playback when it reaches that point
[01:32:48 CET] <xtina> i'm looking for some advice on adding timestamps to the audio and video so they can stay in sync. they both need to be timestamped off of 1 clock
[01:32:56 CET] <dysfigured> marbar: you should get a better irc client :^)
[01:33:07 CET] <marbar> dysfigured, haha, ok, what do you recommend?
[01:33:15 CET] <xtina> i'm currently recording audio using SOX and video using ffmpeg. is anyone familiar with how i might add timestamps for both off of one clock?
[01:33:19 CET] <dysfigured> or, if you're using weechat, configure it so that it tab-completes the last active nick
[01:33:41 CET] <marbar> oh my! i see. sorry dysfigured and dystopia_
[01:33:51 CET] <dysfigured> if weechat,  /set irc.look.nick_completion_smart speakers
[01:34:03 CET] <marbar> dysfigured, i'm on xchat, just sloppy i guess
[01:34:10 CET] <znf> furq, ah, got it, apparently I have to specify tcp mode
[01:34:32 CET] <dysfigured> it's all good, just being a cheeky git. weechat is super nice tho, if you're a cli nerd i highly recommend it
[01:34:39 CET] <xtina> just to add: i want to add timestamps without re-encoding the video
[01:34:42 CET] <xtina> since I'm really CPU bound
[01:34:47 CET] <marbar> dysfigured, mmm, i'll check it out, thanks
[01:34:53 CET] <xtina> can ffmpeg timestamp my h264 encoded video file without reencoding it?
[01:35:00 CET] <dysfigured> i recommend it even if you're not a cli nerd actually. anyway, i'll leave now /me waves bye
[01:35:11 CET] <marbar> dystopia_, do you happen to know of any flags that smooth out videos on ios? like is there some particular thing that they uniquely disallow or choke on?
[01:35:55 CET] <znf> xtina, not really, video doesn't work that way
[01:37:09 CET] <xtina> @znf: hmm, i see. i got the following response to my question on the raspberry pi forums: "Every camera frame has a timestamp on it taken from the SoC STC (System Time Clock). The same STC is used as the Linux reference. The V4L2 driver already converts between the two clock domains, and raspivid has the -pts option to save a file of timestamps that mkvmerge can handle ("mkvmerge -o video.mkv --timecodes 0:timecodes.txt video.h264". D
[01:37:43 CET] <xtina> i'm having a hard time understanding it though :( what are the two different clock domains? what clock is there besides the SoC clock? Why is conversion needed?
[01:38:04 CET] <znf> oh, nevermind, you meant proper timestamps not just overlay, ignore me
[01:38:11 CET] <xtina> znf: yes, not overlay
[01:40:25 CET] <marbar> does anyone know the best place to find presets that work with current ffmpeg builds? an ios preset in particular?
[01:51:19 CET] <xtina> i'm trying to add timestamps to my audio recording
[01:51:39 CET] <xtina> i'm currently using SOX for the recording but might be willing to use a different lib if it allows me to add timestamps
[01:51:44 CET] <xtina> any ideas?
[01:52:18 CET] <xtina> is this something i might have to code myself?
[02:13:49 CET] <xtina> has anyone used mkvmerge before? it does absolutely nothing but hang for me
[02:13:51 CET] <xtina> i'm on a pi zero
[02:14:23 CET] <xtina> @furq: do you have any ideas how I can add timestamps to my audio and video? i thought it would be simple but i'm making no headway
[02:14:49 CET] <xtina> on the video front I have a MKV file as well as a txt file of timestamps.
[02:14:53 CET] <xtina> can ffmpeg accept this?
[02:15:06 CET] <xtina> or do i need to merge them using some other lib first?
[02:40:53 CET] <moneylotion> hey ffmpegers, anyone have some pointers for how to batch remove commercials from ripped videos?
[02:41:16 CET] <moneylotion> i was trying mythtvcommflag, but that only works with myth videos from what i've read
[03:08:52 CET] <Anthropohedron> Is there any way to tell from a .VOB (or .BUP or .IFO) file what language the audio or subtitle tracks are? I've been fighting with a particular DVD for over a week now. The first .VOB of the title has native audio (Chinese, in this case) as its first audio stream and English as its first subtitle stream. But it also has English, French, and Spanish audio dubs, and French and Spanish subtitles. In subsequent .VOB file
[03:08:53 CET] <Anthropohedron> s the audio and subtitle streams are in different orders. It's maddening.
[04:22:11 CET] <alexpigment> moneylotion: i've never found that process to be something you can do automatically
[04:23:17 CET] <alexpigment> perhaps it's because american broadcasts are horrible at allowing lead-ins and lead-outs, but there's often more black frames in between commercials than there are in between the commercial and program itself. i just end up using VideoRedo and do it manually
[04:23:27 CET] <alexpigment> depends on the channel, i'm sure
[04:39:04 CET] <moneylotion> im using playon.tv - it's hit or miss, but it actually added chapters in the video file for ads... anyway ffmpeg could clip chapters out?
[05:27:56 CET] <xtina_> hey guys. i'm struggling to add mmap access to my soundcard.
[05:28:05 CET] <xtina_> i've added mmap_emulation true to my alsa config: http://vpaste.net/rBKeK
[05:28:19 CET] <xtina_> i've also tried adding dtoverlay=i2s-mmap to /boot/config.txt
[05:28:29 CET] <xtina_> but i get the error that mmap_emulation is not a recognized param:
[05:28:50 CET] <xtina_> http://vpaste.net/2k9HP
[05:28:53 CET] <xtina_> any suggestions?
[06:00:07 CET] <xtina_> also, does anyone know how to get timestamps for ALSA audio? i thought it would be easy but... no pointers..
[06:13:50 CET] <xtina_> anyone know if i can pass a timestamp input file to ffmpeg along with a video source?
[06:50:39 CET] <xtina_> hey guys, i'm trying to give ffmpeg a txt file of timestamps along with the video file i'm passing in. i just want to know if that is possible
[06:50:42 CET] <xtina_> ideas?
[07:16:40 CET] <xtina> Hey guys, I want ffmpeg to accept a text file of timestamps along with my h264 input, instead of generating timestamps based on an FPS argument (-r)
[07:16:48 CET] <xtina> is there any way for ffmpeg to accept an input file of timestamps?
[07:17:03 CET] <xtina> if not, any pointers on how I could modify ffmpeg to do so? I am willing to write my own patch if it isn't really complex
[07:29:14 CET] <panpeter> Good morning everyone, ive want to compile gst-plugins-bad myself, my gst-launch says its version 1.8.4, the current master of plugins bad has 1.11.2, do i have to use matching versions or is it possible to mix up versions?
[07:31:41 CET] <xtina> can anyone point me to where in ffmpeg "timestamps" are generated based on the input -r, since that's the part i assume i need to patch?
[08:41:38 CET] <xtina> hey guys. i'm running into an issue where i'm using ffmpeg to stream and i'm sometimes in low wifi-signal areas. when this happens, i end up overflowing my 1MB FIFO audio pipe, and my ALSA buffer_size is a sad 62kb.
[08:41:50 CET] <xtina> I believe 1MB is the upper limit for a FIFO pipe size so i can't increase this
[08:42:16 CET] <xtina> I tried writing the audio input to a regular file instead of FIFO, hoping that my theoretical 'buffer' would not be constrained by the 1MB limit
[08:42:48 CET] <xtina> but in that situation it seems like writing/reading from disk is too slow, since i end up with high delays between audio/video packets, and errors thrown in ffmpeg
[08:43:21 CET] <xtina> does anyone have ideas for how else i can increase the buffer for my audio packets beyond a paltry 1MB? I'm on a pi zero with 512MB of RAM and I'm not using more than 100MB while streaming, so I'd really like to be able to use the remaining 400MB instead of losing those packets
[09:33:52 CET] <hiihiii> hello
[09:34:12 CET] <hiihiii> how do you describe time in milliseconds?
[09:36:44 CET] <hiihiii> I've looked one example which had the x & y parameters of drawtext filter change every 30 seconds
[09:37:04 CET] <hiihiii> x=if(eq(mod(t\,30)\,0)\,rand(0\,(w-text_w))\,x)
[09:37:43 CET] <hiihiii> I'd like to know is it possible to express 30seconds in milliseconds?
[09:50:37 CET] <Rajko> please tell me how to use dxva through C api ?
[09:50:49 CET] <Rajko> the find_decoder_by_name of h264_dxva2 does not exist
[09:53:10 CET] <Rajko> i can find AHWAccell but i do not know where to plug it into my AVContext
[09:53:40 CET] <Rajko> AVCodecContext
[11:13:48 CET] <kerio> on a raspberry pi there's unnecessary copying going on with ffmpeg even if i use accelerated decoding and encoding, right?
[11:18:06 CET] <BtbN> what do you mean?
[11:19:01 CET] <kerio> ffmpeg -c:v h264_mmal -i foo.mp4 -c:v h264_omx bar.mp4
[11:30:58 CET] <mad_ady> hello all!
[11:31:51 CET] <mad_ady> quick question - I want to make a movie out of an image sequence. The thing is - I already have the sequence file names in a variable space separated (e.g. ./83dad79139d8cccd498d_1490320502.png ./83dad79139d8cccd498d_1490320562.png ./83dad79139d8cccd498d_1490320621.png)
[11:32:15 CET] <mad_ady> do I need to prefix each file with -i, or one -i should be enough at the beginning?
[11:32:34 CET] <mad_ady> the order I'm giving them in is chronological
[11:33:01 CET] <DHE> I think the easiest answer is to make a symlink sequence to put them into numerical order and import tnat
[11:33:02 CET] <DHE> that
[11:34:12 CET] <mad_ady> hmm, sounds like a lot of work, but I understand the concept
[11:34:28 CET] <mad_ady> let's see if adding -i before each file works
[11:34:42 CET] <mad_ady> though I'm talking about thousands of files
[11:34:55 CET] <DHE> no. it won't
[11:35:12 CET] <mad_ady> that's because it should all be one input?
[11:35:13 CET] <DHE> ffmpeg supports multiple input files, usually for the intention of taking audio from file 1 and video from file 2, etc.
[11:35:23 CET] <mad_ady> I see
[11:35:34 CET] <DHE> you'd need to use a concatenation filter to turn it into a movie, and that becomes its own logistical nightmare
[11:36:05 CET] <mad_ady> there is no separator between files to show ffmpeg it's an explcit list, right?
[11:36:13 CET] <DHE> ffmpeg
[11:36:41 CET] <DHE> ffmpeg's commandline is loosely:   ffmpeg [input1 options] -i input1 [input2 options] -i input2 .... [output1 options] output1 [output2 optinos] output2 ...
[11:36:50 CET] <DHE> though most times it's only 1 input and 1 output
[11:36:55 CET] <mad_ady> yes
[11:36:57 CET] <DHE> with a few options being global
[11:37:14 CET] <mad_ady> I've played with stream mapping in the past, and that's not what I'm trying to do in this case
[11:37:24 CET] <mad_ady> I'll give your symlink suggestion a try
[11:37:45 CET] <mad_ady> maybe use /tmp or a ramfs to reduce wear on the disk
[11:37:58 CET] <mad_ady> will try it after lunch, thanks :)
[11:39:43 CET] <jkqxz> It would probably be easier to cat the files together in order and use image2pipe.
[12:12:26 CET] <animaonline_> Hi there, I was wondering how I can programmatically re-stream a local RTSP stream to an external one. Any starting directions would be greatly appreciated.
[12:18:04 CET] <vlt> Hello. What is the recommended way to stream a USB webcam video in most realtime possible to another machine? (Both in same 100 Mbit/s LAN, Linux)
[12:32:43 CET] <furqan> i have problem executing this http://pastebin.com/FJTvci6j
[12:32:52 CET] <furqan> ffmpeg hangs while executing that
[12:38:16 CET] <panpeter> Hello, ive checked out gstreamer-plugins-bad repo and compiled everything, is it possible to only add a single plugin to my working installation? i.e. copy [plugin].so into my /usr/lib/x86_64-linux-gnu/gstreamer-1.0/ folder, will this work?
[12:39:46 CET] <BtbN> this is #ffmpeg
[12:48:01 CET] <hiru> hello everyone. I have a question, may I?
[13:06:25 CET] <DHE> hiru: don't ask to ask, just ask
[13:12:32 CET] <mad_ady> @DHE: I implemented your method and worked wonderfully
[13:12:36 CET] <mad_ady> Thank you for the suggestion
[13:12:44 CET] <mad_ady> and have a nice weekend
[13:38:26 CET] <furqan> hi, i want to trim video depending concated video length in complex filter
[13:38:31 CET] <furqan> how can i do that?
[13:49:46 CET] <vlt> Can ffmpeg take a video source (like /dev/video0 created by what appears as "usbcore: registered new interface driver uvcvideo" in dmesg), encode it using a low latency (libx264?) mode and then provide it via network to any "client" that wants to play that?
[13:50:04 CET] <vlt> I mean can ffmepg itself act as the "server" part of that setup?
[13:51:57 CET] <JEEB> it can do multicast if that's what you mean as means for low-latency streaming
[13:52:24 CET] <JEEB> for HTTP and stuff there's some "listen" mode, but generally I wouldn't recommend it, and you have better alternatives
[13:52:44 CET] <JEEB> also low latency wise you will most likely have to tweak some parameters, but libx264 most definitely can do it
[13:53:17 CET] <JEEB> libavformat and libavcodec's shenanigans etc require tweaking, but people have done it
[13:53:20 CET] <vlt> JEEB: What would be a better alternative to provide (a) client(s) with the video stream?
[13:53:32 CET] <JEEB> well with low latency you need UDP
[13:53:43 CET] <JEEB> and multicast/unicast is supported in lavf
[13:54:30 CET] <vlt> JEEB: Ok, so ffmpeg _is_ the "server" but just not in HTTP mode, right?
[13:56:18 CET] <JEEB> yes
[14:12:49 CET] <marbar> hi folks
[14:13:16 CET] <marbar> so, i've been trying to encode this video all sorts of ways, and it always fails on ios right around the same place (a little over a minute in)
[14:13:41 CET] <marbar> does anyone know of a set of flags that guarantee ios playback?
[15:07:59 CET] <alexpigment> marbar: what device are you trying to play it on?
[15:08:39 CET] <alexpigment> and what is your current command line?
[15:14:05 CET] <alexpigment> marbar: well, in absense of an answer, here's a sample that will work on iphone 4s and above (and probably the iphone 4, but unofficially so)
[15:14:38 CET] <alexpigment> ffmpeg -i [infile] -f mp4 -vcodec libx264 -s 1920x1080 -aspect 16:9 -crf 18 -level 4.1 -profile:v high -r 30000/1001 -g 30 -acodec aac -ar 44100 -ac 2 -ab 160000 [outfile]
[15:22:16 CET] <obamoose> hello!
[15:22:43 CET] <alexpigment> hey obamoose
[15:23:03 CET] <obamoose> http://pastebin.com/1PP7S6ig
[15:23:13 CET] <obamoose> will this eternally loop my cgn directory in a single stream?
[15:23:59 CET] <alexpigment> that's outside of my wheelhouse
[15:24:04 CET] <obamoose> ;_;
[15:24:13 CET] <alexpigment> however, why do you have -f -f flv at the end?
[15:24:18 CET] <obamoose> I have no idea
[15:24:21 CET] <obamoose> I'm pretty bad with ffmpeg
[15:24:39 CET] <alexpigment> k
[15:25:01 CET] <alexpigment> well i'd wait for others here to chime in
[16:43:34 CET] <marbar> alexpigment!!! sorry, i was afk :) thanks for the tip, running it now
[16:43:45 CET] <marbar> alexpigment, i'm testing on an ipad
[16:44:22 CET] <marbar> alexpigment, i modified your sample because i'm using hls
[16:44:31 CET] <marbar> so: ffmpeg -i 9 -vcodec libx264 -s 1920x1080 -aspect 16:9 -crf 18 -level 4.1 -profile:v high -r 30000/1001 -g 30 -acodec aac -ar 44100 -ac 2 -ab 160000 -f ssegment -segment_list hls/9/list.m3u8 -segment_time 10 hls/9/%03d.ts
[16:54:26 CET] <alexpigment> k, does that appear to work or no?
[16:54:59 CET] <marbar> it's still running... i'm gonna stop it and just do a small chunk.
[16:55:56 CET] <marbar> ok this'll be a couple minutes but that's enough to test the part that messed up. thanks again alexpigment you're my hero.
[16:56:26 CET] <alexpigment> ok, if it doesn't work, it may be because older ipads don't officially support 1080p
[16:56:26 CET] <marbar> alexpigment, are you a dev?
[16:56:33 CET] <marbar> ah i c
[16:56:41 CET] <marbar> yeah, before i was encoding with way lower settings
[16:56:44 CET] <alexpigment> no, just a guy who has had to do a lot of custom encoding in the past
[16:56:44 CET] <marbar> going off old info i guess
[16:56:57 CET] <marbar> oh man, i feel for ya
[16:57:26 CET] <alexpigment> i have a spreadsheet with all the specs of all Apple devices. some of the devices support higher video specs than they advertise, but it's a good guideline
[16:57:38 CET] <marbar> oh dude that's awesome. is it online somewhere?
[16:57:49 CET] <alexpigment> no, not at the moment
[16:58:06 CET] <marbar> man, i feel like it's been hard to find up to date info on this stuff
[16:58:12 CET] <alexpigment> the world of video encoding is so small that i doubt it would benefit many people
[16:58:18 CET] <marbar> haha
[16:58:21 CET] <alexpigment> well, that's the thing, it's old info and it's very googleable
[16:58:35 CET] <alexpigment> e.g. google "iphone 4 tech specs"
[16:58:44 CET] <alexpigment> they have a page and they list supported video information
[16:59:20 CET] <marbar> hm doesn't seem to work :(
[16:59:32 CET] <marbar> with that command it dies on the first chunk
[16:59:33 CET] <alexpigment> so usually it's just a matter of matching the resolution, h264 profile, h264 level, and audio bitrate (the last part i think is bs, but still)
[16:59:46 CET] <marbar> previous command was this: ffmpeg -i 9 -map 0 -c:v libx264 -c:a aac -deinterlace -f ssegment -segment_list hls/9/list.m3u8 -segment_list_flags +live -segment_time 10 hls/9/%03d.ts
[16:59:50 CET] <marbar> got as far as the 8th chunk, hah
[17:00:05 CET] <alexpigment> are you getting any errors?
[17:00:24 CET] <alexpigment> also, do you know the specs of your source material?
[17:00:26 CET] <marbar> in the encoding?
[17:00:33 CET] <marbar> ah, that's something i should figure out
[17:00:36 CET] <marbar> ffprobe, right?
[17:01:05 CET] <marbar> "Video: h264 (High) ([27][0][0][0] / 0x001B), yuv420p, 1920x1080 [SAR 1:1 DAR 16:9], 29.97 fps, 29.97 tbr, 90k tbn, 59.94 tbc"
[17:01:12 CET] <marbar> "Audio: aac (LC) ([15][0][0][0] / 0x000F), 44100 Hz, stereo, fltp"
[17:01:42 CET] <marbar> is what it says for "ffprobe list.m3u8"
[17:02:06 CET] <marbar> oh i mean that's the output
[17:02:16 CET] <marbar> the source:
[17:03:41 CET] <marbar> http://pastebin.com/nHJWmjqR
[17:04:11 CET] <marbar> above is the ffprobe of the source file
[17:04:21 CET] <marbar> i didn't get any errors in the encoding process
[17:05:46 CET] <alexpigment> ok, so i don't know why the error is happening - that seems more like an HLS-related issue to me
[17:06:48 CET] <alexpigment> but based on this information in the pastebin, i would modify the code above with the following parameters (replace what's already there) -level 4.0 -r 24000/1001 -g 48
[17:07:02 CET] <marbar> here's encoding output: http://pastebin.com/nRbu3GYT
[17:07:38 CET] <alexpigment> oh i see
[17:07:42 CET] <alexpigment> your source is 480x270
[17:07:51 CET] <marbar> oooh that's doing it?
[17:07:56 CET] <alexpigment> no
[17:08:01 CET] <alexpigment> but my settings are all wrong
[17:08:15 CET] <marbar> oh *head smack* shoulda told you this stuff up front, sorry
[17:08:21 CET] <alexpigment> again, i was basing it off your initial request, which was just that you were trying to encode for iOS and it was failing
[17:08:25 CET] <alexpigment> no worries
[17:08:50 CET] <marbar> also that's not the original original
[17:08:50 CET] <alexpigment> lemme get something better for you; 1 sec
[17:08:59 CET] <marbar> i suppose it's good practice to start with the closest to original file
[17:10:10 CET] <marbar> orig file ffprobe: http://pastebin.com/hLWm6nip
[17:10:30 CET] <marbar> i can try your command on the orig, might work better
[17:10:58 CET] <alexpigment> wait
[17:11:01 CET] <alexpigment> i'm confused
[17:11:20 CET] <marbar> sorry. i was starting from a file that i had already encoded because i thought it would maybe help.
[17:11:24 CET] <alexpigment> the pastbin earlier of "the source" was also 480x270
[17:11:29 CET] <alexpigment> so is *this* the actual source?
[17:11:38 CET] <alexpigment> the newest link you sent?
[17:11:38 CET] <marbar> well the first paste was the source i was using
[17:11:51 CET] <marbar> the later paste is the source that i encoded to get it
[17:11:57 CET] <marbar> which i should probably use instead, right?
[17:12:03 CET] <marbar> i mean, of course
[17:12:15 CET] <alexpigment> yeah, i think it's better to do one encode at a time
[17:12:22 CET] <alexpigment> rather, one encode period
[17:12:25 CET] <alexpigment> rather than two
[17:12:31 CET] <marbar> good call
[17:12:41 CET] <alexpigment> but your goal is to output to 480x270?
[17:12:49 CET] <alexpigment> or remain in 720p?
[17:13:30 CET] <marbar> well i think the 480x270 came in because i resized based on some old info
[17:13:37 CET] <marbar> thinking it would make it work
[17:13:48 CET] <alexpigment> ok, let's go for 1280x720 first
[17:13:50 CET] <alexpigment> 1 sec
[17:13:51 CET] <marbar> so i guess i should just forget that intermediate file. i'll try your command on the orig.
[17:13:55 CET] <marbar> ok thanks :)
[17:14:43 CET] <alexpigment> so this is for streaming?
[17:14:57 CET] <alexpigment> i guess the bitrate should be constained a little bit then
[17:15:08 CET] <marbar> basically when users upload videos i want to make an hls version for ios
[17:15:21 CET] <marbar> (hls process currently works for android tablet)
[17:15:30 CET] <alexpigment> ok, so this 720p is not going to be a static thing?
[17:15:32 CET] <marbar> so streaming yes, but not live
[17:15:48 CET] <alexpigment> like, the input resolution will constantly be changing based on what is uploaded?
[17:15:57 CET] <marbar> well i could try to constrain that initially
[17:16:22 CET] <marbar> but you would essentially need different commands depending?
[17:16:27 CET] <alexpigment> no
[17:16:34 CET] <alexpigment> lemme refresh myself on constaining resolution
[17:16:35 CET] <marbar> oh ok. then i suppose i need something generic
[17:16:48 CET] <marbar> dude thank you so much. you're a real lifesaver.
[17:17:54 CET] <alexpigment> don't thank me yet - i don't know if any of this will work. my experience is piping from a renderer at any size and formatting to the upper limit of the device's capabilities
[17:17:59 CET] <alexpigment> no HLS whatsoever
[17:18:27 CET] <marbar> that sounds cool though
[17:19:40 CET] <marbar> oh geez
[17:19:43 CET] <marbar> i feel like such a troll
[17:20:11 CET] <marbar> gosh i just realized this is the orig: http://pastebin.com/n7tJpfsu
[17:20:19 CET] <marbar> i have so many video files floating around
[17:20:40 CET] <marbar> this came fresh outta my friend's final cut pro
[17:20:40 CET] <alexpigment> man, the confusion stacks ;)
[17:20:47 CET] <marbar> i know :( haha
[17:20:56 CET] <marbar> you know how these all-nighters go...
[17:21:26 CET] <alexpigment> ok, so here's the deal - your original code is very bare-bones. there isn't a whole lot of limiting to the constraints of the device
[17:21:28 CET] <alexpigment> BUT
[17:21:48 CET] <alexpigment> at 480x270, you could encode this any number of ways and it wouldn't cause a problem
[17:22:53 CET] <alexpigment> what i do find interesting is that the stream 0:0 of this original file is audio, and stream 0:1 is video
[17:22:58 CET] <alexpigment> usually it's the other way around in my experience
[17:23:08 CET] <marbar> oh right i noticed that!
[17:23:14 CET] <marbar> weird right? i wonder if that's it...
[17:23:24 CET] <alexpigment> well, you were using -map earlier
[17:23:27 CET] <alexpigment> and i didn't know why
[17:23:33 CET] <marbar> heh, neither did i
[17:24:51 CET] <marbar> i'm just gonna try the command you suggested on the _real_ orig
[17:24:52 CET] <alexpigment> well, i think i'm officially at that point - which i often get to here - where a simple question that i can certainly answer reveals itself to be a complex problem involving things i've never touched before
[17:24:59 CET] <marbar> maybe everything will be fine once i'm not encoding it ten times
[17:25:05 CET] <alexpigment> well, the real original is 480x270
[17:25:10 CET] <alexpigment> so you're upscaling
[17:25:25 CET] <marbar> oh yeah so it is
[17:25:31 CET] <marbar> does that cause problems sometimes?
[17:25:51 CET] <alexpigment> how about just doing -i [infile] -c:v libx264 -crf 20 -c:a aac -b:a 128000
[17:26:05 CET] <marbar> ok i'll give it a shot...
[17:26:14 CET] <alexpigment> no, upscaling usually works fine, but it's just unnecessary work if you're going from 480x270 to 1920x1080
[17:29:10 CET] <marbar> :-\ ran it on orig, chokes on second chunk
[17:31:37 CET] <obamoose> hello
[17:32:00 CET] <obamoose> i'm trying to create a stream that never stops
[17:32:09 CET] <obamoose> but my stream ends after every video file I stream
[17:32:13 CET] <obamoose> how do i fix this ;_;?
[17:32:17 CET] <obamoose> http://pastebin.com/1PP7S6ig
[17:32:21 CET] <marbar> alexpigment, ok, i tried making a regular mp4 instead of doing hls anything and it doesn't play in ios
[17:32:31 CET] <marbar> there's something messed up about the orig that i can't seem the shake off
[17:33:33 CET] <alexpigment> that seems so weird
[17:33:46 CET] <alexpigment> you're playing it on iOS how?
[17:33:54 CET] <marbar> yeah totally. my friend exported it from final cut pro
[17:34:05 CET] <marbar> i'm opening it in chrome
[17:34:07 CET] <marbar> on an ipad
[17:34:11 CET] <alexpigment> oh ok
[17:34:13 CET] <marbar> served from my computer
[17:34:16 CET] <alexpigment> well, there's a potential point of failure there
[17:34:42 CET] <alexpigment> again, out of my wheelhouse because i'm not a webdev
[17:34:52 CET] <alexpigment> but i know that sometimes you have to define the mime types in a certain way
[17:35:11 CET] <marbar> hm interesting
[17:35:20 CET] <marbar> because it's only happening every once in a while
[17:35:25 CET] <marbar> that is, not on every chunk
[17:35:42 CET] <marbar> it's possible, i suppose, that certain chunks are misidentified for whatever reason, resulting in wrong mime types like you say
[17:36:04 CET] <alexpigment> you could try renaming it to .m4v as a test
[17:36:36 CET] <alexpigment> Apple generally prefers .m4v for mp4 files, but in my experience, it's only because it allows them to associate a non-standard extension with iTunes
[17:37:29 CET] <alexpigment> but a stopgap solution would be to change the mimetype to video/m4v
[17:37:45 CET] <alexpigment> type="video/mp4" in your source tag
[17:37:47 CET] <alexpigment> er
[17:37:51 CET] <alexpigment> video/m4v
[17:37:56 CET] <marbar> still messes up :-\
[17:38:02 CET] <marbar> but differently interestingly
[17:38:06 CET] <alexpigment> yeah, not sure then
[17:38:18 CET] <alexpigment> i suppose you can't use safari in ios for this?
[17:38:24 CET] <marbar> as a .m4v it gets really messed up but continues playing
[17:38:33 CET] <marbar> whereas as an .mp4 it doesn't even play
[17:39:09 CET] <marbar> but yeah it's super messed up. all green, sound and video both in and out, now frozen
[17:39:12 CET] <alexpigment> ok, but what about keeping .mp4 and changing the source to type="video/m4v"
[17:39:27 CET] <alexpigment> green video usually means missing information in the video stream
[17:39:30 CET] <marbar> oh i'm not even loading this up on a web page
[17:39:39 CET] <marbar> i'm just loading the file directly (in chrome)
[17:39:47 CET] <marbar> so no video tag or anything eveen
[17:42:45 CET] <alexpigment> ok, well as a test, take the original video and re-encode it with this command: ffmpeg -i [infile] -f mp4 -c:v libx264 -crf 20 -c:a aac -b:a 128000 -movflags faststart [outfile.mp4]
[17:43:03 CET] <alexpigment> play it on your computer and make sure it works
[17:43:10 CET] <alexpigment> then try the test again in chrome
[17:43:59 CET] <marbar> ok
[17:45:41 CET] <marbar> oh crap, works for about ten seconds
[17:46:05 CET] <marbar> ya know, the guy just called me. i can probably get a different cut of this video
[17:46:08 CET] <marbar> to work with
[17:46:55 CET] <marbar> (and yes, it works fine on my computer)
[17:47:29 CET] <marbar> is there some basic export setting you'd recommend from final cut pro to avoid this kinda thing?
[17:47:52 CET] <alexpigment> i haven't worked in FCP in years - i use Adobe Premiere
[17:47:58 CET] <marbar> ah i c
[17:48:05 CET] <alexpigment> so i don't know what setting to use, but it does seems weird that you're getting a 480x270 video
[17:48:26 CET] <marbar> yeah i think he used an iphone setting of some kind
[17:48:30 CET] <alexpigment> like that's not reference quality for any TV standard, new or old
[17:48:35 CET] <marbar> haha
[17:49:10 CET] <marbar> ok i'm calling him for another cut...
[17:49:36 CET] <alexpigment> yeah, see if you can get it in 1280x720
[17:49:55 CET] <marbar> ok will do
[17:50:06 CET] <alexpigment> some encoders (and decoders I suppose) have a hard time with resolutions that aren't divisible by a certain number
[17:50:21 CET] <alexpigment> 1280x720 is divisible by 16 on both width and height
[17:50:44 CET] <alexpigment> whereas 480x270 doesn't divide evenly by 4 on the height
[17:50:55 CET] <alexpigment> not that i think it should matter at all, but still
[17:57:29 CET] <marbar> dude thanks so much. all great points. i'll let you know as soon as i have a fresh source file...
[18:27:38 CET] <flomko> Hi all! I have some problem with filters/ filter_complex. My system Centos7, to run ffmpeg i use repo nux-dextop/x86_64. On previos system i run "ffmpeg -i input.mp3 -filter_complex "showwavespic=s=640x120" -frames:v 1 output.png" to create waveform, but on fresh ffmpeg it dont work. I found, that i need use -lavfi with "ffmpeg -i input.mp3 -lavfi showwavespic=split_channels=1:s=640x120
[18:27:38 CET] <flomko> output.png" and it still useless. In outpit ffmpef -filter i found that i have only "showwaves" for videooutput. Can anybody help?
[18:28:47 CET] <flomko> i try create videofile with 1 framy, but output definetly wrong "ffmpeg -i input.mp3 -lavfi showwaves=s=640x120 -frames:v 1 output.png"
[18:28:52 CET] <jfmcarreira> heyy guys
[18:29:11 CET] <jfmcarreira> is there any examples or code where i can see the new api uses? avcodec_receive_frame
[18:29:30 CET] <jfmcarreira> i am currently feeling that i am using bleeding edge functions...
[18:29:40 CET] <c_14> flomko: your version of ffmpeg is too old, update
[18:31:18 CET] <flomko> c_14 ffmpeg version 2.6.8 Copyright (c) 2000-2016 the FFmpeg developers. Maybe u can help me with link to fresh repo to centos?
[18:31:41 CET] <c_14> try a static build
[18:31:44 CET] <c_14> http://johnvansickle.com/ffmpeg/
[18:32:47 CET] <c_14> or build from source https://trac.ffmpeg.org/wiki/CompilationGuide/Centos
[18:34:08 CET] <flomko> thank y
[18:37:22 CET] <cesdo> Hello. I watch youtube videos with mpv and often get [ffmpeg] tls: The TLS connection was non-properly terminated.
[18:37:53 CET] <cesdo> Is it serious error?
[18:43:41 CET] <cesdo> [osd/libass] PlayResX undefined, setting to 960
[18:43:52 CET] <cesdo> What is 960?
[18:45:18 CET] <kepstin> a number in pixels. it's an internal detail from the subtitle renderer; if everything looks fine don't worry about it.
[18:51:40 CET] <durandal_1707> jfmcarreira: doc/examples
[19:00:55 CET] <jfmcarreira> thanks
[19:01:04 CET] <jfmcarreira> but there is not example with the new API
[19:06:43 CET] <ZexaronS> hello
[19:07:43 CET] <ZexaronS> i was wondering if it was possible to create a ffmpeg script to run on some portable device that is hooked up to a digital camera for timelapse recording
[19:08:05 CET] <dystopia_> a script to do what?
[19:08:42 CET] <marbar> alexpigment, omg dude i got sidetracked. my friend is sending me a fresh file later tonight. i'll let you know how it goes. thanks again dude!
[19:09:19 CET] <ZexaronS> The idea is, so that it works in real-time (buffer) so that a lot of the video doesn't need to be saved, but that the video is being converted to snapshots highlights, however, the script has to analyze a portion of the video to determine the occurence of highlights, that's where the for example 5 minute buffer comes in
[19:10:07 CET] <ZexaronS> so instead of me having to recoird 10 gigabytes for 24 hours and then extract and do manual screenshots, it would do this onfly
[19:11:18 CET] <ZexaronS> The point of the script would also be to save more screenshots when there is more activity going on in the picture, so when there's night and not much going on it would do only 1 screenshot per 10 minutes, when there's alot going on it would take for example 1 sceenshot every 10 seconds
[19:13:00 CET] <ZexaronS> the script would recognize the video as if it was offline, well, probably has to know it's being recorded live but with a big buffer it doesn't need actual "live" capability, plus the buffer is needed because an area of 1-2 minutes has to be analyzed in whole to determine the number of appropriate highlights based on how much change happens over that 1-2 minute span
[19:13:37 CET] <dystopia_> that seems very possible to do
[19:13:49 CET] <ZexaronS> This would create very good timelapses automatically, without having to sift through hoards of gigabytes of video and doing it manually
[19:13:58 CET] <dystopia_> save 1 frame ever x seconds or somthing
[19:14:13 CET] <ZexaronS> Indeed
[19:14:15 CET] <furq> ZexaronS: mpdecimate can probably do that
[19:14:17 CET] <furq> !filter mpdecimate
[19:14:17 CET] <nfobot> furq: http://ffmpeg.org/ffmpeg-filters.html#mpdecimate
[19:16:18 CET] <mofasa> Is there any way to produce a "libffmpeg.so" when building ffmpeg?
[19:16:41 CET] <furq> bye
[19:17:38 CET] <kepstin> mofasa: like, combine all the different ffmpeg libraries (libavcodec, libavformat, etc.) into one dll? not that I know of.
[19:17:43 CET] <ZexaronS> That's one part, but it would be rough, this is meant for construction sites in my case, ... it's also more complex because the thing has to save minimal, it has to save 1 frame each 2 minutes anyway at least, then the activity-sensor would determine the frequency of saved frames up to 10 per minute max.
[19:17:55 CET] <furq> gcc can merge multiple .a into one .so
[19:18:02 CET] <furq> but there's no way of doing it with the ffmpeg build system
[19:18:13 CET] <furq> you'd have to build with --enable-static and then merge them yourself
[19:18:20 CET] <furq> it also might break shit
[19:18:51 CET] <ZexaronS> I kinda feels so strongly about this I'd contribute myself if I had more experience
[19:19:06 CET] <mofasa> Allright.. Just need h264 for chromium, which isn't provided by openSuse any longer
[19:23:05 CET] <furq> ZexaronS: drop the input framerate to 1/6 and then set mpdecimate's max to 19
[19:33:05 CET] <ZexaronS> furq: I find that a bit too easy to do all the things I had in mind, but how close would it be?
[19:33:34 CET] <furq> "it has to save 1 frame each 2 minutes anyway at least, then the activity-sensor would determine thefrequency of saved frames up to 10 per minute max."
[19:33:37 CET] <furq> it does that
[19:33:50 CET] <furq> and how often it saves frames is based on the difference from the previous frame
[19:34:10 CET] <furq> whether it'll be a functional security camera i don't know
[19:34:51 CET] <furq> the problem with setting the input framerate to 1/6 means that if something happens but only lasts four seconds, it might not get noticed
[19:37:08 CET] <furq> i'm not an expert on all 172 video filters in ffmpeg but i can't think of any other way to do it without using external tools
[19:37:23 CET] <ZexaronS> Oh, sorry, see, it's not a security camera, it's for construction work site timelapse, a key thing is missing, it has to have a way to determine how much something has moved over a range of 1000 frames, so of a excavator has moved only a few pxiels, that's too low, we don't need to snap a highlight, but if a big truck moved in and out of the whole picture in less than 1000 frames then we need to take a highlight of that truck at
[19:37:23 CET] <ZexaronS> least once or twice
[19:37:43 CET] <furq> well yeah that's how mpdecimate works
[19:38:15 CET] <ZexaronS> highway construction timelapse, wee need a sensor of movement over larger frame batches not just difference of current to previous frame, that's too low for detecting any of that slow movement of vehicles
[19:38:26 CET] <furq> maybe
[19:39:10 CET] <ZexaronS> That's why timelapses don't look good if people do it too roughly because there's a lot of rough cuts, this would make it a lot more smooth.
[19:39:14 CET] <ZexaronS> And all automatically
[19:39:25 CET] <thebombzen> yea mpdecimate drops frames that are close the the previous one
[19:39:26 CET] <furq> this sounds like it's getting into opencv territory to do it properly
[19:39:44 CET] <ZexaronS> Forget the realtime on-fly part, i'd rather just have it work offline for starters.
[19:39:59 CET] <thebombzen> so by seting the framerate to 1/6 you are recording 10 frames per minute
[19:40:10 CET] <thebombzen> and mpdecimate will drop ones that are very similar to the previous one
[19:41:00 CET] <furq> yeah bear in mind mpdecimate is analysing one frame every six seconds
[19:41:21 CET] <furq> or whatever you set the framerate to before invoking it
[19:41:26 CET] <thebombzen> then, if you want to timelapse it correctly, you can use the setpts filter to rewrite the timestamps as a constant framerate
[19:41:42 CET] <thebombzen> so like, this will drop frames but it won't change their timestamps so you'll end up with variable framerate
[19:42:04 CET] <furq> i assumed he was outputting to images, in which case you don't need to do that
[19:42:06 CET] <ZexaronS> Yes that's all good to hear that this is roughtly possible, but what if the machines on the construction site simply chugg along and don't do much, for example if an excavator is excavating and rotating it's hand, we don't need highlights of the excavator rotating, but if the excavator moves some distance we might want to keep the a few frames of it moving on it's tracks toward a new section in the pciture
[19:42:13 CET] <furq> but if you want to output a video in the same pass then you'll need setpts
[19:42:34 CET] <furq> ZexaronS: that's about as good as you'll get with ffmpeg as far as i know
[19:42:45 CET] <thebombzen> ZexaronS: mpdecimate has a parameter on how different you want the frame to be
[19:43:02 CET] <furq> mpdecimate isn't really designed for that and i don't think anything else gets any closer
[19:43:25 CET] <furq> maybe one of the libavfilter gurus can tell us otherwise
[19:44:19 CET] <furq> if you know a bit of C then it shouldn't be that difficult to add a new filter or modify mpdecimate
[19:44:22 CET] <ZexaronS> thebombzen, I was about to ask about that as well yes, 1 frame of comparison might simply not be enough no matter how much I try because there's going to be a lot of movement of all the people , even if those people move only a little if there's 15 people strolling, each one of them adds up to the bits changing
[19:45:08 CET] <thebombzen> then you're going to have to do that one yourself probably
[19:45:36 CET] <ZexaronS> Yes I was amusing the idea of trying to have a look at libavfilter myself ... but I'm not a real programmer and if I was I'd be happy to contribute right now
[19:49:25 CET] <ZexaronS> But at least I brought it up to kinda get it familiarized, and anyone knows who these liavfilter guru's are ?
[19:54:28 CET] <ZexaronS> For starters I'll make a trac
[20:37:27 CET] <ZeroWalker> what's a simple way to compile ffmpeg on windows? (i would love to use visual studio, but i guess one has to use mingw or something)
[20:37:36 CET] <JEEB> you can use mingw-w64 or MSVC
[20:37:40 CET] <JEEB> but you need a shell in any case
[20:38:00 CET] <JEEB> MSVC needs to be 2013, 2015 or 2017 with latest update(s) installed
[20:38:27 CET] <JEEB> testing with 2012 was stopped recently due to 2017 coming out and 2012 needing a converter for the source code
[20:39:45 CET] <JEEB> ZeroWalker: for MSVC you basically call the configure script with certain parameters when the MSVC things are in your PATH (just make sure 'link' points to the MSVC link and not the msys2 thing)
[20:41:15 CET] <ZeroWalker> hmm, that's a bit beyond me, really bad at the configure script stuff, i only used it with mingw cause, well that's was the way to do things, but i don't know hot it's setup or how to modify it
[20:41:26 CET] <ZeroWalker> i do use visual studio 2017 so at least that's set
[20:48:35 CET] <JEEB> ZeroWalker: so if you want to build with msys2 + MSVC (first just for the shell and make to configure and run the compilation), make sure the tools like cl and link are in PATH (usually there's an env var which leads to vsvars32.bat or so)
[20:48:49 CET] <JEEB> and then you can try configuring it like the MSVC 2017 FATE machine
[20:49:07 CET] <JEEB> http://fate.ffmpeg.org/report.cgi?time=20170323125228&slot=x86_32-msvc15-windows-native
[20:49:29 CET] <JEEB> that has stuff like enable-gpl and enable-memory-poisoning etc enabled but those aren't required of course
[20:49:53 CET] <JEEB> --toolchain=msvc --arch=x86 or x86_64 seem to be the main things
[20:50:13 CET] <JEEB> setting the toolchain pretty much enables the MSVC build mode
[21:10:00 CET] <faLUCE> Hello. How can I generate a stereo dummy audio (like a sin() or something similar) ?
[21:10:49 CET] <JEEB> libavfilter should have filters for that
[21:11:12 CET] <faLUCE> JEEB: I mean  with ffmpeg command
[21:11:45 CET] <JEEB> yes
[21:11:58 CET] <JEEB> ffmpeg-all.html and see what filters there are that you can use from ffmpeg.c as well
[21:12:03 CET] <JEEB> https://www.ffmpeg.org/ffmpeg-all.html
[21:13:47 CET] <faLUCE> JEEB: what should I search? I searched "dummy" and found only video stuff
[21:14:23 CET] <JEEB> are you fucking incapable of doing even basic fucking searching
[21:14:24 CET] <JEEB> https://www.ffmpeg.org/ffmpeg-all.html#toc-Audio-Filters
[21:14:30 CET] <JEEB> ctrl+f "audio filters"
[21:14:37 CET] <JEEB> or just scrolled a bit down
[21:15:21 CET] <JEEB> or actually, audio sources would probably be better which is just under the filters
[21:15:34 CET] <JEEB> one of them is called "sine", even
[21:15:38 CET] <JEEB> which you can guess what it does
[21:15:52 CET] <alexpigment> anullsrc is what i've used in the past, fwiw
[21:16:05 CET] <alexpigment> but i don't know if faLUCE needs the audio to be audible
[21:16:09 CET] <JEEB> yea, any of those audio sources or filters
[21:18:29 CET] <faLUCE> there's not anything like what I asked...
[21:18:41 CET] <faLUCE> there's only a silence filter
[21:18:59 CET] <faLUCE> JEEB: sorry, I did not read "sine"
[21:19:06 CET] <kepstin> aevalsrc is the best for making dummy streams, it's pretty easy to make e.g. a sine wave
[21:19:24 CET] <JEEB> " there's not anything like what I asked..." <- fuck you
[21:19:33 CET] <JEEB> just read the fucking list you piece of shit
[21:19:44 CET] <alexpigment> well that escalated quickly...
[21:19:45 CET] <JEEB> sorry, but you've been around for long enough
[21:20:29 CET] <JEEB> alexpigment: it just pisses me off that after you've even linked the correct'ish spot on the TOC and then noted the actual specific point you still get "there's nothing like I need here"
[21:20:55 CET] <alexpigment> no, i get it. i saw the tension yesterday too. you should just actually add him to your ignore list
[21:21:03 CET] <faLUCE> JEEB: I wrote: [21:18] <faLUCE> JEEB: sorry, I did not read "sine"
[21:21:19 CET] <faLUCE> JEEB: be calm
[21:21:41 CET] <JEEB> alexpigment: he's been here for months and even used the API IIRC
[21:22:17 CET] Action: JEEB goes back to stabbing himself with google build systems
[21:22:25 CET] <furq> oh wow, a new build system
[21:22:29 CET] <furq> just what we've always needde
[21:22:30 CET] <furq> ed
[21:22:31 CET] <faLUCE> JEEB: are you just stupid or even mad?
[21:23:04 CET] <JEEB> weren't you the guy doing "low latency streaming" with lavf+lavc?
[21:23:06 CET] <furq> well, i tried
[21:23:26 CET] <JEEB> pretty sure I'm not making it up that a person with your nickname has been here for months
[21:23:31 CET] <alexpigment> furq: it was a good attempt
[21:23:42 CET] <furq> now to sit here and pretend like i'm not going to extremely enjoy this
[21:23:50 CET] Action: alexpigment grabs popcorn
[21:24:51 CET] <JEEB> around --- Day changed Thu Feb 18 2016
[21:25:27 CET] <JEEB> anyways, really. back to stabbing myself and others with gclient/ninja
[21:28:12 CET] <faLUCE> JEEB: you should see a doctor, believe me
[21:35:19 CET] <ZeroWalker> hmm
[21:35:34 CET] <ZeroWalker> let's see if i can figure this out
[21:37:09 CET] <JEEB> ZeroWalker: it shouldn't be too hard
[21:37:23 CET] <JEEB> just make sure you have an msys2 shell with MSVC's cl and link in the PATH
[21:38:36 CET] <ZeroWalker> was a long time i did this stuff. But i want to begin with installing this: https://sourceforge.net/projects/msys2/ ?
[21:39:30 CET] <JEEB> http://www.msys2.org/
[21:39:45 CET] <JEEB> the least retarded msys environment, and you will want the x86_64 one generally
[21:40:37 CET] <ZeroWalker> ah, yeah always been confused by all these versions, to me all this stuff is a total mess, link appreciated
[21:41:22 CET] <JEEB> I still haven't moved to msys2 myself since the basic tools I've got from original msys (which are ancient) still work and I've been switching the toolchain myself manually
[21:42:49 CET] <ZeroWalker> oh, or well i say oh, but not really sure what this implies. But, what does msys2 have that the old one doesn't? (i am not even sure what msys lol;P)
[21:43:10 CET] <JEEB> msys2 has a nicer package manager and isn't awfully outdated
[21:43:21 CET] <JEEB> msys bases on ancient times' cygwin
[21:43:27 CET] <JEEB> while msys2 bases on relatively recent cygwin
[21:44:19 CET] <ZeroWalker> Ah i see
[21:44:47 CET] <BtbN> Why not just use actual cygwin then, if it bases on it anyway?
[21:45:45 CET] <JEEB> BtbN: good question but most people do mingw-w64 builds and MSVC builds in msys instead of cygwin. probably because cygwin tries even harder to be not windows
[21:45:49 CET] <JEEB> like windows paths etc
[21:45:58 CET] <Cracki> ffmpeg bindings/ffi for use in python, recommendations?
[21:46:12 CET] <BtbN> Cygwin has proven very useful for me.
[21:46:18 CET] <JEEB> sure, I use it too
[21:46:27 CET] <BtbN> You can just install mingw compilers, and get rather normal binaries
[21:46:29 CET] <JEEB> this very terminal window I write this text into is on cygwin
[21:46:33 CET] <BtbN> But I'm just using MSVC in it now.
[21:46:41 CET] <JEEB> oh, MSVC works in it too? cool
[21:46:48 CET] <ZeroWalker> i have used cygwin before, i kinda like that it was "automated", but it was messy and required cygwin dll for everything which i wasn't a fan of
[21:46:49 CET] <JEEB> oh, right. probably not a surprised :P
[21:46:58 CET] <BtbN> Well, just need to get the environment variables into it
[21:46:59 CET] <JEEB> ZeroWalker: yes if you build with the cygwin compiler
[21:47:15 CET] <ZeroWalker> oh you can use msvc, that's neat
[21:47:16 CET] <JEEB> BtbN: aye. I guess I've just set up my ancient msys environment with the PATH
[21:47:27 CET] <BtbN> it needs quite a bit more stuff than PATH
[21:47:29 CET] <JEEB> which is why I'm not using cygwin for it
[21:47:36 CET] <BtbN> You just launch the cygwin stuff from an MSVC terminal, and it works
[21:47:47 CET] <JEEB> vsvars32
[21:47:55 CET] <BtbN> I made myself a cygwin alias that does that
[21:47:57 CET] <JEEB> I have that in my msys.bat
[21:48:15 CET] <BtbN> So I can just type vc32 or vc64, and get my environment
[21:48:39 CET] <JEEB> anyways, both environments provide you with the necessary tooling
[21:49:09 CET] <ZeroWalker> hmm, so wait do i do that with msys2 as well, just launch it from msvc terminal?
[21:49:20 CET] <JEEB> run the msys_shell.bat from the VS2017 prompt
[21:49:22 CET] <Cracki> would swig-generated language bindings within the source tree be something the community accepts?
[21:49:28 CET] <ZeroWalker> kk
[21:49:29 CET] <JEEB> Cracki: no
[21:49:46 CET] <Cracki> I just want any kind of decent bindings for python
[21:49:47 CET] <JEEB> Cracki: generally I recommend you do it like the rust bindgen stuff does it
[21:49:59 CET] <JEEB> which generates bindings in the build system
[21:50:03 CET] <Cracki> preferably not rolling my own, duplication of effort and such
[21:50:05 CET] <JEEB> so it's always up-to-date to the version
[21:50:08 CET] <BtbN> ffmpeg won't accept anything that's not C. So any kind of bindings will always be 3rd party
[21:50:21 CET] <JEEB> also man, I remember using swig
[21:50:23 CET] <JEEB> it was awful
[21:50:32 CET] <JEEB> I just ended up writing the boilerplate myself
[21:50:35 CET] <Cracki> JEEB, don't take me literally. I don't care about the how, I care about having python bindings at all
[21:50:52 CET] <JEEB> I think vlc uses a custom python script for it
[21:50:59 CET] <JEEB> for libvlc, though
[21:51:04 CET] <JEEB> so you could take a look at that
[21:51:16 CET] <Cracki> so there is no feasible existing solution to bindings
[21:51:22 CET] <Cracki> I'd have to dissect what they did
[21:51:27 CET] <ZeroWalker> okay, did, and it opens the msys2 terminal, is that all, can i now do like ./configure and stuff on the ffmpeg source;P?
[21:51:44 CET] <JEEB> ZeroWalker: check that cl and link are the MSVC tools
[21:51:53 CET] <furq> Cracki: https://github.com/mikeboers/PyAV
[21:51:58 CET] <JEEB> with `which cl` and `which link`
[21:52:16 CET] <Cracki> furq, I assume it's maintained well?
[21:52:17 CET] <JEEB> ZeroWalker: if yes, then you will probably want to install git, yasm
[21:52:22 CET] <furq> no idea, i've never used it
[21:52:38 CET] <ZeroWalker> hmm, says "no cl"
[21:52:40 CET] <Cracki> that's bad. I have already found several efforts such as pyav
[21:52:53 CET] <ZeroWalker> and link says  "/usr/bin/link"
[21:52:55 CET] <Cracki> I'm really looking for endorsements from people who've used the libs
[21:53:18 CET] <JEEB> yes, many have done bindings but they generally tend to end up outdated which is why you need to find something that generates them on the run
[21:53:27 CET] <JEEB> otherwise they will always be out-of-date
[21:53:31 CET] <Cracki> _I am ok with generated bindings_
[21:53:44 CET] <JEEB> yes, just saying how the status is
[21:54:07 CET] <JEEB> I think your best bet would be to look at the vlc script
[21:54:12 CET] <Cracki> so how are bindings generated?
[21:54:15 CET] <Cracki> nvm
[21:54:15 CET] <JEEB> since that parses their headers
[21:54:35 CET] <Cracki> not sure I trust custom-rolled header parsing, but oh well
[21:54:50 CET] <JEEB> sure, if it uses libclang that'd be best
[21:54:56 CET] <JEEB> never looked at that script tbqh
[21:55:08 CET] <JEEB> just know that it's there
[21:55:22 CET] <JEEB> and it seemed like the least retarded way to do python mappings
[21:55:29 CET] <Cracki> ideally i'd like to see someone caring about maintaining the neccessary build structure to generate bindings in *some* way
[21:56:20 CET] <JEEB> well, if someone would make a generic enough thing then that would be a pip package, and then the build script for a package would require the headers and run that tool during the build phase
[21:56:22 CET] <BtbN> you seem to care, so go ahead
[21:56:32 CET] <JEEB> nothing that would IMHO go into FFmpeg main
[21:56:38 CET] <Cracki> I wish someone else cared before me.
[21:57:01 CET] <JEEB> of course :P
[21:57:07 CET] <JEEB> so do all of us
[21:57:15 CET] <JEEB> I was happy to learn of bindgen for rust
[21:57:20 CET] <JEEB> and even more happy when I got it to work
[21:57:30 CET] <Cracki> >_> if only I care, it'll exist in my little world of software, and nobody else benefits
[21:57:56 CET] <Cracki> or it'll go the way of all the other python bindings, more or less maintained
[21:58:00 CET] <JEEB> well, someone has to care and be the progenitor for everything. whether or not it gets used by others is a separate thing
[21:58:09 CET] <Cracki> ... or useless on anything other than linux
[21:58:13 CET] <BtbN> Can always just dlopen the library and dig your way through it that way
[21:58:21 CET] <Cracki> yes I can use ctypes, but I'd rather not
[21:58:30 CET] <Cracki> or use ctypes, but generate the boilerplate
[21:58:40 CET] <JEEB> yea, that's what the vlc script does
[21:58:45 CET] <ZeroWalker> okay seems to work, had to uncomment a line in the bat file. though it seems to point to the x86 cl weirdly enough
[21:58:54 CET] <JEEB> ZeroWalker: 'gratgs
[21:58:59 CET] <JEEB> *'grats
[21:59:02 CET] <ZeroWalker>  c/Program Files (x86)/Microsoft Visual Studio/2017/Community/VC/Tools/MSVC/14.10.25017/bin/HostX86/x86/cl
[21:59:15 CET] <JEEB> ZeroWalker: and for link you might have to remove/move the msys link tool
[21:59:37 CET] <ZeroWalker> ah
[22:00:15 CET] <Cracki> oh hell, that beast is looong http://git.videolan.org/?p=vlc/bindings/python.git;a=blob_plain;f=generated/vlc.py;hb=HEAD
[22:00:25 CET] <ZeroWalker> okay it points to msvc now for link as well:) (though x86)
[22:00:45 CET] <JEEB> ok, now just install git and yasm in msys2
[22:00:55 CET] <ZeroWalker> don't i want it to point to x64?
[22:00:57 CET] <JEEB> and git clone https://git.videolan.org/ffmpeg.git
[22:01:07 CET] <JEEB> ZeroWalker: you could do that as well, but you could just get rolling for now
[22:01:18 CET] <furq> --depth=1 for god's sake
[22:01:37 CET] <ZeroWalker> well i need it to be x64 for my usercase, but if it's not hard to change i guess i can just move on to get it working
[22:02:02 CET] <JEEB> yea, since the difference would just be to load msys2 from the other MSVS shell
[22:02:20 CET] <JEEB> and I think both 32bit and 64bit MSVC tools could build both?
[22:02:44 CET] <ZeroWalker> ah
[22:03:17 CET] <ZeroWalker> not that familiar with msvc terminal, though i have used it once and i can set it to one or the other, i just launched it so it might just be 32bit per default
[22:03:37 CET] <ZeroWalker> to install yasm, do i just add it to the bin folder (as yasm.exe)?
[22:03:47 CET] <JEEB> msys2 has a package manager
[22:03:51 CET] <JEEB> pacman
[22:03:57 CET] <JEEB> you can install git and yasm from it
[22:03:59 CET] <ZeroWalker> oh
[22:04:01 CET] <ZeroWalker> nice
[22:04:18 CET] <JEEB> http://www.msys2.org/7_msys2-install-freely.png
[22:04:18 CET] <ZeroWalker> pacman, the one you can always trust
[22:04:34 CET] <ZeroWalker> ah thanks
[22:04:54 CET] <JEEB> I've never liked the cryptic options for pacman (it comes from arch linux), but otherwise I've seen no issues with it
[22:05:31 CET] <ZeroWalker> do i want any of those options? (i see -S, which i assume might be Silent?)
[22:05:40 CET] <JEEB> -S is actually install
[22:05:44 CET] <JEEB> welcome to pacman
[22:05:46 CET] <ZeroWalker> that makes no sense
[22:05:52 CET] <ZeroWalker> i trusted  you pacman
[22:06:01 CET] <JEEB> now you know why I said "cryptic options"
[22:06:07 CET] <ZeroWalker> indeed xd
[22:06:23 CET] <JEEB> little y was something about updating the caches?
[22:06:24 CET] <JEEB> not sure
[22:06:37 CET] <JEEB> so you ended up with something like pacman -Syu package_name
[22:06:39 CET] <ZeroWalker> totally logical ofcourse
[22:07:08 CET] <JEEB> but if you looked at the arch wiki you would probably find a ton of explanations :P
[22:07:19 CET] <ZeroWalker> probably xd
[22:07:22 CET] <JEEB> so while you can't remember the darn things arch wiki would help as long as you have a browser
[22:07:33 CET] <JEEB> https://wiki.archlinux.org/index.php/pacman
[22:08:08 CET] <ZeroWalker> The goal of pacman is to make it possible to easily manage packages, however the options are cryptic and can only be understood by veterans.
[22:09:17 CET] <furq> christ don't say that pacman has cryptic options
[22:09:27 CET] <ZeroWalker> hmm is ffmpeg git down?
[22:09:31 CET] <furq> i can hear them descending on the channel already
[22:09:43 CET] <ZeroWalker> well cryptic isn't neccesarily bad
[22:09:47 CET] <JEEB> nope
[22:09:50 CET] <ZeroWalker> hmm
[22:09:51 CET] <JEEB> https://git.videolan.org/git/ffmpeg.git seems to be online
[22:09:58 CET] <JEEB> at least the web interface for it
[22:10:02 CET] <JEEB> http://git.videolan.org/?p=ffmpeg.git;a=summary
[22:10:03 CET] <furq> ready to tell us "ah, but did you know, you can remove packages with apt-get install -packagename, THEREFORE ARCH IS BETTER"
[22:10:18 CET] <furq> i don't know why i'm now doing their job for them. i'll stop
[22:10:32 CET] <ZeroWalker> hmm
[22:10:36 CET] <haroldp> I have an ffmpeg command-line that is converting an rtsp stream from a crappy netcam to rtmp stream.  I can end the output to a file and get a valid mp4 file.  Now I'm having trouble getting nginx to serve it.  I set up nginx with the rtmp module, configured server and app (probably wrong, heh), and then try to hit the stream with VLC.  no errors on client or server.  no stream.
[22:10:54 CET] <ZeroWalker> okay this one seems to work
[22:11:04 CET] <ZeroWalker> hmm, maybe i copied the old link wrong or something
[22:11:11 CET] <JEEB> or I just wrote it wrong
[22:11:14 CET] <JEEB> I didn't copypasta it at first
[22:11:20 CET] <furq> haroldp: try opening it with ffmpeg/ffplay
[22:11:23 CET] <furq> that should at least give some errors
[22:11:38 CET] <furq> ffmpeg -v debug -i rtmp://... -f null -
[22:11:44 CET] <haroldp> ok.  I'll try to install ffplay somewhere
[22:11:46 CET] <ZeroWalker> the i put all blame on you, you should be ashamed!
[22:11:53 CET] <ZeroWalker> or unshamed?
[22:12:10 CET] <Cracki> well, thanks. I'll figure this out from here
[22:12:53 CET] <JEEB> huh
[22:12:54 CET] <JEEB> warning: terminate MSYS2 without returning to shell and check for updates again
[22:12:55 CET] <JEEB> warning: for example close your terminal window instead of calling exit
[22:13:09 CET] <JEEB> msys2 sounds like fun with how it tries to do cli package management
[22:13:10 CET] <JEEB> I guess
[22:14:35 CET] <JEEB> but at least it has newer tooling than my olden msys2 thing
[22:15:09 CET] <BtbN> package management from within the environment just doesn't work on Windows
[22:15:17 CET] <BtbN> Because of unavoidable file locking
[22:15:20 CET] <JEEB> yes
[22:15:47 CET] <JEEB> also I think you could avoid it, but only by windows APIs instead of the "standard" ones. plus I think binaries would get locked anyways
[22:16:29 CET] <obamoose> Hello
[22:16:42 CET] <obamoose> Can someone help me figure out how to turn my stream into an eternal stream
[22:17:21 CET] <obamoose> that just keeps on playing the files I add on the end
[22:17:27 CET] <obamoose> This is my current config:
[22:17:33 CET] <obamoose> http://pastebin.com/1PP7S6ig
[22:22:49 CET] <xtina> Hello! I'm trying to understand if I can increase my ALSA buffer size over 65536; I'm unable to do so. does anyone know if this is a hardware constraint?
[22:32:19 CET] <alexpigment> xtina: i don't have a clue, but 65536 = 64 x 1024
[22:32:58 CET] <xtina> @alexpigment: yes, which makes me suspect that the ALSA buffer size does have an upper limit
[22:33:02 CET] <xtina> i've never seen a value over 65536
[22:34:29 CET] <alexpigment> yeah, i don't know enough about ALSA (been a while since I used linux regularly)
[22:34:58 CET] <haroldp> furq: the ffmpeg command you suggested just hangs after, "Sending play command for 'cam1'".  Nothing after that untill I ^c
[22:37:16 CET] <ZeroWalker> ./configure: line 1420: cmp: command not found
[22:37:41 CET] <ZeroWalker> i get this, i installed pkg-config as well as it seemed to need it. but not sure what this error is about
[22:38:17 CET] <JEEB> diffutils
[22:38:26 CET] <alexpigment> xtina: google isn't retuning anything useful when searching for 64mb alsa buffer limit
[22:38:38 CET] <alexpigment> is it possible your audio card is the limiting factor?
[22:38:45 CET] <JEEB> ZeroWalker: diffutils is the package name that provides cmp
[22:38:46 CET] <JEEB> just fyi
[22:39:00 CET] <JEEB> cmp
[22:39:03 CET] <ZeroWalker> ah, thanks, what would i ever do without you
[22:39:36 CET] <JEEB> you would have started typing "msys2 cmp" and gotten "msys2 cmp not found"
[22:40:32 CET] <ZeroWalker> oh
[22:40:49 CET] <ZeroWalker> but i want it to be melodramatic;(
[22:43:40 CET] <xtina> @alexpigments: yea, google doesn't tell me anything. it's possible
[22:47:46 CET] <alexpigment> xtina: did you see this page? https://bbs.archlinux.org/viewtopic.php?id=171477
[22:47:55 CET] <alexpigment> perhaps there's a hint somewhere in there
[22:48:43 CET] <alexpigment> http://stackoverflow.com/questions/43009246/does-alsa-buffer-size-have-a-hardware-defined-upper-limit
[22:48:44 CET] <alexpigment> also that
[22:52:19 CET] <ZeroWalker> hmm, i looks like the configure succeeds, but i can't "make", so i guess it doesn't?
[22:52:59 CET] <JEEB> ah, fun. maybe you're missing good ol' make?
[22:53:13 CET] <JEEB> yes, funny enough that's not installed by default
[22:53:19 CET] <JEEB> package name is make
[22:58:12 CET] <alexpigment> this is a pretty exhaustive compilation guide by JEEB here. i'm going to have to save this chat log for future reference :)
[22:58:23 CET] <alexpigment> (appreciate the info as a spectator)
[22:58:30 CET] <ZeroWalker> oh, yeah i thought it would be
[22:59:05 CET] <furq> this is why i just cross-compile
[22:59:10 CET] <alexpigment> same
[22:59:22 CET] <ZeroWalker> instead of make?
[22:59:54 CET] <haroldp> furq: got any other ideas for debugging this? :)
[23:00:16 CET] <furq> pastebin your ffmpeg command (the one streaming to nginx) and maybe your nginx config
[23:00:41 CET] <JEEB> furq: tbqh this ain't too different from cross-compilation on *nix
[23:00:51 CET] <furq> msys is just generally a bit of a pain
[23:01:06 CET] <JEEB> if your distro doesn't install make, diffutils, yasm, git by default you install the same packages
[23:01:26 CET] <JEEB> the only extra requirement is to have the msys bat run from MSVS's thing
[23:01:27 CET] <JEEB> vOv
[23:03:55 CET] <faLUCE> can ffmpeg/libav compiled WITHOUT aac support?
[23:04:08 CET] <furq> --disable-encoder=aac --disable-decoder=aac
[23:04:29 CET] <faLUCE> thnks furq
[23:04:58 CET] <JEEB> furq: I do agree that *nix is faster with fork() n' all tho :D
[23:05:10 CET] <JEEB> anyways, I guess this is weekend so time to install VS2017 on this box finally
[23:05:27 CET] <haroldp> furq: here's my nginx conf.  The ffmpeg comamand is exec-ed from there: http://pastebin.com/PKXDCUvj
[23:06:11 CET] <furq> does anything show up in error.log
[23:06:18 CET] <haroldp> nada
[23:07:20 CET] <haroldp> I'm pretty sure it's something wrong with my conf.  First time setting something like this up
[23:07:50 CET] <furq> you might need "allow play all"
[23:07:56 CET] <furq> in rtmp{}
[23:07:59 CET] <haroldp> within what black?
[23:08:02 CET] <haroldp> thanks
[23:08:17 CET] <furq> the docs don't mention the default, but i have that so i assume it's there for a reason
[23:09:39 CET] <haroldp> ok, added, restarted, still stalling at ' Sending play command for 'cam1''
[23:10:10 CET] <haroldp> my command is like, ffmpeg -v debug -i  rtmp://10.0.0.42/live360p/cam1
[23:14:28 CET] <furq> actually i'm not sure if that's how exec works
[23:15:01 CET] <haroldp> ok.  Lifted it from various examples I found
[23:15:02 CET] <furq> you might want to use exec_static
[23:15:10 CET] <haroldp> I'll try that
[23:15:17 CET] <furq> outside of the application block
[23:18:19 CET] <haroldp> ooh, actually getting an error message in teh log now
[23:18:28 CET] <haroldp> 2017/03/24 15:17:49 [alert] 15896#103419: kevent() error on 9 filter:-1 flags:4000 (9: Bad file descriptor)
[23:19:44 CET] <haroldp> getting that like every 5 seconds, heh
[23:21:33 CET] <ZeroWalker> wasn't --enable-shared the command for making the dlls ?
[23:21:43 CET] <JEEB> yes
[23:21:48 CET] <JEEB> during configure
[23:21:58 CET] <JEEB> also a little hint, use out-of-tree building
[23:22:00 CET] <ZeroWalker> yeah did that, but ended up with ffmpeg.exe:(
[23:22:08 CET] <ZeroWalker> what's that?
[23:22:09 CET] <JEEB> yes, you always also get the exe
[23:22:14 CET] <JEEB> unless you disable the binaries
[23:22:40 CET] <JEEB> out-of-tree just means that you call the configure outside of the source tree
[23:23:12 CET] <JEEB> so if you have ffmpeg sources under a directory called "ffmpeg", you can a directory called "build" next to it and then run "../ffmpeg/configure"... and that works :)
[23:23:20 CET] <ZeroWalker> not really sure what that means. i just call ./configure inside the git clone map
[23:23:21 CET] <ZeroWalker> Oh
[23:23:36 CET] <JEEB> that way to reset the situation you can just remove that directory and re-create it
[23:23:59 CET] <JEEB> that said, if you already once configured within you'll have to run `git clean -dfx`
[23:24:01 CET] <ZeroWalker> that sounds neat, as now it just compiles it in the middle of everything
[23:24:04 CET] <JEEB> to clean the ffmpeg source directory :)
[23:24:09 CET] <ZeroWalker> ah:d
[23:24:42 CET] <JEEB> also, to utilize the libraries you made, you will also have to install
[23:24:54 CET] <JEEB> for that I recommend utilizing the --prefix option
[23:25:16 CET] <JEEB> in msys2 you can probably just do --prefix=/home/username/ownapps/ffmpeg or so
[23:25:30 CET] <JEEB> and then after you have built it with make and it succeeded, you can run `make install`
[23:25:38 CET] <JEEB> that way you get the correct directory structure
[23:25:50 CET] <JEEB> you will have lib/ bin/ include/ etc
[23:25:53 CET] <JEEB> within  your prefix
[23:27:27 CET] <ZeroWalker> wait, i want to install those into msys2? i thought you only did that if it's used withing msys or the compiler or something
[23:27:45 CET] <JEEB> the install is needed for re-usage of the libraries and headers
[23:27:48 CET] <JEEB> in general
[23:27:53 CET] <JEEB> that's why I noted --prefix
[23:28:00 CET] <JEEB> so you have your little directory with just ffmpeg
[23:28:04 CET] <alexpigment> haroldp: i'm about to head out and it's not clear if furq is still around or not, but i found some people who seem to be running into similar issues. perhaps it's worth a look if furq doesn't come back
[23:28:07 CET] <alexpigment> https://www.google.com/search?q=kevent+()+error+on+9+filter%3A+-1+flags%3A4000+(9%3A+Bad+file+descriptor)&oq=kevent+()+error+on+9+filter%3A+-1+flags%3A4000+(9%3A+Bad+file+descriptor)&aqs=chrome..69i57.20303j0j7&sourceid=chrome&ie=UTF-8#q=ffmpeg+error+(9:+Bad+file+descriptor)+kevent+error+on+9+filter:-1&*
[23:28:28 CET] <haroldp> yeah, I was looking at the same search
[23:28:32 CET] <alexpigment> the 4th result seems to be particularly relevant
[23:28:36 CET] <JEEB> then in that other thing you'd be using FFmpeg's libraries with you can specify <that prefix directory>/lib for libraries and /include for header search paths
[23:28:40 CET] <JEEB> if you get what I say :P
[23:29:13 CET] <haroldp> one guy is on FreeBSD same as me, but suggests redirecting stdin and stdout to /dev/null but that didn't help me at all. Still getting the error.
[23:29:35 CET] <alexpigment> yeah, that's the one i was looking at
[23:29:41 CET] <alexpigment> oh well, figured i'd throw that out there just in case
[23:29:43 CET] <alexpigment> good luck!
[23:29:52 CET] <haroldp> thanks anyway
[23:30:46 CET] <haroldp> was trying to copy something like this: https://www.vultr.com/docs/setup-nginx-rtmp-on-ubuntu-14-04
[23:31:27 CET] <furq> haroldp: change ffmpeg to /usr/local/bin/ffmpeg
[23:33:47 CET] <ZeroWalker> hmm
[23:34:01 CET] <oftc_ftw> hi all. i want to convert some old family videos that already were converted from VHS to DVD. Now I have a bunch of files that I want to upload to my cloud storage and ideally want to be able to play it in my browser (HTML5 player). I guess this leaves webm/VP9 or mp4/H.264 as best possibilities, right? but i also want to make sure i don't have information loss, i.e. artifacts new that were not already there. are these codecs capable
[23:34:01 CET] <oftc_ftw> of such a "lossless" conversion?
[23:34:26 CET] <furq> both of them do lossless but i don't think browsers will play that back
[23:34:26 CET] <oftc_ftw> *new artifacts
[23:34:37 CET] <oftc_ftw> oh :(
[23:35:04 CET] <oftc_ftw> well, i could at least try i guess? is there a special command line parameter for lossless in ffmpeg?
[23:35:16 CET] <furq> -c:v libx264 -qp 0
[23:35:27 CET] <furq> i just tested and that definitely doesn't work in browsers
[23:35:35 CET] <oftc_ftw> that was quick :D
[23:35:41 CET] <oftc_ftw> ok, i'
[23:35:44 CET] <furq> or at least not in mine
[23:35:53 CET] <furq> maybe it'll work in something which uses lavc for decoding
[23:35:56 CET] <oftc_ftw> ok, i'd rather have it lossless and not play it in a browser...
[23:36:13 CET] <furq> don't convert it at all then
[23:36:17 CET] <klaxa> ^
[23:36:18 CET] <haroldp> I swear, if it's a dumb PATH fail...  :)
[23:36:30 CET] <furq> lossless h264 will be much bigger than a dvd source
[23:36:38 CET] <oftc_ftw> right, i will try if it works and if not upload the originals. thanks guys!
[23:37:12 CET] <furq> what cloud storage is this
[23:37:15 CET] <oftc_ftw> furq: btw, are you implying h264 is better than vp9?
[23:37:25 CET] <oftc_ftw> "better" in any way
[23:37:33 CET] <furq> x264 is better than libvpx in some ways
[23:37:39 CET] <oftc_ftw> ok, thanks :)
[23:37:41 CET] <furq> mostly speed and compatibility
[23:38:10 CET] <furq> also i can't remember how to do lossless vp9 off the top of my head
[23:39:12 CET] <furq> but yeah depending on the cloud storage it might generate browser-playable videos for you from a dvd source while keeping the originals
[23:39:23 CET] <furq> i believe google drive can do that
[23:40:45 CET] <oftc_ftw> nextcloud can do it as well, i have a private server
[23:40:56 CET] <haroldp> furq: holy shit.  this is kinda working.
[23:40:58 CET] <oftc_ftw> thanks again, i will simply try it with your parameters
[23:41:07 CET] <furq> well yeah those lossless files will be huge
[23:41:19 CET] <oftc_ftw> :)
[23:41:21 CET] <furq> probably 3-5x bigger than the source
[23:41:52 CET] <furq> it's probably less hassle to back up the dvds and then make lossy copies for browser viewing
[23:42:04 CET] <haroldp> anyway, VLC got at leats one frame out of it, heh
[23:44:47 CET] <xtina> hey guys, i know this isn't really the right chat for this, but is anyone here familiar with mmap?
[23:44:58 CET] <xtina> i'm looking for a way to read/write from memory with a large buffer (say, 50MB)
[23:45:11 CET] <xtina> i'm currently using a FIFO pipe but that has an upper buffer limit of 1MB. i'm hitting htat limit.
[23:45:46 CET] <xtina> i can't read/write from disk quickly enough for my needs (realtime streaming with ffmpeg)
[23:50:50 CET] <kerio> how's that gonna help you
[23:57:01 CET] <xtina> @kerio: my 1mb buffer is currently filling up if i have low connectivity for a bit of time
[23:57:11 CET] <xtina> i don't use more than 100mb of ram for streaming and i have 512
[23:57:28 CET] <xtina> so i'd like to have a bigger leeway to store audio packets if ffmpeg isn't reading them fast enough
[23:57:52 CET] <xtina> i know it's not a 100% solution but if i could use a 100mb buffer, i could get 100x more leeway which is good enough
[23:57:55 CET] <kerio> can you use mbuffer
[23:57:56 CET] <kerio> ?
[23:58:17 CET] <xtina> hmm, let me look into it
[00:00:00 CET] --- Sat Mar 25 2017


More information about the Ffmpeg-devel-irc mailing list