[Ffmpeg-devel-irc] ffmpeg.log.20180916

[burek]

[00:00:05 CEST] <ahoo> which lossy audio codec sounds best?
[00:00:30 CEST] <ahoo> is mp3 obsolete?
[00:00:45 CEST] <c_14> that depends entirely on how much bitrate you throw at it
[00:00:58 CEST] <ahoo> if i would use mp3, 320.
[00:01:12 CEST] <nicolas17> hm
[00:01:14 CEST] <c_14> they'll all sound as good as you want them, but the newer codecs sound better at lower bitrates than the older ones at comprable bitrates
[00:02:30 CEST] <ahoo> i'm converting flac to a lossy format for mobile listening while at the same time i demand near-losslessly audio
[00:03:02 CEST] <ahoo> so& how about aac?
[00:03:16 CEST] <c_14> It's fine
[00:03:23 CEST] <c_14> vbr around 256k and you shouldn't notice
[00:03:26 CEST] <c_14> you can probably even go lower
[00:04:30 CEST] <ahoo> thx
[00:04:36 CEST] <furq> apple aac at 128k is pretty much transparent
[00:04:43 CEST] <furq> or opus if you're on android
[00:05:16 CEST] <c_14> ^depending on the device and android version it may only support opus in .mkv though
[00:05:33 CEST] <furq> how long ago are we talking here
[00:05:42 CEST] <c_14> 2 years?
[00:05:45 CEST] <furq> oh wow
[00:05:54 CEST] <furq> presumably thirdparty players will support ogg anyway
[00:06:00 CEST] <c_14> yeah
[00:06:06 CEST] <c_14> It's just the builtin stuff that won't
[00:06:10 CEST] <furq> on a related note, do any android devices do hwdec for aac
[00:06:25 CEST] <furq> i've seen people claiming opus is "bad for battery life"
[00:06:49 CEST] <furq> but i'm pretty sure that's not true on android
[00:07:05 CEST] <ahoo> i use a first-class player.
[00:07:16 CEST] <ahoo> it plays everything reasonable.
[00:08:45 CEST] <furq> if you mean a dedicated pmp then just use opus i guess
[00:09:12 CEST] <johnnny22> whats more efficient, a wrapped frame or a rawvideo frame for the output ? Is there any difference ?
[00:09:15 CEST] <furq> apple aac is competitive with opus but there's no reason to use it unless you have an iOS device
[00:10:08 CEST] <c_14> johnnny22: wrapped_frames have a memcpy less I believe
[00:11:44 CEST] <c_14> I don't think there's anything that can consume them once they've been output though
[00:12:31 CEST] <johnnny22> c_14: i'm sending to a device so it should be fine :)
[00:22:25 CEST] <johnnny22> though, this device actually also does a av_packet_clone internally. Not sure of the impact of that either.
[01:33:46 CEST] <johnnny22> c_14: using the wrapped frame seems to reduce it's CPU usage by almost 1/5th (~25% to ~20%)
[01:41:00 CEST] <c_14> nice
[01:41:15 CEST] <illuminated_> https://paste.fedoraproject.org/paste/WUYIMFODTYqLnmr4bfnywA <---that's what outputs when I put in the command to copy the streams and set the default audio stream to the second audio stream... but when I play it back in any player, it still defaults to the 1st audio stream.  Anybody know why?
[01:43:53 CEST] <johnnny22> c_14: still not sure 'seems' :)  think i did a little mistake, so i'm not sure yet.
[01:43:54 CEST] <c_14> what's the command you're using?
[01:44:01 CEST] <c_14> ^ illuminated_
[01:44:46 CEST] <c_14> also check ffprobe to make sure the second stream still has the default disposition in that
[01:45:56 CEST] <illuminated_> the command is generated by a script i'm writing so the fields are auto-populated by variables.  let me do it manually and see if that sorts it.  also, i've never used ffprobe.  how do I find out that information with ffprobe?
[01:46:09 CEST] <c_14> just ffprobe file
[01:46:23 CEST] <c_14> it'll list the stseams a lot like in that output you posted
[01:46:43 CEST] <c_14> (there's more detailed options, but just running it should work just fine)
[01:48:48 CEST] <illuminated_> c_14: ffprobe output https://paste.fedoraproject.org/paste/O0RXws1iGqyymGEKjw3R-Q
[01:49:12 CEST] <illuminated_> doesn't say nothing about default anywhere
[01:51:16 CEST] <c_14> I don't think avi has default streams
[01:51:36 CEST] <c_14> what you can do is change the order of the streams so the "second" one is first
[01:51:40 CEST] <c_14> and then that should be chosen by default
[01:51:51 CEST] <c_14> changing the order of your map statements should do it
[01:52:10 CEST] <illuminated_> ok cool
[01:52:20 CEST] <illuminated_> thanks
[01:53:28 CEST] <illuminated_> now I gotta figure out how to do that in a shell script
[01:54:23 CEST] <johnnny22> maybe be lucky and run 'ps aux' and get the command from there ;)
[01:59:39 CEST] <furq> illuminated_: do yourself a favour and put it in something better than avi
[02:19:32 CEST] <johnnny22> c_14: yup, (rawvideo (native) -> wrapped_avframe (native)) seems 1/5 better for the CPU than (rawvideo (native) -> rawvideo (native))
[02:19:47 CEST] <johnnny22> at least, in this context that I have.
[02:22:01 CEST] <johnnny22> I hope this RGB0 output will do, because, if I have to change it to RGBA, i loose all this performance boost.
[02:36:29 CEST] <johnnny22> oh it comes out green .. i did something wrong :P
[02:37:28 CEST] <c_14> should it be black?
[02:39:30 CEST] <c_14> because that sounds a lot like an incorrect yuv->rgb conversion
[02:39:41 CEST] <c_14> (0,0,0) is black in RGB, but green in YUV
[02:40:10 CEST] <johnnny22> yeah, supposed to be black, but i'm guessing it's actually xrgb ordered bits
[02:40:29 CEST] <johnnny22> it's a 'dark green' that's coming out.
[02:41:43 CEST] <johnnny22> think they got their API all screwed up on the decklinks :P or things not very neatly named.
[02:42:52 CEST] <johnnny22> bmdFormat8BitARGB  is BGRA bitwise, while their bmdFormat8BitBGRA  is XRGB bitwise
[02:44:41 CEST] <johnnny22> the source is x11grab, so i think i'm limitted to BGR0 (i might be wrong though)
[02:46:10 CEST] <johnnny22> trying to get things go unchanged :P
[02:46:42 CEST] <c_14> x11grab should be bgra, yeah
[02:52:42 CEST] <johnnny22> my x11grab is coming out as bgr0
[02:53:52 CEST] <johnnny22> cool filter though :P lol
[03:00:55 CEST] <johnnny22> probably got myself all screwed into the endian'ness .. i'll get to the bottom ;)
[04:16:53 CEST] <johnnny22> got them colors fixed, i missed a part that returned the frames pixel format.
[04:17:56 CEST] <johnnny22> it's running at 10% of a core
[04:19:27 CEST] <johnnny22> and their BGRA fits with av's BGR0
[05:13:36 CEST] <johnnny22> codecpar is a short term for what exactly ?
[06:15:24 CEST] <yagiza> johnnny22, codec parameters?
[06:35:36 CEST] <johnnny22> aaah quite possible
[07:59:27 CEST] <illuminated_> hmmm... having issues with subtitle disposition now
[08:00:06 CEST] <illuminated_> mapped the subtitle track and trying to set the disposition, but ffprobe and players are not recognizing the new default subtitle
[08:09:05 CEST] <illuminated_> hmm.. I was able to get it to work, but I don't know why it worked.  But how I got it to work is not how I want it to work in my script
[08:12:57 CEST] <illuminated_> ahh.. now I get it.
[08:56:14 CEST] <illuminated_> programming is hard
[08:59:27 CEST] <illuminated_> trying to provide a shell script interface to ffmpeg...
[08:59:39 CEST] <illuminated_> for something specific
[08:59:45 CEST] <mixfix41> why not broad
[08:59:52 CEST] <illuminated_> broad?
[09:00:04 CEST] <mixfix41> well bash scripting can apply to a lot of inputs
[09:00:12 CEST] <illuminated_> yeah
[09:00:24 CEST] <illuminated_> well what i'm attempting to do is manipulate streams and stream defaults
[09:01:09 CEST] <mixfix41> so youre encoding files or something with stream
[09:01:16 CEST] <illuminated_> kinda..
[09:03:19 CEST] <illuminated_> basically making it easy to choose which audio streams you want and which subtitle streams you want... where you can put in which streams you want in a a file, and then tell it which default stream you want, and then regenerate the file with those details.
[09:05:32 CEST] <illuminated_> but.. I want to create a file list as well.. say you download a season of files and they all have the same streams in them.. you'll be able to quickly say I want these audio streams, these subtitle streams, and I want this audio stream, and this subtitle stream as default, and automatically regenerate all the files
[09:06:38 CEST] <illuminated_> i wrote a script to do this for me with 6 seasons of green acres.  It was a russian torrent and had russian/english audio tracks.
[09:07:15 CEST] <illuminated_> but i thought...if I generalize it it would be even cooler...
[09:09:35 CEST] <illuminated_> after I wrote the script for green acres, I just sat back and let it re-generate 6 seasons of files with only the english audio track for me.
[09:10:12 CEST] <illuminated_> just issued 1 command and it re-created 200 files for me lol.
[09:12:58 CEST] <illuminated_> but I'm not the worlds best shell scripter
[09:13:06 CEST] <illuminated_> i've not really shell scripted
[09:13:12 CEST] <illuminated_> cuz I couldn't think of anything to script
[09:13:33 CEST] <mixfix41> i think as daily everybody uses shell when theyre using a shell
[09:13:34 CEST] <illuminated_> but this seemed like an interesting idea..so off I go
[09:13:54 CEST] <mixfix41> therefore progressing but yeah *progressing* varies
[09:14:04 CEST] <illuminated_> yeah
[09:15:13 CEST] <illuminated_> audio seems pretty straight forward but the subtitles are different
[09:15:29 CEST] <mixfix41> ofcourse but thats where the fun is at
[09:15:52 CEST] <mixfix41> theres actually a prety good one subtitle editor or something its a learning curve but take notes its worth it
[09:17:16 CEST] <illuminated_> well, the thing is... audio stream numbers are related to the inputs audio stream numbers for default stream numbers.  evidently, it appears, subtitle stream default numbers are related to the output audio stream numbers.
[09:18:58 CEST] <mixfix41> are you indicating thats why subtitles arent universal or something im not sure
[09:20:03 CEST] <mixfix41> actuailly subtitles are just text files
[09:20:24 CEST] <mixfix41> so with the numbers is just time numbers in the textfile
[09:21:39 CEST] <illuminated_> no... i'm just saying that the way you get stream numbers is different between audio and subtitle
[09:21:55 CEST] <illuminated_> for the purposes of setting default streams
[09:22:17 CEST] <mixfix41> or yeah you wanting to embed the subtitles or something when the film is in russian
[09:22:26 CEST] <illuminated_> no
[09:22:32 CEST] <illuminated_> I don't really want to explain it
[09:22:40 CEST] <illuminated_> too difficult to explain
[09:29:32 CEST] <fromBeyond> Hi, I am getting the error "Unknown input format: 'dshow'" on ubuntu 18. I am trying to record my webcam. I have installed from the PPA
[09:29:55 CEST] <fromBeyond>  ffmpeg -list_devices true -f dshow -i dummy
[09:30:03 CEST] <fromBeyond> this command gives me the error
[09:30:27 CEST] <fromBeyond> i am googling the error, but i dont get any fruitfull results
[09:30:40 CEST] <fromBeyond> ffmpeg version 3.4.4-0ubuntu0.18.04.1
[09:30:45 CEST] <fromBeyond> is the version
[09:30:48 CEST] <tdr> weird, that command gives me:  Unrecognized option 'list_devices'.
[09:31:15 CEST] <fromBeyond> do you perhaps know how i can show my webcam?
[09:31:22 CEST] <fromBeyond> what command can i try
[09:32:07 CEST] <fromBeyond> ffmpeg -y -f vfwcap -i list
[09:32:12 CEST] <fromBeyond> this also gives me an error
[09:32:27 CEST] <fromBeyond> Unknown input format: 'vfwcap'
[09:35:36 CEST] <tdr> ffmpeg -devices  maybe?
[09:37:43 CEST] <fromBeyond> https://pastebin.com/pUQ5TTFi
[09:37:47 CEST] <fromBeyond> this is the output
[09:39:24 CEST] <tdr>  D  video4linux2,v4l2 Video4Linux2 device grab
[09:40:22 CEST] <fromBeyond> tdr: yes, thank you
[09:40:29 CEST] <fromBeyond> I am a novice to this kind of use
[09:41:16 CEST] <tdr> try ffmpeg -f v4l2 -list_formats all -i /dev/video0
[09:41:30 CEST] <tdr> your camera should be video0 unless you have more than one
[09:42:30 CEST] <fromBeyond> i did, and it worked. This is what I meant with above. I am now wondering, how to inlcude audio. I see pulse video output, but i think i need some input. openal audio capture device perhaps
[09:46:19 CEST] <fromBeyond> i did it in a different way. Thank you for your help, I have enough to start experimenting and learning
[09:46:28 CEST] <tdr> np, good luck
[10:26:40 CEST] <mixfix41> ~ls
[10:29:18 CEST] <mixfix41> oh yea illuminated_ you might have thought of dialog utility
[10:31:07 CEST] <illuminated_> mixfix41:  ? not sure what you're talking about
[10:31:53 CEST] <mixfix41> oh nvm
[11:44:10 CEST] <GuiToris> Hey, I dont fully understand these presets. Do the affect onlybthe file size or the quality too? If i wish to edit the video files later shoul i use ultrafast or veryslow?
[12:11:07 CEST] <yagiza> Hello!
[12:11:36 CEST] <ahoo> HELLO!!!!!!
[12:12:40 CEST] <yagiza> I finally implemented RTP communication in my app using FFMpeg libraries.
[12:13:35 CEST] <ahoo> conspirations
[12:15:02 CEST] <yagiza> The next step is SRTP. How can I get master key and master seed to specify them in srtp_in_params/srtp_out_params?
[12:15:16 CEST] <yagiza> ahoo, what do you mean?
[12:15:39 CEST] <ahoo> nothing behind the unturned stone
[14:32:40 CEST] <terminalator> How does one take multiple screenshots from a movie file?
[14:32:50 CEST] <fling> Which filter to use for black frame subtraction?
[14:36:34 CEST] <fling> terminalator: hit S in mpv
[14:37:31 CEST] <terminalator> fling: Thanks, but I mean from the command line with ffmpeg let's say for example 10 shots from one file
[14:37:56 CEST] <BtbN> convert to an image sequence and set a stupidly low fps that ends up with however many frames you want
[14:37:57 CEST] <terminalator> I tried a command, but it isn't working at all
[14:38:05 CEST] <terminalator> I see
[14:38:12 CEST] <BtbN> You can also use the thumbnail filter which has some logic to find "meaningful" frames
[14:38:34 CEST] <terminalator> BtbN: From the command line?
[14:47:00 CEST] <fling> terminalator: this works for me -> for m in {1..10} ; do ffmpeg -ss $m:00 -i input.mp4 shot-$m.jpg ; done
[14:47:23 CEST] <fling> terminalator: ten shots, one each minute
[14:53:43 CEST] <terminalator> fling: Thanks, I'm trying that command out. I've bit trouble getting "ffmpeg -ss $m:00" part working. Could you give another example, please?
[14:53:44 CEST] <fling> terminalator: now you help me to do black frame subtraction!
[14:54:00 CEST] <fling> terminalator: hat is the issue with this part?
[14:54:17 CEST] <fling> s/hat/what/
[14:55:21 CEST] <fling> -ss skip seconds from the input in this case
[14:56:23 CEST] <terminalator> Okay, give a minute
[15:01:38 CEST] <terminalator> fling: It keeps saying "Invalid duration specification for ss"
[15:05:41 CEST] <akaWolf> https://github.com/akaWolf/FFmpeg/commit/52673d0760a4ad751c2329a30b4e9f565eaa8a67 -- does that looks okay for commit in master?
[15:07:37 CEST] <JEEB> akaWolf: post it on the -devel ML? although I thought max attempts were configured in some AVOption?
[15:07:48 CEST] <JEEB> so hard-coding 5 doesn't sound like a good idea (TM)
[15:09:06 CEST] <terminalator> fling: Thank you so much Sir! Managed to get it working. I'm a happy camper again :)
[15:21:37 CEST] <akaWolf> JEEB: there are hardcoded auth attempts also, so I decide that connect attempts also can be hardcoded..
[15:29:23 CEST] <fling> now you help me getting dark frame subtraction
[15:29:28 CEST] <fling> I can't figure out the command
[15:32:56 CEST] <fling> terminalator: what was the issue?
[15:33:52 CEST] <terminalator> fling: Simply, misplaced a semicolon ;)
[15:35:18 CEST] <terminalator> Did you manage to solve the black frame thingy?
[16:25:42 CEST] <fling> terminalator: no, have no idea how to do this
[16:26:28 CEST] <fling> terminalator: I kindof need to subtract two videos somehow probably
[16:26:43 CEST] <fling> like each frame sample?
[16:28:57 CEST] <terminalator> What are you're trying to achieve exactly?
[16:28:59 CEST] <terminalator> Hmmm
[16:31:58 CEST] <fling> terminalator: to remove hot pixels and other static sensor noise from a video
[16:50:52 CEST] <kepstin> fling: should be fairly straightforwards. give ffmpeg the black frame and video as inputs, and write a -filter_complex string that loops the black frame then combines then using the 'blend' filter (supports a bunch of blend modes similar to what e.g. photoshop has)
[16:51:57 CEST] <kepstin> actually, blend supports framesync options, you don't even need to loop it
[16:53:56 CEST] <kepstin> fling: could be as simple as ffmpeg -i video.mp4 -i blackframe.jpg -filter_complex '[0:v][1:v]blend=all_mode=subtract' output.mp4
[16:54:08 CEST] <kepstin> (adding additional audio/video encoding options as appropriate)
[18:16:05 CEST] <fling> kepstin: thanks! Will try this later.
[18:18:01 CEST] <fling> kepstin: How do I construct the dark frame properly? I think I should capture like a one minute video, then blend all the frames together to get rid of the random noise right?
[18:18:56 CEST] <kepstin> i'm not the expert on that. Probably look into some photography forums - maybe astronomy in particular? to see what they recommend
[18:20:17 CEST] <kepstin> also it probably doesn't make sense to try to do this if your camera has any sort of automatic gain control :/
[18:26:52 CEST] <ahoo> do camcorders usually allow capturing from it via dshow?
[18:51:56 CEST] <m_saba01> Hey guys, is it possible to improve the perceived audio quality of highly compressed speech?
[18:52:23 CEST] <m_saba01> Some 64kbps audio books feature quite audible compression artifacts. I know that it's impossible to restore data that has been discarded during the compression.
[18:52:38 CEST] <m_saba01> What I'm looking for is a way to conceal the artifacts/distortion that compression generally creates. Someting analogous to slightly blurring a JPEG in order to remedy these 8x8 blocks.
[18:59:52 CEST] <ChocolateArmpits> m_saba01, well there are filters for video compression noise removal out there, I'm guessing there must be something similar for audio
[19:14:09 CEST] <ChocolateArmpits> m_saba01, can only find delossifier http://www.help.stereotool.com/7.42/delossifier.shtml
[19:22:28 CEST] <durandal_1707> that is full of shit
[19:22:37 CEST] <durandal_1707> you cant improve lossy audio
[22:09:45 CEST] <GuiToris> hey, do presets affect only the file size or the quality of the images too? If I'd like to edit the videos later should I use ultrafast or veryslow?
[22:10:13 CEST] <ChocolateArmpits> GuiToris, for editing you should be mostly concerned with GOP size
[22:10:31 CEST] <BtbN> If you want to edit it later you should use a lossless intermediate codec
[22:11:00 CEST] <GuiToris> group of pictures ChocolateArmpits?
[22:11:05 CEST] <ChocolateArmpits> GuiToris, yes
[22:11:41 CEST] <ChocolateArmpits> large gop sizes will slow down seeking considerably
[22:12:12 CEST] <ChocolateArmpits> you should try around 1 second long gop for your framerate
[22:12:18 CEST] <GuiToris> I'll look this up
[22:12:27 CEST] <ChocolateArmpits> but import the video into your NLE as well to check if that works for you
[22:12:47 CEST] <ChocolateArmpits> if it's choppy, then consider lowering the value even more
[22:13:11 CEST] <furq> GuiToris: presets specifically affect quality, not necessarily filesize
[22:13:33 CEST] <JEEB> furq: libx264 presets control how much effort/cpu time the encoder puts into encoding the content
[22:13:45 CEST] <JEEB> so if you are using the same bit rate, you are going to lose quality
[22:13:51 CEST] <GuiToris> furq, does that mean I should always use -crf 18 -preset veryslow than ultrafast?
[22:13:54 CEST] <furq> i assume this is in crf mode
[22:14:05 CEST] <JEEB> if you are somehow matching the quality between encodes, then it will just affect the file size
[22:14:05 CEST] <furq> in which case slower presets will look nicer and not necessarily be smaller
[22:14:53 CEST] <JEEB> basically, you get less bang for your buck (bit rate) :P
[22:15:09 CEST] <JEEB> GuiToris: you should basically always use the slowest preset that is fast enough for you
[22:15:18 CEST] <JEEB> you can test with like 5min of content or so :P
[22:15:20 CEST] <furq> for a final encode, yeah
[22:15:28 CEST] <furq> but if you're editing later i'd be inclined to use a much lower crf
[22:15:35 CEST] <furq> in which case the preset is less important
[22:15:53 CEST] <JEEB> well if this is for a source of NLE :P then I would just go with Ut Video or something
[22:15:58 CEST] <furq> or that, yeah
[22:16:02 CEST] <GuiToris> furq, even lower than 18?
[22:16:10 CEST] <JEEB> because you have VFW/DShow/MF and QT modules for Ut Video
[22:16:17 CEST] <furq> 18 is pretty much fine for a final export
[22:16:32 CEST] <furq> but if you're going to reencode it then i'd want to minimise loss as much as possible
[22:16:34 CEST] <JEEB> 18 is not magical, please teach people to actually figure out the best CRF for their content
[22:16:52 CEST] <JEEB> it depends on the type of content, resolution and frame rate
[22:17:07 CEST] <GuiToris> furq, final :O I usually use much higher numbers around 25
[22:17:15 CEST] <furq> ew
[22:17:58 CEST] <GuiToris> -oh yes, I'm talking about x264-
[22:19:22 CEST] <GuiToris> ChocolateArmpits, which option control that gop value that you were refering to?
[22:19:28 CEST] <furq> -g
[22:19:52 CEST] <ChocolateArmpits> ^
[22:20:20 CEST] <GuiToris> hmm, thanks I found something completely different
[22:21:00 CEST] <ChocolateArmpits> -x264-params keyint ?
[22:22:01 CEST] <GuiToris> I found that too
[22:22:08 CEST] <GuiToris> and different codecs
[22:22:16 CEST] <GuiToris> I looked up gop specifically
[22:27:51 CEST] <GuiToris> ChocolateArmpits, sooo I did some research. The higher the gop number, the bigger the filesize is, but seeking is faster. It doesn't affect the quality of the pictures, does it?
[22:28:17 CEST] <ChocolateArmpits> that research seems faulty
[22:28:57 CEST] <ChocolateArmpits> gop number has no direct connection to the filesize mostly
[22:29:11 CEST] <ChocolateArmpits> and high gop numbers certainly do not make seeking faster
[22:29:35 CEST] <GuiToris> how can I find out the current gop number?
[22:30:22 CEST] <ChocolateArmpits> GuiToris, of any file ?
[22:31:32 CEST] <GuiToris> I have a video file here, I'd like to try experimenting with the different gop numbers
[22:31:47 CEST] <GuiToris> it's a 29.97 framerate video
[22:33:09 CEST] <GuiToris> ChocolateArmpits, https://ptpb.pw/PeHC
[22:33:20 CEST] <ChocolateArmpits> you need some stream analyzer software and then look for i-frame pattern
[22:34:38 CEST] <GuiToris> "you should try around 1 second long gop for your framerate"   is it -g 29?
[22:35:27 CEST] <ChocolateArmpits> 30 is more common
[22:36:08 CEST] <GuiToris> I can't use -c copy now, can I?
[22:36:16 CEST] <ChocolateArmpits> nope
[22:37:11 CEST] <GuiToris> and if the framerate is 25 than should I use -g 25?
[22:37:17 CEST] <ChocolateArmpits> ye
[22:37:24 CEST] <GuiToris> thanks you ChocolateArmpits
[00:00:00 CEST] --- Mon Sep 17 2018