[Ffmpeg-devel-irc] ffmpeg.log.20191011

Sat Oct 12 03:05:03 EEST 2019

[00:00:11 CEST] <kepstin> the better tvs even have telecine cadence detection so they can playback 24p stuff in 60i smoothly
[00:00:32 CEST] <gmaciolek> I guess people have gotten used to a lack of "unison" like back in the day w/analog stuff
[00:00:58 CEST] <gmaciolek> I hate to say this, but, I wish TV was more like Apple. XD
[00:01:29 CEST] <gmaciolek> "The hell with backward compatability (past a little while)" - 68k to PPC, OS 7/8/9 --> OS X, and PPC ->x86
[00:01:31 CEST] <kepstin> the annoying thing was that someone went "well, we have this motion interpolating stuff in our tvs anyways for deinterlacing, why not apply it to all video and upconvert everything to 60p (or more)?"
[00:01:51 CEST] <gmaciolek> kepstin: yeah, that stuff drives me NUTS
[00:01:57 CEST] <gmaciolek> some famous actor has a spot about it?!
[00:02:41 CEST] <kepstin> if we didn't have interlaced HD video, then we probably wouldn't have ever had tvs with builtin "smooth motion" features :)
[00:02:44 CEST] <gmaciolek> Tom Cruise, of all people, made a whole video about disabling frame interpolation
[00:03:55 CEST] <gmaciolek> ... I sure wish my clipboard was working, so I could put the link here
[02:46:05 CEST] <Atlenohen> ffmpeg does do SVT-VP9 already ?
[02:46:18 CEST] <Atlenohen> dang it, I don't have AVX2
[04:04:19 CEST] <AlexLearner> Good evening! Please, how could I duplicate an audio stereo pair through ffmpeg? Meant, on a certain video file, I have two audio channels and, on target, I would like to have a copy of these channels, as a third and a fourth audio channels. Could you help me?
[04:05:28 CEST] <DHE> so 4 channels where channels 1 and 3 are actually the same, and 2 and 4 are also the same?
[04:05:40 CEST] <AlexLearner> yes!
[04:05:53 CEST] <AlexLearner> 1-2, 1-2
[04:06:20 CEST] <DHE> probably the `pan` filter
[04:07:38 CEST] <AlexLearner> Pan filter, gonna check. I was trying something concerning those -map and -map_channel options; it is a wrong path, isn't it?
[04:08:13 CEST] <DHE> -map is for selecting which audio feed to use (eg: on a DVD with a commentary audio track, map would let you choose regular audio vs that commentary track)
[04:09:37 CEST] <DHE> hmmm... map_channel might work. -map_channel 0.0.0 -map_channel 0.0.1 -map_channel 0.0.0 -map_channel 0.0.1   # or maybe 0.1.x for a movie file
[04:10:50 CEST] <AlexLearner> Gonna try. Thank you very much, DHE++
[04:16:02 CEST] <AlexLearner> Huh, tried: ffmpeg -i source.mpg -map_channel 0.0.0 -map_channel 0.0.1 -map_channel 0.0.0 -map_channel 0.0.1 -y target.mpg
[04:16:15 CEST] <AlexLearner> but didn't work, unfortunatelly.
[04:18:53 CEST] <AlexLearner> the "target" file won't have the extra audio channels.
[04:23:11 CEST] <furq> AlexLearner: that should work but the second digit needs to be the track number in the file
[04:23:16 CEST] <furq> so it's probably 0.1.0
[04:23:28 CEST] <furq> or just -af "pan=quad|c0=c0|c1=c1|c2=c0|c3=c1" will be shorter anyway
[04:23:38 CEST] <furq> also you'll want to do -c:v copy or else it'll reencode the video
[04:24:41 CEST] <AlexLearner> Gonna try. Thank you, furq++
[04:25:58 CEST] <furq> map_channel just gets passed to pan anyway so you might as well use it directly
[04:32:03 CEST] <AlexLearner> Understood. Tried the pan filter right now but it doesn't do the trick, still. Will revise, here.
[04:32:20 CEST] <AlexLearner> Thank you again, furq++
[09:56:29 CEST] <symtab> hi...i'm trying to add a intro to my video using concat, however for videos that has height > width the resulting video does not play in chrome (it plays in firefox, with ffplay, with vlc, but NOT in chrome), so i decided to convert the video and add black bars, but it still doesnt work, any ideas why this is happening? I attached my conversion
[09:56:29 CEST] <symtab> commands:
[09:56:31 CEST] <symtab> https://pastebin.com/hnGdbEbK
[09:56:34 CEST] <symtab> thank you!
[10:19:20 CEST] <symtab> anyone?
[10:24:58 CEST] <poutine> symtab: instead of "didn't work" do you think you could state what is occurring that is unexpected, or what isn't occurring that is expected?
[10:40:19 CEST] <symtab> basically when using a html5 player to display the video in chrome, the video stops after the intro part is displayed, but this only happens in google chrome, in firefox, vlc, ffplay...it works...
[10:41:10 CEST] <symtab> and by the video stops i mean suddenly the player displays the play button, cant fast forward, nothing...and this only happens if the source video has height > width (vertical video)
[10:41:25 CEST] <symtab> so it must be something related to this, but i was not able to figure this out in 2 days of testing...
[11:02:50 CEST] <symtab> poutine: any ideas?
[11:02:53 CEST] <symtab> thank you!
[14:42:04 CEST] <transhuman> anyone know how you use or run ffms2? No directions!
[14:44:59 CEST] <JEEB> transhuman: simplest for end users (if you don't want to utilize its C/C++ API) is to either utilize it through avisynth or vapoursynth
[14:46:07 CEST] <transhuman> ah it works as a plugin for them?
[14:46:22 CEST] <transhuman> which is easier for beginners would you say JEEB?
[14:47:55 CEST] <transhuman> NM going to try both thanks
[14:48:03 CEST] <JEEB> at this point it depends on which is simpler to install for you. on windows both vapoursynth and avisynth have installers, so you might as well go vapoursynth. avisynth doesn't really run on native *nix so it'd be a wine thing there, vapoursynth would have to be built or if packaged for your distro just installed
[14:48:17 CEST] <JEEB> avisynth has its own language and with vapoursynth you write python scripts
[14:48:48 CEST] <transhuman> ah so vopoursynth would be better for me since I use both platforms
[14:48:51 CEST] <transhuman> thanks
[14:49:37 CEST] <JEEB> avisynth has a long history of generally working nicely under wine, but yea - it's not "native" (among other issues with it in general)
[14:49:48 CEST] <Chagall> avisynth is simpler
[14:49:55 CEST] <JEEB> vapoursynth is just not packaged in all the distros and I've honestly never tried to build it :P
[14:50:06 CEST] <JEEB> thus avs under wine can be simpler to get running
[14:50:08 CEST] <transhuman> ok thanks for the warning
[14:50:31 CEST] <JEEB> Chagall: dunno, both have an editor with preview (AvsPmod, vapoursynth editor)
[14:50:54 CEST] <JEEB> but I guess avs not being a python thing can make it appear simpler to utilize at first :)
[14:50:57 CEST] <JEEB> as it's a domain specific language
[14:51:07 CEST] <Chagall> I find avspmod a better experience overall and more importantly the syntax is less verbose
[14:51:19 CEST] <JEEB> :)
[14:51:43 CEST] <Chagall> though if it's just to use ffms2 I'd use vs
[14:51:50 CEST] <JEEB> I partially got tired of having to go through hoops with high bit depth on avs, although I did use it until 2015 or so?
[14:51:57 CEST] <JEEB> then I switched my usage to vs
[14:51:59 CEST] <Chagall> it's just a bit annoying when you are doing a lot of filtering and have to prefix package names every time
[14:52:47 CEST] <transhuman> once I figure out a pipeline to do what I want, then I want to be able to script it all with python
[15:04:59 CEST] <transhuman> compiling all this crap makes me wish I had one of those xeon 64 core 128 thread processors...
[15:05:10 CEST] <DHE> you mean Epyc
[15:05:15 CEST] <transhuman> ya
[15:05:29 CEST] <DHE> well I want 2 of those. 128 cores, 256 threads
[15:05:58 CEST] <transhuman> I take that and raise you to 256 by 512 lol
[15:06:32 CEST] <transhuman> one single 10 THz processor core would be nicer though, now that I am dreaming
[15:13:10 CEST] <transhuman> DHE problem is compiling can only get so fast with multicore, multithreading. Not sure what the limit is but I liked the days where everything was simple and single core!
[15:14:15 CEST] <DHE> transhuman: you'd be amazed how well compiling can be multi-core scaled. especially big projects like the linux kernel or ffmpeg
[15:14:26 CEST] <transhuman> now I have a serious problem, one of the vapoursynth packages requires python3.6-dev and ubuntu no longer carries it
[15:14:49 CEST] <transhuman> ah so we haven't really reached that limit yet, thats good
[15:26:19 CEST] <zerodefect> I want to decode files from AWS S3 (C SDK). I was going to create an AWS presigned url so that I get an HTTPS url and then use the https libavformat plugin to stream the file off S3.  I don't see a problem there, but I think I'll have issues when, for example, I have MOV reference files that have DREFs. Is there anyway I can intercept the calls in FFmpeg SDK to open the elementary files and
[15:26:19 CEST] <zerodefect> create presigned URLS for those files too before trying to open files?
[16:24:02 CEST] <ncouloute> So I finally figured out a way to tell where ffmpeg will move the timestamp to. Just have to turn on loglevel debug and look for [concat @ xxxx] file: n ..... pts_time:y.yyyy . Although this log file can get really large.  Looking for the first occurance of that for a given file and the last occurance gives me the start and stop time.  Looking at the source code that seems to be the only way unless I use the api
[16:24:02 CEST] <ncouloute> directly. Thought there might be a clever way using concat filter and filter complex but doesnt seem like it. Since showinfo is not support with concat. How easily does the ffmpeg api translate to the given flags? I'm assuming there is a function for each flag?
[16:36:37 CEST] <kepstin> many of the flags (command line options) on the ffmpeg cli map directly to AVOption options that can be set on various contexts
[16:37:03 CEST] <kepstin> some of the ffmpeg cli options control functionality implemented within the ffmpeg cli code, and doesn't exist in the ffmpeg api
[16:43:21 CEST] <ncouloute> Gotcha so I will have to just see what happens then... Think I'm going to try and get away with the parsing the log for now.. I just dont know how the performance of the reading a file that could be 2gb+. I do want to explore the ffmpeg api though as that will make what I'm doing a lot easier. Since I can write actual code and not parsing command line output.
[16:47:26 CEST] <kepstin> ncouloute: if you're using the api, it's entirely possible for you to do the timestamp adjustment for concatenation yourself rather than use the concat filter, so you know exactly what's happening.
[16:47:49 CEST] <kepstin> (or you can do something like track frames going into the filter chain vs. frames going out)
[16:49:22 CEST] <ncouloute> yeah the latter is what I would like to do. I dont necessarily need it to stay where it is. Just need to know where it ends up. Since I do things like converte to 60000/1001 frame rate.
[19:43:50 CEST] <ekacnet> What is the support policy for ffmpeg for the different versions in term of security updates
[19:48:13 CEST] <BtbN> I doubt there is one.
[19:50:10 CEST] <ekacnet> so when should we expect 3.2 or 3.4 to be dropped in term of support ?
[19:51:06 CEST] <BtbN> Usually whenever no major Linux distros still uses it
[19:51:26 CEST] <JEEB> officially, possibly at any time. unofficially, asking the person doing most back-porting to those branches is probably most accurate
[19:51:29 CEST] <JEEB> https://trac.ffmpeg.org/wiki/Downstreams
[19:51:32 CEST] <JEEB> there's also this
[20:19:40 CEST] <ekacnet> JEEB: could a link to that page be added on https://ffmpeg.org/security.html ?
[20:24:27 CEST] <BtbN> The downstreams page isn't strictly related to security.
[21:01:18 CEST] <Chagall> I'd like to get objective quality metrics from 2 clips but only as far as the length of the shorter one, is there a quick way to do that where I could reuse the command line regardless of the minimum length between the two clips?
[21:01:56 CEST] <Chagall> or at least knowing which of the clips is the shorter one but independently of the actual length?
[21:04:00 CEST] <Chagall> hm, actually I guess it already does that
[21:11:29 CEST] <Chagall> oh nevermind it doesn't
[21:18:03 CEST] <JEEB> kepstin: https://github.com/google/ExoPlayer/commit/62618f24ec012020b84200c37c0a8bc215018241
[21:18:11 CEST] <JEEB> now we know where gav1 was going :P
[21:18:47 CEST] <kepstin> i still wonder "why not dav1d?" :/
[21:19:03 CEST] <kepstin> not like there's a license issue there
[21:19:15 CEST] <JEEB> we don't have visibility, but it clearly seems like they dislike pulling in stuff that's not under their control
[21:19:47 CEST] <JEEB> they seemingly say it's because dav1d doesn't have 32bit ARM SIMD, but IIRC gav1 is not at all faster :P
[21:24:36 CEST] <kepstin> seems like it would be easier to fix that than write an entire new decoder from scratch :)
[21:24:53 CEST] <kepstin> (fix dav1d not having arm simd)
[21:53:49 CEST] <JEEB> kepstin: but then it wouldn't be something causing you to get a bonus as it's not a GOOG project :)
[21:56:18 CEST] <Chagall> I enjoy NIHing software, maybe I should apply at google :P
[22:12:04 CEST] <^Neo> Hello! I noticed a bunch of pending BMD DeckLink patches to support multichannel/format SDI audio and AFD... they're about 2 years old now and I'm curious if anyone can comment what's the next steps to getting them merged if anyone is familiar with the situation
[22:47:31 CEST] <void09> I found a .ts video file that plays fine, but ffmpeg refuses to cut/transfor it into mkv. it stops after a few tens of kb written
[22:47:52 CEST] <void09> [matroska @ 0x558e73fb84c0] Timestamps are unset in a packet for stream 0. This is deprecated and will stop working in the future. Fix your code to set the t
[22:47:52 CEST] <void09> [matroska @ 0x558e73fb84c0] Can't write packet with unknown timestamp
[22:47:52 CEST] <void09> av_interleaved_write_frame(): Invalid argument
[22:55:25 CEST] <^Neo> try setting the framerate explicitly?
[22:55:35 CEST] <^Neo> or setPTS?
[22:55:43 CEST] <void09> dunno what setPTS is
[22:57:58 CEST] <void09> these are hdtv dvb-c captures, i've cut a bunch before and got these errors, but it would cut find, now this one for some reason doesn't
[22:58:44 CEST] <void09> cut fine*
[23:02:00 CEST] <void09> this is the full error log: https://pastebin.com/7qGuvKkz
[23:02:43 CEST] <^Neo> can you share the CLI command?
[23:03:32 CEST] <void09> just ffmpeg -i input.ts -framerate 25 -vcodec copy -acodec copy outputvideo.mkv
[23:03:45 CEST] <void09> framerate 25 was just added per your suggestion  ,but doesn't help
[23:04:31 CEST] <kepstin> what's the source of this ts file? sounds like it has a corruption or timestamp jump in the middle that can't be represented in mkv
[23:04:32 CEST] <void09> tried cutting from the 23rd minute onwards, also does not work. i think the whole stream has something ffmpeg does not like
[23:04:58 CEST] <void09> kepstin: it's a dvb-c tuner capture, although not mine. all of mine worked so far
[23:06:03 CEST] <void09> but I also usually get the timestamps errors and the conversion runs fine, despite them
[23:06:10 CEST] <^Neo> https://video.stackexchange.com/questions/24462/timestamps-are-unset-in-a-packet-for-stream-0-this-is-deprecated-and-will-stop
[23:07:26 CEST] <kepstin> most of the options in the answers on that stackexchange are only appropriate for live sources, not for a recorded file :/
[23:07:43 CEST] <kepstin> does it work if you re-encode the video rather than copy?
[23:08:03 CEST] <kepstin> also note that seeking to cut the video probably won't work if the timestamps are broken
[23:08:04 CEST] <void09> I did no try this
[23:09:00 CEST] <void09> kepstin: it plays perfectly, no errors whatsoever.. also got the timestamps error on almost all the recordings i made (even though the recorder counted no errors when capturing), and they cut/convert to mkv fine
[23:09:32 CEST] <void09> fI assume if it can be played, it can be converted to mkv somehow, maybe with some tweaks
[23:09:44 CEST] <^Neo> -fflags +genpts should help?
[23:09:55 CEST] <^Neo> or only for transcoding?
[23:10:08 CEST] <void09> where do i put that, after input, or after output ?
[23:10:12 CEST] <^Neo> uhhhhhhhhhhhhh
[23:10:29 CEST] <kepstin> genpts will probably cause sync issues, or it might change the playback speed. It's an input option.
[23:10:37 CEST] <^Neo> before the input
[23:10:51 CEST] <kepstin> note that genpts requires that there's a dts present
[23:11:02 CEST] <kepstin> so it might work in certain kinds of corruption
[23:11:05 CEST] <^Neo> ffmpeg -fflags +genpts -i file.ts ...
[23:11:08 CEST] <^Neo> @kept
[23:11:13 CEST] <void09> does not work :\
[23:11:14 CEST] <^Neo> yarrrr
[23:11:15 CEST] <shibboleth> since genpts is said to be such a poor option with re to multiinput sync
[23:11:19 CEST] <shibboleth> what is recommended?
[23:11:48 CEST] <void09> is anyone willing to take a look at the file ? I can do a raw cut of the first 200MB or so
[23:15:58 CEST] <void09> https://pastebin.com/MNi8iBuu
[23:16:03 CEST] <void09> this is what I get when tryign encoding
[23:19:26 CEST] <^Neo> Can you do something like `ffmpeg -i "input.ts" -an -c:v rawvideo -f raw -vframes 100 tmp.yuv`
[23:19:39 CEST] <^Neo> and get output?
[23:19:49 CEST] <^Neo> or can you ffplay the file?
[23:21:38 CEST] <void09> [NULL @ 0x55a79000c940] Requested output format 'raw' is not a suitable output format
[23:21:38 CEST] <void09> tmp.yuv: Invalid argument
[23:22:34 CEST] <void09> ffplay input.ts works
[23:25:16 CEST] <^Neo> sorry, -f rawvideo
[23:27:01 CEST] <pink_mist> is that -c:v rawvideo should be change to -f rawvideo or is that -f raw should be changed to -f rawvideo or some combination of the above?
[23:27:06 CEST] <pink_mist> *changed
[23:27:53 CEST] <void09> ^Neo: that seems to work, done 5000 frames so far
[23:28:21 CEST] <^Neo> ok, so it's not the decoder
[23:28:45 CEST] <^Neo> `ffmpeg -i "input.ts" -an -c:v rawvideo -vframes 100 -f rawvideo tmp.yuv`
[23:29:03 CEST] <^Neo> so you're going to fill out your hard disk if you don't have vframes set
[23:29:23 CEST] <void09> 29GB for 10k frames :D
[23:29:52 CEST] <void09> ffplay tmp.yuv
[23:30:02 CEST] <void09> [IMGUTILS @ 0x7fb9d67fb870] Picture size 0x0 is invalid
[23:30:02 CEST] <void09> tmp.yuv: Invalid argument
[23:30:58 CEST] <void09> should the .yuv be playable with vlc/mpv ?
[23:31:20 CEST] <durandal_1707> .yuv is rawvideo, without headers
[23:31:39 CEST] <void09> ok, cause no player wants to play it
[23:31:41 CEST] <durandal_1707> you need to remember size, pixel format and etc to play it back
[23:32:02 CEST] <durandal_1707> store video in some container next time
[23:32:06 CEST] <durandal_1707> like .y4m
[23:32:20 CEST] <durandal_1707> it is still rawvideo, but with headers
[23:32:22 CEST] <void09> ok I maanged to cut and convert to mkv with mkvtoolnix.. trying to see if it errors on the mkv
[23:32:25 CEST] <durandal_1707> mpv will play it
[23:34:03 CEST] <void09> ok, ran ffmpeg on the split remuxed video by mkvtoolnix, I get a continuous spam of these errors: [matroska @ 0x55961767ea00] Non-monotonous DTS in output stream 0:0; previous: 1389700, current: 1389660; changing to 1389700. This may result in incorrect t
[23:34:28 CEST] <void09> [matroska @ 0x55961767ea00] Non-monotonous DTS in output stream 0:0; previous: 1389700, current: 1389660; changing to 1389700. This may result in incorrect timestamps in the output file.
[23:35:05 CEST] <void09> but it processed the file
[23:35:12 CEST] <durandal_1707> mkvtoolnix problem
[23:35:58 CEST] <void09> durandal_1707: have you been following the problem I reported since the beginning ? it was about a .ts file that ffmpeg can't conver to mkv
[23:37:12 CEST] <durandal_1707> void09: feel free to upload .ts somewhere sane place and report bug to ffmpeg trac
[23:40:58 CEST] <void09> https://drive.google.com/open?id=1ZQUrBhJjLOpvaNBQIBkLuIs5Gm-1nUul
[23:41:48 CEST] <durandal_1707> i said to report it on trac
[23:41:55 CEST] <durandal_1707> i have no time now
[23:45:13 CEST] <void09> heh, seems my ip is  blacklisted by bl.spamcannibal.org [103.224.212.227]
[23:48:54 CEST] <durandal_1707> void09: what is your problem with that file?
[23:49:19 CEST] <void09> https://trac.ffmpeg.org/ticket/8256#ticket
[23:49:22 CEST] <void09> did i do it right?
[23:54:55 CEST] <durandal_1707> yes
[23:54:57 CEST] <void09> ffmpeg -i "input.ts" -vframes 1000 outputfile.y4m - produces a playable file
[23:57:14 CEST] <void09> however, "ffmpeg -i "input.ts" -an -c:v rawvideo -f rawvideo -vframes 1000 tmp.y4m" does not produce a playable file. what is wrong with this line?
[00:00:00 CEST] --- Sat Oct 12 2019