[Ffmpeg-devel-irc] ffmpeg.log.20170402

[burek]

[00:49:04 CEST] <angular_mike_> how to play video with ffmpeg, without ffplay?
[00:49:26 CEST] <angular_mike_> I want to apply the filter and preview results without dumping to disk
[01:07:07 CEST] <klaxa> angular_mike_: how about: ffmpeg [whatever] -f matroska pipe:1 | mpv -
[01:07:59 CEST] <angular_mike_> klaxa: Failed to recognize file format
[01:08:42 CEST] <klaxa> huh can you pastebin your command + output?
[01:08:48 CEST] <angular_mike_> oh nvm. typo
[01:57:27 CEST] <velus> hello all im looking to create a streaming server but i need to be able to stream using rtmps can this be done with ffmpeg
[02:39:07 CEST] <DHE> ffmpeg isn't a server platform. but it can feed a server
[02:55:45 CEST] <petecouture> velus: Wowza maybe a better option for you
[03:01:58 CEST] <marbar> unomystEz, are you doing hls?
[03:04:42 CEST] <unomystEz> hey
[03:04:55 CEST] <unomystEz> marbar: hls?
[03:06:54 CEST] <marbar> unomystEz, yeah, are you slicing up your video for ios playback?
[03:07:12 CEST] <marbar> unomystEz, w/ an m3u8 file?
[03:07:39 CEST] <unomystEz> marbar: no, just slicing it up for youtube
[03:08:27 CEST] <marbar> unomystEz, ah, ok. yeah, i'm trying to process a video for ios playback, but so far i haven't gotten it to play back all the way without freezing :'(
[03:12:19 CEST] <unomystEz> marbar: how are you splitting your videos?
[03:14:15 CEST] <b0bby__> hey
[03:14:16 CEST] <b0bby__> trying to learn how to use the ffmpeg libraries (libav) and Im getting errors([vp8 @ 0x19492a0] Invalid profile 5.
[03:14:16 CEST] <b0bby__> [vp8 @ 0x19492a0] Header size larger than data provided) with the code https://gist.github.com/anonymous/2cd6151c667146dceb0b7b08762de903(taken for the examples page). It worked fine with mpeg formates but I tried changing the codec to vp8 and giving the program a webm only to get errors.
[03:15:47 CEST] <marbar> unomystEz, i've tried a lot of different techniques. here's the one i just tried (that also didn't work):
[03:15:54 CEST] <marbar> ffmpeg -i 9 -map 0 -c:v libx264 -c:a aac -r 25 -force_key_frames 'expr:gte(t,n_forced*10)' -f ssegment -segment_list hls/9/list.m3u8 -segment_list_flags +live -segment_time 10 -segment_time_delta 0.02 hls/9/%03d.ts
[03:16:03 CEST] <marbar> unomystEz, by didn't work i mean the video stalls at some point
[03:16:13 CEST] <marbar> unomystEz, on an ipad
[03:16:34 CEST] <marbar> unomystEz, w/ an html5 video tag pointing at the m3u8 file
[03:16:56 CEST] <unomystEz> marbar: hmm, I don't have any apple products but I did read on the ffmpeg website about enabling support for legacy devices?
[03:17:13 CEST] <furq> marbar: pastebin the m3u8 somewhere
[03:17:37 CEST] <marbar> unomystEz, i see, i'll look that up, thanks.
[03:17:43 CEST] <marbar> furq, thanks dude will do!
[03:20:32 CEST] <marbar> furq: here's the m3u8 generated by the above command: https://pastebin.com/yqHrbi5p thanks for helping!
[03:21:02 CEST] <b0bby__> can I have some help(ffmpeg c api)
[03:23:02 CEST] <furq> well it's not the thing i suspected
[03:23:25 CEST] <furq> apple have an hls verification tool but you need an apple developer id to even download it and i have no idea what platforms it runs on
[03:23:30 CEST] <furq> but that might be worth looking into
[03:23:38 CEST] <furq> "media streaming validator tool"
[03:24:31 CEST] <marbar> furq, thanks dude. i'll look into crapple's tool :)
[03:25:28 CEST] <furq> apparently iOS recently started bailing out if it found segments more than 20% smaller than EXT-X-TARGETDURATION
[03:25:38 CEST] <marbar> no way, seriously?
[03:25:39 CEST] <marbar> that's insane
[03:25:41 CEST] <marbar> hahahahaha
[03:25:48 CEST] <furq> you have one of those but it's the last segment, and i can't imagine that would break it
[03:26:30 CEST] <marbar> right, it breaks pretty early, too :-\
[03:26:32 CEST] <furq> the guy who was reporting that in here had every segment 50% smaller
[03:27:07 CEST] <marbar> interesting...
[03:28:47 CEST] <furq> fwiw if you can't run the tool then all the warnings it might spit out are here
[03:28:48 CEST] <furq> https://developer.apple.com/library/content/technotes/tn2235/_index.html
[03:28:56 CEST] <furq> you can check some of those manually
[03:29:29 CEST] <furq> although because it's apple, there's no official word on which of those warnings are actually errors
[03:30:15 CEST] <marbar> hahahahaha, oh man, thanks, i'll give that a shot. apple is hilarious.
[03:31:31 CEST] <furq> well it's not as if you're only using hls because apple refuses to let ios devices play back dash
[03:31:46 CEST] <marbar> right, of course
[03:31:47 CEST] <furq> you're actually only using hls because apple refuses to let ios devices play back dash on any site that isn't netflix.com
[03:32:09 CEST] <marbar> yup, definitely intentional. they're dorks.
[03:32:51 CEST] <furq> it is pretty funny that i only ever see people complaining about iOS support for hls in here
[03:33:20 CEST] <furq> all the various hacks to get hls to work in other browsers work better than the platform which is forcing you to support it
[03:33:23 CEST] <marbar> well their goal is to funnel everyone into their app store
[03:33:26 CEST] <marbar> they don't want nice web apps
[04:02:34 CEST] <main> hi all i have a question... what is the option like "-vb/-b:v" for a variable/maximum bitrate?
[04:02:44 CEST] <main> i always thought "-vb/-b:v" is already variable suddently wondering aboud the output file size and realizing by lookin at the manual that its a fix bitrate
[04:02:55 CEST] <main> i want to have an output with max 3Mbps and min as less as possible withoud quality loss
[04:03:08 CEST] <furq> main: which video codec
[04:03:18 CEST] <main> theora
[04:04:20 CEST] <furq> oh
[04:04:22 CEST] <furq> no idea then
[04:04:23 CEST] <main> but i could use anotherone i just think its the smallest
[04:04:28 CEST] <furq> it really isn't
[04:04:34 CEST] <furq> theora is ancient and pretty poor
[04:04:47 CEST] <main> :_)
[04:05:06 CEST] <main> i could use mp4 aswell
[04:05:17 CEST] <main> i just want a variable bitrate
[04:05:33 CEST] <furq> well yeah with x264 you'd use -maxrate 3000k -bufsize 6000k -crf 20
[04:05:42 CEST] <furq> change -crf as you see fit, lower values are higher quality
[04:06:06 CEST] <main> what does crf mean?
[04:06:12 CEST] <furq> constant rate factor
[04:06:29 CEST] <main> ok ty ill trye
[04:06:45 CEST] <furq> it pretty much just means constant quality
[04:07:03 CEST] <furq> rather than trying to figure out an average bitrate for each clip which looks good
[04:07:05 CEST] <main> and just cause im curious... would  -maxrate 3000k -bufsize 6000k -crf 20 work with theora aswell?
[04:07:10 CEST] <furq> no
[04:07:33 CEST] <furq> if you want something that isn't patent-encumbered then you can use libvpx-vp9
[04:07:42 CEST] <furq> in which case use something like -b:v 3000k -crf 30
[04:08:00 CEST] <furq> but otherwise x264 is generally a better choice
[04:08:07 CEST] <main> ok ty for the fast answer ill make a trye
[04:13:02 CEST] <main> i just see i cant use x264 cause the inputfile is h264... what container do i need for libvpx-vp9?
[04:27:15 CEST] <kepstin> why wouldn't you be able to use x264? the input file format doesn't matter...
[04:27:31 CEST] <kepstin> and libvpx-vp9 should work in mkv (matroska) or webm
[04:29:53 CEST] <main> i dont understand but if i convert the infile with the same codec as the out file e.g h264->x264 it is damaged but i just convertet a h264 withoud to specify the vcodec option but the -f mov and then it worked to convert h264 to x264
[04:30:04 CEST] <main> ha matroska ty
[04:30:12 CEST] <main> *ah
[04:49:21 CEST] <mavi> Just curious why does youtube reccomend closed GOP and limiting bframes to 2 for uploads?
[04:49:33 CEST] <mavi> Does it allow for less compression losses during a reencode?
[04:50:58 CEST] <furq> i don't think it really matters if you're not live streaming
[04:51:22 CEST] <mavi> https://support.google.com/youtube/answer/1722171?hl=en
[04:51:32 CEST] <mavi>       Progressive scan (no interlacing)     High Profile     2 consecutive B frames     Closed GOP. GOP of half the frame rate.     CABAC     Variable bitrate. No bitrate limit required, though we offer recommended bit rates below for reference     Chroma subsampling: 4:2:0
[04:51:54 CEST] <mavi> Does it make live streams look worse? Usually more bframes means better compression, no?
[04:54:30 CEST] <furq> iirc they don't reencode live streams so you need to use broadly compatible settings
[04:55:13 CEST] <furq> but stuff like "gop of half the framerate" is just ridiculous
[04:55:33 CEST] <furq> also 384k aac for stereo audio
[04:55:44 CEST] <mavi> isn't that pretty much lossless?
[04:55:58 CEST] <furq> 128k stereo is pretty much lossless with a halfdecent encoder
[04:57:05 CEST] <furq> the only setting on there which actually makes a difference is making sure your mp4 is faststart
[04:57:09 CEST] <furq> because that'll speed up processing
[04:57:20 CEST] <furq> although you could also just use mkv or mpegts
[04:57:39 CEST] <furq> oh and i guess no interlacing is useful
[04:57:48 CEST] <furq> you can disregard more or less everything else
[05:00:21 CEST] <mavi> sounds good. Thanks
[06:15:48 CEST] <marbar> hey folks
[06:16:00 CEST] <marbar> so i was complaining here about hls earlier today
[06:16:09 CEST] <marbar> but i think my issue doesn't really actually involve hls
[06:16:33 CEST] <marbar> i experimented w/ slicing out the problem segment from the original file
[06:16:54 CEST] <marbar> and it's messed up no matter how i've encoded it (so far)
[06:17:09 CEST] <marbar> i just tried with:
[06:17:10 CEST] <marbar> ffmpeg -ss 60 -t 60 -i 9 -c:v libx264 -c:a aac -b:a 32 -r 30 -deinterlace -x264opts "keyint=120:min-keyint=120" -forced-idr 1 -movflags +faststart ipoo5.mp4
[06:17:30 CEST] <marbar> trying to get it running in ios (ipad) chrome, btw
[06:18:09 CEST] <marbar> any obvious problems? or missing ios-sy flags?
[06:19:01 CEST] <marbar> (thanks for any ideas)
[06:20:19 CEST] <furq> do you get any warnings during encoding
[06:21:03 CEST] <marbar> furq, ooo i'll check, thanks
[06:21:30 CEST] <furq> also probably don't use -deinterlace
[06:21:49 CEST] <furq> i'm not even sure what that does these days
[06:22:38 CEST] <marbar> oh i see thanks
[06:22:55 CEST] <marbar> also it warned me that the audio bitrate was too low, so removing that
[06:23:21 CEST] <furq> oh right yeah
[06:23:24 CEST] <furq> -b:a 32 is 32bps
[06:23:28 CEST] <furq> you want 32k (or something higher)
[06:23:47 CEST] <marbar> i see
[06:23:54 CEST] <marbar> well trying now with: ffmpeg -ss 60 -t 60 -i 9 -c:v libx264 -c:a aac -r 30 -x264opts "keyint=120:min-keyint=120" -forced-idr 1 -movflags +faststart ipoo6.mp4
[06:24:04 CEST] <furq> use -vf yadif if you want to deinterlace
[06:25:02 CEST] <marbar> ok, well my target is ipad, so i think i need that in general, although maybe not with this source file
[06:25:18 CEST] <furq> well yeah only do it if your source is interlaced
[06:25:21 CEST] <furq> otherwise you'll lose quality
[06:25:39 CEST] <marbar> i see. well my problem is  that i get green blocks and other glitches in playback
[06:25:53 CEST] <marbar> usually to the point of freezing up
[06:26:09 CEST] <furq> i take it you don't see any warnings during encoding
[06:27:04 CEST] <marbar> not this time around (minus the absurdly low audio bitrate)
[06:27:17 CEST] <furq> you can check if ffmpeg can decode the output file with ffmpeg -i foo.mp4 -f null -
[06:27:20 CEST] <marbar> but on ipad chrome it's a glitchfest. is there any flag in particular for smoothing stuff out for ios?
[06:27:25 CEST] <marbar> oh cool ok
[06:27:43 CEST] <furq> which ipad
[06:27:56 CEST] <marbar> oh good question i'll find out...
[06:28:34 CEST] <thebombzen> is deinterlacing even a problem that can be solved
[06:28:44 CEST] <thebombzen> like could you theoretically deinterlace video perfectly
[06:28:53 CEST] <furq> apparently the original only supports main at 3.1
[06:29:07 CEST] <furq> thebombzen: depends on the encoder
[06:29:24 CEST] <thebombzen> so with lossless video yes
[06:29:28 CEST] <furq> if you had interlaced rawvideo then sure
[06:29:48 CEST] <furq> if you've got artifacts that bleed across fields then probably not
[06:29:57 CEST] <thebombzen> but that's like all lossy encoding
[06:30:05 CEST] <furq> it sure is
[06:30:11 CEST] <furq> that's why interlacing sucks
[06:30:20 CEST] <thebombzen> but then again if you have artifacts that bleed agross fields then how much does the deinterlacing magnify them
[06:30:28 CEST] <furq> this is why we need shit like nnedi which is slower than x264 veryslow
[06:30:56 CEST] <thebombzen> also wouldn't deblocking fudge the fields together?
[06:31:03 CEST] <furq> probably
[06:31:06 CEST] <thebombzen> or is deblocking not really a risk at high enough bitrates
[06:31:30 CEST] <furq> try using yadif on a crap mpeg2video source if you want to give yourself nightmares
[06:31:52 CEST] <furq> and then bear in mind yadif is relatively good
[06:32:09 CEST] <marbar> hahaha
[06:32:34 CEST] <marbar> ok it's an ipad mini (model ME277LL/A) running 10.2.1
[06:33:00 CEST] <furq> that does high at 4.1 (according to this forum post) so it should be ok
[06:33:32 CEST] <marbar> and ffmpeg can decode it (the command you gave me didn't throw any errors for that file)
[06:33:48 CEST] <furq> weird
[06:34:06 CEST] <marbar> i'll try w/ 4.1
[06:34:09 CEST] <furq> pastebin the output of `ffprobe -show_streams output.mp4"
[06:34:15 CEST] <marbar> ok thanks
[06:34:26 CEST] <furq> unless this is 1080p60 then it won't be exceeding level 4.1
[06:34:36 CEST] <furq> and you normally shouldn't set that manually anyway
[06:34:59 CEST] <furq> i would also expect it to just bail out rather than decoding it wrong
[06:36:58 CEST] <marbar> here furq https://pastebin.com/TS6MAazv thanks :)
[06:38:32 CEST] <furq> does it play back ok in another player
[06:39:13 CEST] <marbar> plays back fine in debian chrome
[06:39:43 CEST] <furq> weird
[08:46:59 CEST] <oneforall2> libcdio_paranoia.so: undefined reference to `cdio_cddap_track_audiop at CDIO_CDDA_2'  ffmpeg-3.2.4  libcdio-paranoia-10.2+0.94+1  gcc 6.3.0
[09:01:21 CEST] <sfan5> oneforall2: tried make clean?
[09:23:13 CEST] <oneforall2> looks like a bad build of libcdio-paranoia-10.2+0.94+1
[10:01:02 CEST] <kerio> does opus fit in mp4s?
[10:02:09 CEST] <furq> no
[10:02:28 CEST] <furq> there's a draft spec but nothing official
[10:02:51 CEST] <furq> and even when there is it won't work anywhere until after we're all dead
[10:09:13 CEST] <thebombzen> hey, vorbis in mp4 works
[10:09:17 CEST] <thebombzen> and nobody ever expected that
[10:11:58 CEST] <oneforall2> yup thats what it was rebuilt libcdio-paranoia-10.2+0.94+1  ..
[10:15:28 CEST] <kerio> how are you supposed to stream opus then
[10:15:34 CEST] <thebombzen> matroska
[10:15:37 CEST] <furq> ogg
[10:15:39 CEST] <thebombzen> or mpegts
[10:15:40 CEST] <thebombzen> or ogg
[10:15:46 CEST] <kerio> yeah but ogg how
[10:15:47 CEST] <thebombzen> you're also not supposed to stream mp4 files
[10:15:50 CEST] <kerio> i wanted to do HLS
[10:15:52 CEST] <furq> icecast?
[10:15:55 CEST] <kerio> furq: eq
[10:15:56 CEST] <kerio> ew
[10:16:02 CEST] <thebombzen> with HLS? webm or mpegts afaik
[10:16:06 CEST] <thebombzen> probably mpegts
[10:16:13 CEST] <kerio> you can put opus in mpegts? :o
[10:16:16 CEST] <thebombzen> yea
[10:16:20 CEST] <furq> yeah but you can't do opus over hls
[10:16:24 CEST] <furq> opus is strictly h264+aac
[10:16:28 CEST] <thebombzen> HLS is*
[10:16:28 CEST] <furq> s/opus/hls
[10:16:31 CEST] <thebombzen> yea
[10:16:35 CEST] <thebombzen> well there you go
[10:16:48 CEST] <kerio> fuck it i'm just going to serve .nut
[10:16:53 CEST] <furq> i'm pretty sure an html5 audio tag will accept an icecast stream
[10:16:56 CEST] <furq> just do that
[10:16:57 CEST] <thebombzen> can't with HLS xD
[10:17:02 CEST] <thebombzen> NUT at least
[10:17:38 CEST] <kerio> yeah but icecast requires some extra servering
[10:17:39 CEST] <thebombzen> kerio: also if your'e actually serious about opus-in-nut make sure you have a very recent build of ffmpeg
[10:17:46 CEST] <furq> yeah but on the other hand, it's actually good
[10:17:50 CEST] <furq> unlike hls which is dogshit
[10:18:03 CEST] <thebombzen> opus-in-nut only works as of like a week or two ago
[10:18:05 CEST] <furq> just imagine the thrill of using an actual streaming protocol in a browser
[10:18:07 CEST] <kerio> :o
[10:18:11 CEST] <kerio> furq: but it's just http
[10:18:25 CEST] <furq> it's still streaming
[10:18:31 CEST] <thebombzen> kerio: I know because I'm the one who submitted the patch to put Opus and HEVC support inside .nut
[10:18:48 CEST] <kerio> nice
[10:18:51 CEST] <thebombzen> it was literally just two lines but it works now b/c nut is great
[10:22:13 CEST] <thebombzen> https://git.ffmpeg.org/gitweb/ffmpeg.git/commitdiff/744916908167951c8d3590bc8efeac0ffe35c1ee?hp=14764b93e282f1c1a30d093da915df1b692db126
[10:22:36 CEST] <furq> also it's not really a big deal to host an icecast server
[10:22:43 CEST] <furq> the most arduous part of that is the bit where you have to edit some xml
[10:24:59 CEST] <kerio> D:
[10:25:23 CEST] <kerio> how do you feel about webrtc
[10:26:00 CEST] <furq> i don't
[10:26:14 CEST] <kerio> Beam uses webrtc
[10:26:43 CEST] <furq> i don't know what that is
[10:26:52 CEST] <kerio> the new xbox livestreaming platform
[10:26:58 CEST] <furq> oh dear
[10:27:11 CEST] <kerio> it's like twitch but with webrtc support
[10:27:13 CEST] <kerio> and no Kappa
[10:27:18 CEST] <furq> i don't know what that is
[10:27:33 CEST] <kerio> https://twitchemotes.com/emote/25
[10:27:43 CEST] <furq> i didn't say i wanted to know
[10:28:13 CEST] <kerio> >open a stream
[10:28:24 CEST] <kerio> >guy with green shirt that gets chromakeyed away
[10:28:38 CEST] <furq> what does "greater than open a stream" mean
[10:28:39 CEST] <kerio> >microstuttering like there's some wrong deinterlacing going on
[10:28:49 CEST] <kerio> 10/10 streaming platform
[10:28:58 CEST] <kerio> furq: it's meme arrows
[10:29:05 CEST] <kerio> i'd make them green to signify that but this channel is +c
[10:29:08 CEST] <furq> it sounds terrible and you should stop it
[10:29:19 CEST] <kerio> https://beam.pro/Outpost how does this guy not fix it
[10:29:49 CEST] <furq> if this uses webrtc does that not mean it's stuck with h264 baseline
[10:30:01 CEST] <furq> given that that's what all the desktop browsers use for webrtc decoding
[10:30:26 CEST] <furq> or do they use the system h264 for decoding
[10:30:40 CEST] <kerio> i have no idea, i use safari :^)
[10:30:45 CEST] <furq> oh dear
[10:31:13 CEST] <kerio> you're just jealous of my native HLS support
[10:31:36 CEST] <furq> "ah fuck...goddammnit...if only i had, native hls support" -- me
[10:32:05 CEST] <furq> look at how annoyed i must have been to pronounce goddammnit with two ms and one n
[10:32:37 CEST] <furq> that's the jealousy phoneme
[10:32:54 CEST] <kerio> phonememe
[10:33:03 CEST] <furq> you disgust me
[10:33:28 CEST] <kerio> anyway that site works like shit in chrome canary
[10:33:35 CEST] <kerio> maybe firefox fares better
[10:34:00 CEST] <furq> read that sentence back to yourself
[10:34:19 CEST] <kerio> apparently i don't have webrtc in firefox nightly
[10:34:28 CEST] <kerio> why is everything shit
[10:34:45 CEST] <kerio> also why can't we just embed ffmpeg in everything
[10:34:58 CEST] <furq> not everything is shit
[10:35:02 CEST] <furq> for instance, i am quite good
[10:35:04 CEST] <furq> and uh
[10:35:09 CEST] <furq> there are probably some other things
[10:35:14 CEST] <kerio> pizza is good
[10:35:24 CEST] <furq> it's only good when i'm eating it
[10:35:29 CEST] <furq> when other people are eating it, i don't care for it
[10:35:30 CEST] <kerio> HLS is good
[10:35:35 CEST] <bencoh> nope
[10:35:42 CEST] <furq> what he said
[10:37:13 CEST] <kerio> single-file HLS is neato
[10:37:36 CEST] <bencoh> broken mpeg-ts you mean? :*
[10:37:38 CEST] <furq> that's true. there is no other way to play a single video file
[10:38:32 CEST] <kerio> isn't fragmented isobmff better than mpegts
[10:38:43 CEST] <kerio> or is it worse but less patented
[10:39:03 CEST] <furq> well it's definitely not less patented
[10:41:04 CEST] <bencoh> and most probably "worse"
[10:43:04 CEST] <furq> don't the mpegts patents expire soon
[10:45:55 CEST] <furq> january 2018 if i'm reading this right
[10:46:06 CEST] <furq> or they might have expired already
[10:46:20 CEST] <bencoh> neat
[10:46:26 CEST] <furq> http://www.google.co.uk/patents/US6181712
[10:46:38 CEST] <furq> this is january 2018 and it looks pretty likely to cover any usage of mpegts
[10:46:49 CEST] <furq> but i'm not giving it to a lawyer to find out for sure
[10:47:08 CEST] <furq> that's the last mpeg-2 patent to expire
[20:14:32 CEST] <ZexaronS> hello
[20:14:46 CEST] <dystopia_> hello o/
[20:15:12 CEST] <ZexaronS> I was wondering if it possible to analyze a video file and report back a graph of pixel activity, and sound activity, useful for skipping to areas of interest on security camera ?
[20:16:07 CEST] <dystopia_> i can't think of a way :(
[20:16:46 CEST] <dystopia_> if it's vbr, looking at points when the bitrate increases would tell you there is activity in the shot rather than a static image
[20:32:19 CEST] <ZexaronS> dystopia_ - don't the encoders have analyze pixel difference when they ... ehm, encode and compress ?
[20:40:55 CEST] <xeche> Hello guys. I'm having a bit of trouble with the new encode/decode API. My use case is decoding a 264 stream. I use av_read_frame to read packets, but avcodec_send_packet always fails with an unknown errorcode (-10021541205 something)
[20:41:34 CEST] <xeche> [h264 @04eee40] No start code is found.
[20:41:44 CEST] <xeche> [h264 @04eee40] Erro splitting the input into NAL units.
[20:45:26 CEST] <xeche> This happens with three separate files, so I doubt it's a data issue. Anybody got any pointers?
[21:14:40 CEST] <rose54> Many thanks to all the developers and contributors to ffmpeg.
[23:45:24 CEST] <topher> Hi, I am trying to use FFMpeg to extract frames from a video. The video stream changes from 1920x1080 to 1280x710 to 640x480 several times throughout the duration and I want to extract frames at the correct resolution. However during the extraction process there is an auto-inserted scaler which scales everything to 1080p. Does anyone have any ideas as to how to approach this? command and logs https://pastebin.com/nkejaTz5
[00:00:00 CEST] --- Mon Apr  3 2017