[Ffmpeg-devel-irc] ffmpeg.log.20180313

burek burek021 at gmail.com
Wed Mar 14 03:05:02 EET 2018


[01:00:28 CET] <FishPencil> Is there any way to actually get 60fps out of gdigrab
[01:01:20 CET] <JEEB> no idea, also there are more optimized screen capture APIs since Win8.1+
[01:01:40 CET] <JEEB> not in FFmpeg, but if you like a challenge you could try making something using both that and FFmpeg
[01:02:06 CET] <JEEB> I'd link to the virtualdub guy's blog on the subject
[01:02:18 CET] <JEEB> but unfortunately that is currently giving 500
[01:07:34 CET] <CoreX> JEEB looks up to me http://www.virtualdub.org/docs_capture.html
[01:10:07 CET] <JEEB> no, it was on the blog
[01:10:11 CET] <JEEB> regarding the actual APIs
[06:54:10 CET] <rahul_-> Hello all
[06:54:22 CET] <rahul_-> any one there ?
[07:02:09 CET] <uprhyme> no
[07:05:47 CET] <rahul_-> uprhyme, :D
[07:06:55 CET] <rahul_-> I am trying https://pastebin.com/jaEjJRdS
[07:07:04 CET] <rahul_-> i want to get video codec name
[07:07:21 CET] <rahul_-> Now i am getting NONE everytime
[07:07:25 CET] <rahul_-> but it is H264
[07:23:00 CET] <rahul_-> any suggestions ? why i am getting AV_CODEC_ID_NONE
[07:31:08 CET] <gnarface> rahul_-: my only guess is maybe the file is corrupted
[07:32:15 CET] <gnarface> (but i'm not familiar with the api so i can't verify anything one way or another about that code snippet)
[07:33:33 CET] <rahul_-> gnarface, But the command ffprobe -show_streams -i  /opt/big_buck_bunny_1080p_h264.mov  | grep code
[07:33:44 CET] <rahul_-> is giving me the codec_name=h264
[07:38:16 CET] <rahul_-> gnarface, https://pastebin.com/PgbY2Cj4 i update the code
[07:38:21 CET] <rahul_-> but still same
[07:39:48 CET] <gnarface> rahul_-: i got nothing man, sorry.  same result with other h264 files?
[07:40:41 CET] <rahul_-> gnarface, I haven't checked .. with other files, but the same file with ffprobe command give me that :|
[13:34:51 CET] <leonardo98> I need to convert an ISO image to mp4 it contains menus and chapters how can I do that from ffmpeg?
[13:35:07 CET] <leonardo98> ?
[13:38:33 CET] <BtbN> you won't be able to convert the menus if you mean that
[13:46:01 CET] <leonardo98> BtbN: no i dont want to convert menues i just want skip those menus and convert only the movie part
[13:46:19 CET] <leonardo98> BtbN: no i dont want to convert menus i just want to skip those menus and convert only the movie part
[13:48:30 CET] <BtbN> I guess it's a DVD. I usually use makemkv for those
[13:50:12 CET] <leonardo98> how can I do it from makemkv then
[13:51:54 CET] <BtbN> That should be very obvious
[13:53:27 CET] <leonardo98> is it possible from ffmpeg
[13:53:37 CET] <leonardo98> i mean things like selecting chapters
[13:53:41 CET] <furq> no
[13:54:14 CET] <furq> you need to demux the title first and then run ffmpeg on that
[13:54:20 CET] <furq> if you want to avoid creating a temporary file then use tccat
[14:00:08 CET] <brian__> im trying to concat 2 mp4 files without reencoding, im converting them to ts and then concat them to mp4. Now the audio of the 2nd file is slowed down
[14:00:35 CET] <brian__> and its logging alot of  Non-monotonous DTS in output stream 0:1; warnings
[14:00:41 CET] <brian__> can i do something about this?
[14:02:16 CET] <brian__> https://gist.github.com/Brianvdb/8433e5df411aa8cccfe3c7830d348874 this I execute
[15:31:59 CET] <saml_> brian__, ffmpeg -copyts -i "concat:in1.ts|in2.ts" -muxpreload 0 -muxdelay 0 -c copy joint.ts
[15:36:45 CET] <brian__> saml_ I already found my problem, the audio sample rates were different
[15:37:13 CET] <saml_> ah thanks
[15:37:55 CET] <brian__> the only problem i'm having now is little audio 'pops' at the part where im concatenating the video
[15:38:00 CET] <brian__> like there is no audio for 3ms
[15:38:18 CET] <brian__> and then continues
[15:49:20 CET] <Nacht> Anyone happen to have an example of what '-progress' outputs ? Is it somewhat identical of  -stats ?
[15:50:02 CET] <alexp> brian__: out of curiosity, does the audio also pop midstream if you just pause then unpause the video during playback?
[15:51:53 CET] <brian__> alexpigment nope, only at the parts where the concats are
[15:52:07 CET] <alexpigment> just checking
[15:52:21 CET] <Nacht> Sounds like the same audiopriming issue I'm running into
[15:52:27 CET] <alexpigment> because pops at the start of audio playback using indicate a dcbias issue
[15:52:36 CET] <saml_> don't they provide separate audio file and make .ts files with no audio?
[15:54:29 CET] <saml_> ffmpeg -f concat -i mylist.txt -c copy all.ts;   ffmpeg -i all.ts -acodec copy -vcodec copy all.mp4
[15:54:38 CET] <saml_> does this have same audio popping issue?
[15:55:57 CET] <Nacht> The transcode shouldn't be giving the issue
[16:15:19 CET] <brian__> saml_ ye it does have the same popup issue
[16:15:28 CET] <brian__> popping*
[16:15:51 CET] <brian__> with ur commands
[16:17:29 CET] <brian__> alexpigment anything i can do about that?
[16:18:02 CET] <alexpigment> yeah, check the dcbias filter
[16:18:07 CET] <alexpigment> you have to re-encode
[16:18:13 CET] <alexpigment> the audio i mean
[16:19:47 CET] <brian__> nothing else i can do? I am on android and re-encoding will take too much time
[16:20:05 CET] <alexpigment> re-encoding audio?
[16:20:23 CET] <brian__> i can encode the individual audio parts in each segment
[16:20:39 CET] <brian__> ye ffmpeg on android is slow..
[16:20:54 CET] <brian__> encoding audio which is like 1min takes 40secs or something
[16:21:09 CET] <alexpigment> can you confirm in an audio editor that the waveforms are not on the center line?
[16:21:23 CET] <brian__> any audio editor u can recommend me?
[16:21:55 CET] <alexpigment> audacity?
[16:26:31 CET] <brian__> alexpigment: https://imgur.com/a/w1Klu
[16:27:47 CET] <alexpigment> ok, that's not a dcbias issue
[16:28:07 CET] <alexpigment> the popping could be that the waveform isn't on the centerline at startup
[16:28:26 CET] <brian__> there's no sound for like 0,2s
[16:28:33 CET] <brian__> :/
[16:33:37 CET] <Nacht> You can use "-filter_complex "showwavespic=s=640x120" -frames:v 1" to plot the audio. I'm betting you'll see a gap
[16:35:15 CET] <brian__> there are gaps ye
[16:36:20 CET] <Nacht> I was running into the same issues when I wanted to take a series of TS files. Trim the first and last segment, and concat the whole together. It caused a small gap in the output file at the point where the first trimmed segment flowed into the original 2nd one
[16:36:29 CET] <brian__> Nacht alexpigment : i used this to create the individual video's & soundtracks https://gist.github.com/Brianvdb/7e634b4b1141079f28b5455a8295310d
[16:36:49 CET] <Nacht> I've read a bit that AAC Audio Priming might be the cause, but I haven't been able to confirm it
[16:37:11 CET] <brian__> the commands in that gist already have gaps inside them audio wise at start and end
[16:38:36 CET] <brian__> ye audio priming seems to be the problem
[16:39:11 CET] <brian__> https://developer.apple.com/library/content/documentation/QuickTime/QTFF/art/aac_encoded_bit_stream.jpg
[16:39:16 CET] <brian__> exactly that
[16:43:23 CET] <brian__> https://stackoverflow.com/questions/46577437/ffmpeg-concat-drops-audio-frames seems to solve that problem
[16:43:32 CET] <brian__> but i would have to re-encode :(
[16:48:18 CET] <gh0st3d> Hey everyone, I've put together a command that pipes images from phantomjs to create an mp4 (with silent audio track) ... trying to recreate the same line but to produce a webm and I'm not sure how to handle the audio, or if  I did the video right at all... here's my two lines: https://pastebin.com/gsPaQrQ6
[17:11:54 CET] <alexpigment> gh0st3d: the audio part to generate the silent audio is the same
[17:12:11 CET] <alexpigment> all you need to do is to change the output extension to .webm instead of mp4 like you have it now
[17:12:49 CET] <furq> it's yuv420p, not 240p
[17:13:11 CET] <alexpigment> the audio codec should be automatically chosen based on the file extension, i believe. if not, just specify -c:a libvorbis
[17:13:16 CET] <alexpigment> also yes, what furq said too :)
[17:13:45 CET] <alexpigment> scratch what i just said -c:a vorbis is the codec
[17:15:49 CET] <furq> libvorbis
[17:15:57 CET] <furq> vorbis is the internal encoder which is no good
[17:16:08 CET] <furq> generally just don't set the codecs for mp4 and webm though, the defaults are what you want
[17:16:39 CET] <alexpigment> vorbis is still the default for webm, right?
[17:16:44 CET] <alexpigment> or did that get switched to opus?
[17:16:44 CET] <furq> opus is the default for webm
[17:16:54 CET] <furq> vp9 and opus
[17:16:58 CET] <alexpigment> hmmm
[17:17:02 CET] <alexpigment> i guess that's a valid choice
[17:17:16 CET] <alexpigment> although i only ever use webm for html5 compatibility with windows xp
[17:17:35 CET] <furq> does that not support opus
[17:17:36 CET] <alexpigment> in which case, i'd rather have it be vp8/vorbis
[17:17:39 CET] <alexpigment> my two cents
[17:17:58 CET] <alexpigment> oh, it probably supports it from a software standpoint
[17:18:14 CET] <alexpigment> it's just that anything running xp is probably old and sucks
[17:18:44 CET] <alexpigment> so the decoding complexity of vp8 vs vp9 and vorbis vs opus seems kinda important
[17:18:49 CET] <furq> vp8, sure
[17:20:11 CET] <alexpigment> on the other hand, i do realize that some people just encode to webm only because of licensing or just because they prefer open source stuff
[17:20:23 CET] <alexpigment> so they probably want vp9 and vorbis for quality
[17:22:18 CET] <gh0st3d> tbh it's all over my head, trying to get a semi-complicated set up running to create a personalized video clip and attach it to existing video clips
[17:22:46 CET] <gh0st3d> All of the settings have to match between the videos so that I can merge the videos in the fastest way (although I don't even know if that works at all for webm files)
[17:23:14 CET] <gh0st3d> I had everything working fine for mp4 but for cross browser compatibility I have to get it working with webm files too
[17:24:30 CET] <alexpigment> gh0st3d: for cross-browser compatibility, you do realize it's *only* firefox on windows xp that needs webm at this point, right?
[17:25:35 CET] <alexpigment> so if your target audience is younger, you can kinda accept the small chance that it's an issue. hell, anyone using a non-IE web browser on xp at this point is being told daily that their shit is unsupported
[17:26:15 CET] <gh0st3d> It's real estate agents so most of them are older... But I thought it looked like for most browsers only the newest versions were playing mp4s
[17:26:27 CET] <gh0st3d> But... that would be amazing.
[17:27:56 CET] <furq> gh0st3d: if you looked on caniuse then the chart format they use is wonky
[17:28:12 CET] <furq> view by "date relative" and it's much more obvious that everything has supported it for ages
[17:28:14 CET] <gh0st3d> Ah, the wiki chart shows IE support back to 9, but caniuse just shows 11 which made me think it wouldn't work in ie 9/10
[17:28:45 CET] <gh0st3d> yeah that's much better. Screw webm.
[17:30:31 CET] <furq> the other advantage of vp9 is that it's license fee free
[17:30:40 CET] <furq> which is a bit ruined by the fact that you need an mp4 fallback anyway
[17:31:14 CET] <furq> i guess also it's more efficient than h264 but not enough to be bothered about
[17:31:17 CET] <alexpigment> well, if it's a web-based service, i don't think it matters
[17:31:35 CET] <alexpigment> i.e. if the encoding is done on the server side, i don't think you need to license x264
[17:31:39 CET] <alexpigment> rather
[17:31:42 CET] <alexpigment> h.264
[17:32:00 CET] <alexpigment> same with aac as i recall
[17:32:51 CET] <furq> i'm too lazy to remember if you actually need a license for this type of usage but i'm pretty confident it'd be free even if you need one
[17:33:08 CET] <furq> and obviously if you're not in the US hten it's not even worth thinking about
[17:33:55 CET] <alexpigment> well, h264 has a minimum-unit license that most companies don't even hit
[17:34:01 CET] <alexpigment> but again, server side it doesn't matter
[17:34:16 CET] <alexpigment> aac licensing is a pay-per-unit license
[17:34:55 CET] <alexpigment> and i think you don't need to license unless an end user is provided with an encoder or decoder
[17:35:22 CET] <alexpigment> of course, i am not a lawyer, but i've been around the block with all this licensing stuff before
[17:35:42 CET] <gh0st3d> ayiyi. looking at their page about licensing now. more headaches.
[17:36:18 CET] <alexpigment> they did just change their rules just recently
[17:36:29 CET] <alexpigment> via licensing, i mean
[17:36:36 CET] <alexpigment> *Via
[17:36:48 CET] <alexpigment> so like a lot of countries are exempt
[17:37:07 CET] <furq> i wouldn't worry about it unless you charge for this service
[17:37:10 CET] <alexpigment> MPEG-2 AAC is now exempt but MPEG-4 AAC isn't
[17:37:30 CET] <alexpigment> that's also a fair point
[17:37:35 CET] <alexpigment> let's just assume you're good to go
[17:37:37 CET] <alexpigment> ;)
[17:37:56 CET] <gh0st3d> I think it looks like we're in the clear
[17:38:28 CET] <furq> last i checked you didn't need an h264 license for delivering h264 video if you don't charge for it
[17:38:38 CET] <furq> and if you do there's a minimum unit count before you have to pay
[17:38:58 CET] <furq> and i don't think there has ever been a fee for delivering aac, the licenses are just for encoders and decoders
[17:39:20 CET] <gh0st3d> Hmm, well we're charging for a "higher tier" of a membership. They will get personalized clips attached to our existing content clips and then they can post it to Facebook / viewers can watch it for free on their personalized sites
[17:39:42 CET] <furq> 16:37:10 ( alexpigment) MPEG-2 AAC is now exempt but MPEG-4 AAC isn't
[17:39:51 CET] <furq> by mpeg-2 do you mean all lc-aac or just adts
[17:39:57 CET] <furq> if it's the latter then that's really weird
[17:40:03 CET] <alexpigment> lc-aac
[17:40:08 CET] <furq> ok that makes more sense
[17:40:20 CET] <furq> i guess that's good to know
[17:40:56 CET] <alexpigment> i've looked into the difference between MPEG-2 and MPEG-4 LC-AAC many times now, and it's so insignificant that i always forget what the difference is
[17:41:11 CET] <alexpigment> it's like one feature is unavailable
[17:41:40 CET] <alexpigment> perceptual noise substituion, according to my notes
[17:41:51 CET] <furq> PNS and LTP according to what i just looked up
[17:43:21 CET] <alexpigment> gotcha
[17:43:40 CET] <alexpigment> either way, neither have a very noticeable affect on quality at normal bitrates
[17:44:09 CET] <furq> i assume that's just as a result of all the mpeg-2 patents having expired now
[17:44:38 CET] <alexpigment> yeah, the old DVD-era stuff has all been ending lately
[17:44:42 CET] <alexpigment> AC-3 just ended
[17:44:50 CET] <furq> the last mpeg-2 patent expired last month apparently
[17:44:54 CET] <alexpigment> MPEG-2 stuff mostly expired a year or two ago
[17:44:56 CET] <furq> you'd think there'd have been more fanfare
[17:45:44 CET] <alexpigment> maybe ATSC broadcasters are happy about it
[17:45:56 CET] <alexpigment> although again, most of that cost reduction already happened
[17:46:22 CET] <furq> dvb-t still uses it as well
[17:46:46 CET] <alexpigment> oh right, non-t2
[17:46:52 CET] <furq> but most people broadcasting that are using it alongside t2 which probably already cost way more
[17:47:16 CET] <furq> especially the poor suckers that are using hevc
[17:47:47 CET] <alexpigment> i was surprised that anyone was using it yet, but it's been proven to me that people definitely are in europe
[17:47:56 CET] <furq> yeah i know cz is using hevc
[17:48:05 CET] <furq> i assume a few other smaller countries are
[17:48:28 CET] <furq> i assume their set top boxes all cost 1 million dollars
[17:48:36 CET] <alexpigment> haha
[17:48:52 CET] <alexpigment> maybe to save costs the boxes do hevc but not 4k ;)
[17:49:09 CET] <alexpigment> in which case, these cost less, but they'll need to re-buy them in a few years
[17:49:36 CET] <ritsuka> you can find a dvb-t2 receiver with hevc for less than 25 $
[17:50:21 CET] <alexpigment> i assume it's basically just a SoC?
[17:50:25 CET] <furq> i guess in europe they just don't bother paying the fee
[17:50:32 CET] <alexpigment> with a hardware decoder chip?
[17:50:37 CET] <furq> as long as there's no american components in the box i don't think they have to
[17:51:35 CET] <ritsuka> yes
[17:52:11 CET] <ritsuka> they are cheap, and anyway recent tv already have it
[17:52:41 CET] <furq> i'm happy enough with my h264 dvb-t2
[17:52:49 CET] <furq> it's not like 4k broadcast is happening any time soon
[17:53:24 CET] <alexpigment> i'd be happy with all of the current (older) standards if the cable companies would just use the fucking bandwidth allotted to them
[17:54:22 CET] <alexpigment> what's the point of an MPEG-2 channel at 38mbps if you're only getting like 8-10mbps?
[17:54:50 CET] <sfan5> >cable tv
[17:54:51 CET] <alexpigment> if they switched to H.264, i guarantee you they'd lower the bitrate to 4-5mbps
[17:54:52 CET] <sfan5> there's your problem
[17:55:02 CET] <alexpigment> of course that's my problem :)
[17:55:33 CET] <alexpigment> waiting for google fiber to just come, you know, a quarter of a mile north into the main part of the city
[17:56:20 CET] <sfan5> satellite tv is not really a thing in the US, correct?
[17:56:52 CET] <alexpigment> it's fairly big, but it just requires that you have two television-capable services to get both internet and tv
[17:57:14 CET] <sfan5> huh
[17:57:15 CET] <alexpigment> and so most people don't want to deal with the hassle or re-wiring to support it
[17:57:21 CET] <sfan5> I've only ever heard of people having cable tv
[17:57:24 CET] <sfan5> (and complaining about it)
[17:57:41 CET] <alexpigment> well, you need cable or something equivalent for internet
[17:57:51 CET] <alexpigment> and the satellite you get for the tv
[17:57:58 CET] <sfan5> DSL :^)
[17:58:01 CET] <alexpigment> there are other options, but that is the ideal scenario
[17:58:22 CET] <alexpigment> and then you have to make sure that one of the lines coming into your place is from the cable company and the others are for the satellite company
[17:59:13 CET] <alexpigment> but fiber is more than able to exceed the quality of satellite
[17:59:22 CET] <sfan5> sure, if it's available
[17:59:23 CET] <alexpigment> and cable would be too if they even gave a shit about quality
[17:59:32 CET] <alexpigment> which none do
[17:59:50 CET] <alexpigment> it's all monopolies with cable. no need to improve
[18:01:54 CET] <sfan5> you only have you(r politicians) to blame for that
[18:02:23 CET] <alexpigment> i mean that's anecdotal
[18:02:35 CET] <alexpigment> i had no choice in the matter, and i will not accept indirect blame for it
[18:39:05 CET] <leonardo98> exit
[20:25:54 CET] <calle> When changing from one container to another does ffmpeg use FALLOCATE?
[20:26:21 CET] <julesverne> bit of rage going in the ffmpeg email list.
[20:26:28 CET] <durandal_1707> calle: what's that?
[20:27:29 CET] <calle> Say i have a .mov and i want to make it in to a .mp4 the video codec is the same i just want to chnage the container then most of the file should stay the same just the container parts need to be chnaged
[20:27:55 CET] <DHE> it doesn't work that way...
[20:28:15 CET] <calle> durandal_1707: http://man7.org/linux/man-pages/man2/fallocate.2.html a function that allowes one to add and remove sections to a file.
[20:30:09 CET] <calle> DHE: so ffmpeg does not work like that? But in theory it should work right? unless the video codec is filled with escape sequences somehow.
[20:30:10 CET] <julesverne> I know I've simply renamed the extension before from .mod to mpeg before. Can't remember if i've done that with .mov though.
[20:30:19 CET] <durandal_1707> calle: no such function is used in codebase
[20:30:35 CET] <durandal_1707> julesverne: ignorance is bliss
[20:30:57 CET] <julesverne> that's my motto.
[20:31:18 CET] <calle> julesverne: your media player probaly just ignored your file ending and made it work anyway.
[20:32:06 CET] <julesverne> i saw that the codec was the same so I gave it a shot. it worked.
[20:33:31 CET] <julesverne> I'm not foolish enough believe that is the norm, but I was curious.
[20:35:54 CET] <calle> like: ffmpeg -i "file.mkv" -vcodec copy -acodec copy -aq 200 "file.mp4"
[20:38:36 CET] <calle> That thakes quite some time to execute but the amount of acctual stuff that needs to chnage in the original file is small. Probably less than 1%.
[20:39:05 CET] <sfan5> tip: -aq 200 is pointless
[20:39:50 CET] <DHE> it should be fast. spinning disks seeking around might slow it down but that should be the extent of it
[20:39:58 CET] <calle> I figured that most of the time would be spent reading and writing video data that would be excatly the same.
[20:40:25 CET] <julesverne> i'd guess roughly the same amount of time as copy\paste.
[20:40:42 CET] <calle> Um i changed the line before i posted it.
[20:41:03 CET] <calle> It used to be: ffmpeg -i "$ARG" -vcodec copy -acodec ${Codec} -aq 200 ${MetaAtFront} "$ARG.mp4"
[20:41:36 CET] <calle> but somtimes ${Codec} = copy
[20:42:48 CET] <DHE> well if you specify an audio codec then you wil transcode the audio. depending on codec it takes 1/20th the length of the video to process (eg: 1 hour video would take 3 minutes to transcode)
[20:43:12 CET] <DHE> in full copy mode it's usually the disks that limit speed
[20:45:01 CET] <calle> DHE: Yes i can see how it would be a minor point from the perspective of the entire ffmpeg project to care about the specific case when one only alters the container.
[20:47:22 CET] <DHE> well you can't edit the file in-place. there does need to be some hard copying.
[20:49:24 CET] <calle> DHE: Mabye that could be fixed with some sort of copy on write feature of the os.
[20:52:05 CET] <sfan5> that doesn't really help since most likely the same chunk of data will be at a different location
[20:52:23 CET] <sfan5> so it would not be any better than creating a file from scratch
[20:54:13 CET] <calle> sfan5: but that would be the point of fallocate.
[20:55:02 CET] <DHE> there is no feature in any OS that I know of to insert 100 bytes into the middle of a file quickly
[20:55:13 CET] <furq> different containers pack data differently anyway
[20:55:33 CET] <DHE> fallocate() is to quickly ensure space is reserved for applications that need to make large files that are not intended to be sparse
[20:55:48 CET] <furq> one obvious example being avcc/annex-b h264 in mp4/mpegts
[20:55:57 CET] <furq> but also just the way the streams are interlaved
[20:56:05 CET] <furq> interleaved
[20:56:09 CET] <DHE> or just mpegts which chunks its data into 184 byte segments with a small 4-byte header between each
[20:56:14 CET] <DHE> that'd be fun
[20:56:22 CET] <calle> DHE: No it is more nuanced than that: http://man7.org/linux/man-pages/man2/fallocate.2.html
[20:56:40 CET] <calle> look at the section: Increasing file space
[20:57:35 CET] <sfan5> that sounds very bothersome for barely any benefit
[20:57:58 CET] <DHE> "This mode has the same limitations as FALLOC_FL_COLLAPSE_RANGE"...
[20:58:17 CET] <DHE> "Typically, offset and len must be a multiple of the filesystem logical block size"  which is typically 4096 bytes these days
[20:58:38 CET] <calle> DHE: aha yes that makes sense.
[20:58:58 CET] <DHE> plus that's going to be hell on file fragmentation
[20:59:11 CET] <DHE> so at the end of the day, I'm still going to suggesting you just do it normally
[21:02:33 CET] <calle> I guess what im am acctualy after is reducing the amount of writes to disk. (I am sort of obbsesed with that)
[21:06:38 CET] <c_14> do it in a tmpfs?
[21:07:03 CET] <DHE> or a network share?
[21:08:48 CET] <dragm__> Anyone know the difference between CRF constrained mode vs 2 pass vbr with the same max peak bitrate, if one looke at it from a quality perspective ?
[21:10:15 CET] <calle> c_14: yes that is probably the most reasonable solution that i could do :) But i started spawning about using fuse to "reencode on the fly", completly insane but not imposible.
[21:10:48 CET] <c_14> mp3fs
[21:10:57 CET] <c_14> can probably fork it to work with videos
[21:15:54 CET] <furq> dragm__: they're identical
[21:16:35 CET] <furq> 2-pass vbr is functionally just crf with the crf value predetermined by the first pass
[21:16:54 CET] <dystopia_> it's also sloooow
[21:16:59 CET] <furq> yeah
[21:17:17 CET] <furq> so it's good that it doesn't look any nicer
[21:17:25 CET] <furq> it's only there if you absolutely must hit a target size
[21:23:05 CET] <Soni> how do I speed up ffmpeg?
[21:23:23 CET] <Soni> or rather uh
[21:23:36 CET] <Soni> I have this 700MB video file
[21:23:49 CET] <Soni> how do I compress it, without changing things like framerate and stuff?
[21:25:49 CET] <BtbN> don't specify any new framerate
[21:32:04 CET] <Soni> input says 24.54 fps but it wants to output 24.58 fps
[21:32:26 CET] <Soni> BtbN: this is with not specifying framerate
[21:32:48 CET] <BtbN> did the duration change?
[21:33:26 CET] <BtbN> With some containers, those are all just guesses though
[21:33:31 CET] <Soni> BtbN: how do I view that?
[21:33:46 CET] <BtbN> ask ffprobe
[21:33:53 CET] <Soni> it's too slow for me to do it like this, can I enable hwaccel?
[21:34:10 CET] <BtbN> ffprobe without any furhter parameters should finish instantly
[21:35:45 CET] <Soni> that won't give me the stuff for the output, will it?
[21:35:56 CET] <Soni> unless I already have the output
[21:35:58 CET] <Soni> which I don't, yet
[21:39:24 CET] <Soni> guess I'm just gonna have to wait :/
[22:10:48 CET] <Soni> ok so it finished, but it's 400MB :/
[22:11:06 CET] <Soni> I need it smaller... 720p?
[22:21:38 CET] <alexpigment> Soni: do you care about quality?
[22:22:15 CET] <Soni> alexpigment: the original has quality
[22:22:29 CET] <Soni> mainly... what's the scale for 720p?
[22:22:29 CET] <alexpigment> dropping resolution is an ok way to decrease file size (with the right parameters)
[22:22:39 CET] <alexpigment> but you can always drop bitrate or increase the CRF value
[22:22:51 CET] <alexpigment> what do you mean by "scale"/
[22:23:43 CET] <Soni> how do I bring the 1080p 700MB file down to 720p?
[22:23:49 CET] <alexpigment> oh
[22:23:54 CET] <alexpigment> just do -s 1280x720
[22:24:31 CET] <alexpigment> not that if you're specifying a bitrate explicitly, the file size won't change much
[22:24:41 CET] <Soni> ah, 1280, thanks
[22:24:46 CET] <alexpigment> but at the same CRF value, the 1280x720 should be a smaller file size
[22:25:19 CET] <Soni> alexpigment: it changed by about 50% without any arguments (just ffmpeg -i in out)
[22:25:33 CET] <alexpigment> ok so you're just using the implied CRF 23 then
[22:26:18 CET] <Soni> so, y'know, 50%, that's a lot isn't it?
[22:26:31 CET] <alexpigment> well, 23 is pretty lossly
[22:26:34 CET] <alexpigment> *lossy
[22:27:32 CET] <Soni> what do you recommend for a 5 minute 700MB 1080p mp4 -> <something here>?
[22:27:52 CET] <alexpigment> i mean you're asking the wrong person
[22:28:02 CET] <alexpigment> i'd personally recommend that you keep the original 700MB 1080p mp4
[22:28:16 CET] <alexpigment> save your time, save your quality
[22:28:19 CET] <Soni> alexpigment: oh I'm keeping it
[22:28:30 CET] <Soni> but I mean for sharing
[22:28:41 CET] <alexpigment> like on dropbox or something?
[22:28:42 CET] <alexpigment> youtube?
[22:28:50 CET] <Soni> uh I was thinking torrents
[22:28:55 CET] <alexpigment> oh gotcha
[22:28:58 CET] <Soni> encrypted torrents
[22:29:05 CET] <alexpigment> well, it's up to you
[22:29:14 CET] <Soni> I don't like everyone having the keys to my porn y'know
[22:29:34 CET] <Soni> and I think encrypted torrents are pretty good for that
[22:29:43 CET] <alexpigment> but i know that i'd appreciate getting a file that i knew someone re-encoded randomly just to get the file size down to an arbritrary number
[22:29:49 CET] <alexpigment> *arbitrary
[22:30:11 CET] <alexpigment> gah. typing is hard
[22:30:18 CET] <sfan5> >encrypted torrents
[22:30:26 CET] <Soni> who waits 6 hours to be able to watch some 5 minute porn
[22:30:32 CET] <alexpigment> i'm saying i would appareciate *not* getting a file that was reencoded arbitrarily
[22:30:36 CET] <Soni> sfan5: the files are encrypted, not the torrents
[22:31:34 CET] <alexpigment> ok, i guess if it takes 6 hours to get 700MB, that is a concern
[22:32:25 CET] <Soni> altho I could probably seed it from my phone as well, that would speed it up slightly
[22:32:45 CET] <alexpigment> this all sounds like a very weird setup, but i'll trust there's a reason for all this
[22:33:17 CET] <Soni> oh I'm trying to get it to one person specifically I guess
[22:33:17 CET] <sfan5> just re-encode it to hevc with low settings + low-bitrate opus audio
[22:33:40 CET] <sfan5> the quality will probably drop to a "watchable on phone" level
[22:33:54 CET] <alexpigment> yeah, except if your phone doesn't have HEVC hardware decoding ;)
[22:34:02 CET] <alexpigment> or doesn't support opus
[22:34:07 CET] <sfan5> mpv-android :^)
[22:34:13 CET] <alexpigment> <- sheep
[22:34:24 CET] <sfan5> also if you scale it down somewhat, you won't even need hwdec
[22:34:28 CET] <alexpigment> hell, i don't even know if my phone supports HEVC natively
[22:35:24 CET] <alexpigment> i guess my phone does do hevc in hardware. that's pretty cool i guess?
[00:00:00 CET] --- Wed Mar 14 2018


More information about the Ffmpeg-devel-irc mailing list