[00:00:08 CEST] Action: Nolski begins googling for acronyms [00:00:34 CEST] <klaxa> blue green red alpha [00:00:45 CEST] <klaxa> the second video is in rgb color space while the others are in yuv color space [00:00:55 CEST] <klaxa> i'm guessing that's the culprit [00:01:08 CEST] <c_14> set -pix_fmt to yuv422p when encoding that second one (or reencode to yuv422p) [00:03:42 CEST] <Nolski> Hey it worked! [00:04:10 CEST] <Nolski> That's so strange, why would it change the color space when the encoding is manually set? [00:08:10 CEST] <KarlFranz> What does the following message mean? "Non-monotonous DTS in output stream 0:0; previous: 70605251, current: 70603200; changing to 70605252. This may result in incorrect timestamps in the output file." <- When encoding an *ac3 dump from a dvd to vorbis. [00:09:03 CEST] <Nolski> KarlFranz: I'm no expert but I think it's just how the timestamps differ between encodings and sometimes it doesn't line up perfectly. I seee it all the time when encoding from one format to another [00:09:15 CEST] <Nolski> *citation needed* [00:09:42 CEST] <KarlFranz> I hope it does not cause desyncing. [00:10:09 CEST] <loop3r_> how to find out which values are k1 and k2 in matrixes http://www.ffmpeg.org/ffmpeg-all.html#lenscorrection ? [00:10:10 CEST] <KarlFranz> I am still encoding the video stream that goes with that sound, I will check when I mux it all together,. [00:13:18 CEST] <ocrete> how am I supposed to know the final format of the decoded picture when get_buffer2() is called? for example, I can't find out of the stream is interlaced or not before a frame comes out? [00:37:01 CEST] <edoceo> I've got a video that reports 10738 hours long, it's not. I process through ffmpeg to re-transcode but it still shows this errant time [00:39:51 CEST] <DHE> probably has corrupted timestamps then [00:40:07 CEST] <DHE> never seen THAT high before. 18 hours sure... [00:41:31 CEST] <c_14> Maybe try genpts? [00:42:33 CEST] <chungy> I always see that with VOB dumps from DVDs, haha... -fflags +genpts fixes it right up. Dunno if it'll help in your case. [00:46:33 CEST] <edoceo> I'll know in 8 minutes... [00:55:26 CEST] <edoceo> Dang; that did not fix my issue. Other things I could try? [00:55:44 CEST] <c_14> Input option? Output option? Both? [00:57:46 CEST] <edoceo> http://edoceo.io/paste?p=kfvYF5 [00:59:24 CEST] <chungy> move the "-fflags +genpts" to before the -i option. (and you can use "-c copy" for output to avoid re-encoding, btw) [01:00:23 CEST] <edoceo> I can't -c copy this stream cause the source audio codec is opus which confuses a down-stream consumer of this media [01:00:40 CEST] <edoceo> I'm trying the flag before -i [01:00:42 CEST] <chungy> -c:v copy then [01:00:51 CEST] <chungy> if the input and output are both VP8 anyway... heh [01:02:11 CEST] <edoceo> Our source video is not the best quality. Like this time stamp issue, among others. Before entering our processing engine I do this re-transcode to help it clean up [01:02:28 CEST] <edoceo> Otherwise the -filter_complex grid making and overlay and amix all goes to sh.. [01:04:37 CEST] <edoceo> This ffmpeg 2.2.x version works as expected, output duration is 27 minutes: http://edoceo.io/paste?p=5yWRCp [01:04:58 CEST] <c_14> Wait, is this a regression? [01:05:27 CEST] <chungy> If it works with 2.2, then yeah, it's a regression [01:05:44 CEST] <c_14> If it's a regression, open a bug on trac. [01:08:10 CEST] <edoceo> Hm, I keep getting a bunch of Captcha - and SpamBayes determined spam probability of 61.91% [01:09:31 CEST] <edoceo> Well, account is created now. I gotta do a thing, and I'll fiddle with -fflags a bit, get some repro and -report output [01:10:02 CEST] <edoceo> Well, -fflags didn't work in front of -i either. [01:10:39 CEST] <edoceo> If I upload videos to the bug tracker FTP those are not public right? And get removed/not-shared? Our sample video might be a little private [01:12:54 CEST] <c_14> They're usually attached to the bug report. Can you reproduce the issue with a small portion of the video? [01:13:48 CEST] <llogan> IIRC, the files uploaded to FTP aren't viewable by the public [01:25:47 CEST] <Nolski> I must not understand colors correctly. I'm using a simple drawtext filter but it's giving me this cannot find color error for '00)' but that isn't anywhere in the filter. [01:25:49 CEST] <Nolski> https://gist.github.com/Nolski/cee93d5c47745170057a [01:27:01 CEST] <c_14> You need to escape the ':' in the between time code thingies [01:27:37 CEST] <c_14> The last error from the drawtext filter there [01:37:02 CEST] <coops> is there a specifcation for the FFM file format anywhere? [01:42:41 CEST] <Nolski> ah right. thanks :) [03:34:53 CEST] <nardev> guys is there any quick example hoe to scale down videos only if with resolution is more than 1024 for example? [03:50:34 CEST] <c_14> you could probably just -vf scale=max(1024,x):-2 or something [04:04:27 CEST] <nardev> c_14, thank you [05:07:42 CEST] <Renari> Would anyone be willing to look at this https://gist.github.com/Renari/409f415994e9dd94930c it seems I'm having timing issues with seeking even though I'm putting -ss after the input file. In the above command intermediate2.ts is completely absent from the output file. [05:09:37 CEST] <Renari> I should mention that the output for intermediat2.ts is a 0 byte file. [05:12:29 CEST] <Renari> Is it possible to specify frame numbers when seeking instead of times? I think that might fix my issue. [05:12:56 CEST] <pzich> I think -seek and -vframes can do that? but I'm no expert [05:13:57 CEST] <pzich> nope, I'm wrong [05:14:08 CEST] <Renari> :( [05:14:37 CEST] <Renari> I don't understand why a 4 second time-window of the video is being output as a 0 byte file. [05:15:02 CEST] <pzich> anything else in the ffmpeg output for that one? does it work with 5s? [05:15:24 CEST] <pzich> oh [05:15:36 CEST] <pzich> I think the problem is that you're using -c copy [05:15:56 CEST] <Renari> It works with ts1 and ts3 though. [05:16:01 CEST] <pzich> and if your keyframes aren't where your boundaries are, it can't just copy [05:16:08 CEST] <Renari> Ah [05:16:18 CEST] <Renari> that would explain the weird timing issues I've had with ts1 and ts3 then [05:16:20 CEST] <pzich> are you sure that the other ones are exact? [05:16:38 CEST] <Renari> They are in-fact wrong based on the video timings. [05:16:50 CEST] <Renari> But work with the output. [05:18:54 CEST] <Renari> Currently trying it with a codec set. [05:19:03 CEST] <Renari> Takes a lot longer to seek though. [05:19:17 CEST] <pzich> well it has to actually encode it this time :D [05:19:56 CEST] <Renari> yup that fixed it [05:20:39 CEST] <Renari> Oh, thanks. [05:20:48 CEST] <pzich> no problem :) [12:24:23 CEST] <lilibox> hi [12:24:57 CEST] <sagax> hi! [12:25:47 CEST] <lilibox> does exist any script allows to build ffmpeg static version here? [12:26:13 CEST] <lilibox> i kept stuck on centos5 and need use the latest ffmpeg [12:26:38 CEST] <sagax> get ffmpeg archive [12:26:43 CEST] <sagax> unpack [12:26:53 CEST] <sagax> run configure in unpack dir [12:27:06 CEST] <sagax> run make and make install [12:27:11 CEST] <lilibox> with some "fork" ffmpeg already installed on system [12:27:27 CEST] <sagax> fork - removing [12:28:00 CEST] <lilibox> i am not sure if it is possible, it is used for something different [12:28:39 CEST] <sagax> try run removed and show different [12:28:43 CEST] <lilibox> soo my guess is use - create the latest ffmpeg static build and call it within full path [12:29:36 CEST] <sagax> you might configure with custom path for your self [12:30:31 CEST] <lilibox> yes, and which is the best tutorial how to create the latest ffmpeg build? any official how to? [12:30:46 CEST] <lilibox> i need static of course [12:31:31 CEST] <lilibox> mxf and h264 with sound is ok for me, at least pcm [12:31:35 CEST] <chungy> Might look at Nix and see if it runs on CentOS 5. Your biggest hurdle here is actually probably going to be the kernel; the default is 2.6.18, very ancient at this point in time. Just for Nix or other static builds to work, I wouldn't be surprised if you will be required to upgrade the kernel (which isn't too difficult in most cases) [12:32:13 CEST] <chungy> beyond that, the ./configure script is fairly self-explanatory I'd think, it has an option for static builds [12:34:54 CEST] <lilibox> chungy yes, no way to move with kernel, i must be able to do it at current state of os [12:36:42 CEST] <lilibox> ad ./configure did you speak about nix's configuration or about ffmpeg's general configuration? [12:36:44 CEST] <chungy> I would give nix a spin, but i don't know if it'll run. If it does, it's pretty easy to get going (and they probably have a pre-built ffmpeg for your needs) [12:36:51 CEST] <chungy> ffmpeg's [12:38:55 CEST] <lilibox> i must be sure that all codecs are packed in one statics binary, i am not sure if i am able to avoid situation part of my "statics" ffmpeg will call some dynamics part of existing ffmpeg which will probably criple output [12:39:29 CEST] <lilibox> that is reason i am looking for script which is able build everything iself [13:24:29 CEST] <hefest> so, when I do this: ffmpeg -y -i 2rotated.mp4 -crf 20 -vf 'transpose=2,transpose=2' -metadata:s:v:0 rotate=0 output.mp4 [13:24:34 CEST] <hefest> fps increases to 120fps [13:26:03 CEST] <hefest> original file fps is 29fps and the output.mp4 has 120fps , [13:27:21 CEST] <hefest> any way to fix this except -r 29 ? [14:04:12 CEST] <rimdeker> Any ffmpeg devs here? Just wanted to say that you guys do a hell of a job. [14:07:10 CEST] <durandal_170> most devs are on another channel [15:32:53 CEST] <mathieu> Hi, I want to copy & modify the H264 decoder [15:33:21 CEST] <mathieu> How would I make this work & include in ffmpeg? [15:33:48 CEST] <mathieu> I'd like to keep the avdec_h264 but add the modified one [15:34:51 CEST] <mathieu> is it possible to have 2 decoders simultaneously? [15:42:12 CEST] <DHE> you can have multiple decoders for the same codec. such a thing already exists for VDPAU vs software decoders [16:14:35 CEST] <well0ne> Canar you there? [16:17:43 CEST] <well0ne> here is my php solution, to my "background-process" dilemma http://privatepaste.com/58263bc365 [17:48:03 CEST] <ocrete> how am I supposed to know the final format of the decoded picture when get_buffer2() is called? for example, I can't find out of the stream is interlaced or not before a frame comes out? [18:04:15 CEST] <AndrewMock> My brain is fried. How do I apply -vf "scale=1280:533" AFTER -vf "crop=1920:800:0:140" ? [18:04:25 CEST] <AndrewMock> i know use commas somewhere [18:04:29 CEST] <AndrewMock> i just need sleep [18:08:36 CEST] <AndrewMock> makes sense now i think [18:12:46 CEST] <AndrewMock> just comma seperation unless ambigious [18:12:49 CEST] <AndrewMock> k [18:22:51 CEST] <Guest90758> I found why my command wasn't working... What was it that HEVC resolutions had to be a multiple of again? [18:22:54 CEST] <Guest90758> 4? 8? [18:23:19 CEST] <JEEB> that depends on your chroma subsampling [18:23:25 CEST] <JEEB> the format itself has no other limitations [18:24:01 CEST] <JEEB> 4:2:0 subsampling would be mod2 width and height, 4:4:4 would be mod1, 4:2:2 would be mod2 width, mod1 height [18:26:53 CEST] <Guest90758> i know what modulus is.. is that what you mean? [18:27:05 CEST] <JEEB> yes [18:27:09 CEST] <Guest90758> mod1 vs mod2 though? [18:27:16 CEST] <BtbN> h264 has a restriction of it beeing a multiple of 16 [18:27:26 CEST] <JEEB> BtbN: yes but that is of the coded region [18:27:38 CEST] <JEEB> the picture can be of whatever depending on the chroma subsampling [18:27:51 CEST] <Mavrik> Guest90758, he means that 4:2:0 has to be divisible by 2 and 4:4:4 divisible by 1 [18:27:53 CEST] <Mavrik> :) [18:28:15 CEST] <Guest90758> oh ok thx [18:28:37 CEST] <JEEB> and with HEVC the CTU can be at most 64x64, so it really doesn't make any sense to mention the coding block sizes [18:28:47 CEST] <JEEB> esp. since the padding has very close to zero cost [18:29:06 CEST] <Guest90758> so i changed my scaling to 1280x534 instead of 1280x533 and it worked... thx! [18:29:36 CEST] <JEEB> there are ways to automatize your scaling to be always within the limitations of your chroma subsampling [18:29:44 CEST] <JEEB> see the trac wiki on scaling [18:29:46 CEST] <loop3r> hello, have someone worked with ffmpeg lenscorrection I want to ask about little thing :P [18:29:50 CEST] <JEEB> and ffmpeg-all.html [18:30:09 CEST] <Mavrik> Guest90758, scale filter has a helpful "-2" constant [18:30:14 CEST] <Mavrik> e.g. "scale=1280:-2" [18:30:24 CEST] <Guest90758> oh solid [18:33:06 CEST] <Guest90758> when my h264 decoder mumbles something like "not enough frames to estimate rate; consider increasing probesize", does that mean bitrate or framerate? [18:33:51 CEST] <Guest90758> reading between the lines this means bitrate & no effect on my output bitstream. [18:36:42 CEST] <Nolski> What's the difference between the -vf flag and -filter_complex? [18:38:08 CEST] <c_14> -vf is a filterchain, -filter_complex is a filtergraph [18:38:20 CEST] <c_14> A filtergraph can contain several filterchains [18:39:23 CEST] <Nolski> ah, I see [18:41:58 CEST] Action: Nolski sighs [18:42:22 CEST] <Nolski> why does a drawtext filter kill the audio on my video that I'm applying it to? [18:52:47 CEST] <DHE> Nolski: if you're using -map to pull in filter items, be sure to explicitly pull in audio as well. -map 0:a [18:53:26 CEST] <Nolski> DHE: but if I'm not filtering any of the audio with this command do I just reference it in the -map? [18:53:37 CEST] <DHE> filter_complex will let you build a very complicated filter graph with multiple inputs and multiple outputs [18:53:57 CEST] <DHE> you're not filtering the audio, but you are explicitly building an output from built elements. you'll need to specify audio manually in this case [18:54:07 CEST] <Nolski> Ah I see [18:54:16 CEST] Action: Nolski levels up [18:54:57 CEST] <DHE> for example I'm making multiple HLS feeds (several bitrates) from a single ffmpeg instance. audio is split several ways, video is deinterlaced once and then piped through several scale filters for each input [18:59:58 CEST] <Nolski> DHE: I think I see what you're saying. I've mixed different audio streams together just fine using the amerge filter but since this is just one video with text being rendered on it, I'm a bit confused how to just carry that one stream over. [19:00:01 CEST] <Nolski> https://gist.github.com/Nolski/d9edc88da76f204441ec#file-gistfile1-txt-L26 [19:00:38 CEST] <Nolski> That stream [0:1] is being mapped using `-map` but the output doesn't have it so it breaks? [19:04:34 CEST] <Guest90758> why does my bitrate keep going up??? lol [19:11:14 CEST] <DHE> Nolski: don't use [brackets] for file inputs [19:13:08 CEST] <Nolski> DHE: not quite sure I follow [19:13:34 CEST] <DHE> -map [videoout] -map 0:a [19:26:41 CEST] <Nolski> DHE: sorry for making you spell it out for me but prior, I was never mapping the video (or [videoout]). If I assign it in the filter graph it gives the same error as with the audio [20:17:03 CEST] <AndrewMock> [h264 @ 0000000004866780] Stream #0: not enough frames to estimate rate; consider increasing probesize [20:18:30 CEST] <AndrewMock> Does that mean that the decoder can't figure out the framerate? [20:21:27 CEST] <kepstin-laptop> I think that message is actually about bitrate. Unless you're trying to figure out the bitrate of an input file, it'll just be a minor display issue. You can probably ignore it unless you have other problems. [20:24:58 CEST] <AndrewMock> well i think that it may have something to do with my apparent timing issue http://i.imgur.com/vYbsAEe.png [20:26:15 CEST] <kepstin-laptop> AndrewMock, that 'past duration too large' thing is a fairly new message in ffmpeg; it happened before but used to not print a message. It apparently has something to do with the rate control on the output encoding [20:31:36 CEST] <loop3r> who is good with lens correction filter ?:D [20:34:21 CEST] <JEEB> the person who wrote it? [20:34:38 CEST] <JEEB> I hope you read the documentation at least [20:37:26 CEST] <AndrewMock> what is rate control distortion? [20:44:30 CEST] <loop3r> JEEB: I did, I just want to ask few question about opencv in that example thats it, calm down [20:46:35 CEST] <prelude2004c> hello everyone [20:46:57 CEST] <prelude2004c> question ... how do i setup ffmpeg to allow playback of a file in a loop ? eg.. i have 10 files.. i want to play them one by one without delay [20:47:19 CEST] <c_14> https://trac.ffmpeg.org/wiki/Concatenate [20:47:48 CEST] <kepstin-laptop> prelude2004c, that sounds like something that many players can do, when given an appropriate playlist... [20:48:50 CEST] <kepstin-laptop> but yeah, you could use ffmpeg to concatenate them into a single video if you can't just play them directly. [20:49:58 CEST] <AndrewMock> I really need help with understanding what this means: http://i.imgur.com/vYbsAEe.png [20:50:06 CEST] <AndrewMock> Shall I file a bug report? [20:50:24 CEST] <AndrewMock> At least to ask for more-understandble strings? [20:51:02 CEST] <AndrewMock> some people have murmored about framerate sync and about rate control [20:51:08 CEST] <AndrewMock> but nothing explicit [20:51:43 CEST] <prelude2004c> kepstin-laptop < .. i am trying to create a stream though [20:51:43 CEST] <JEEB> loop3r: I'm calm, I just am pretty sure that most folk have no idea on such specific filters added by someone at some point for a very specific use case [20:51:51 CEST] <prelude2004c> that si why i want ffmpeg so i can output to RTMP or hls or whatever [20:52:14 CEST] <kepstin-laptop> prelude2004c, ah, gets a bit more complicated then :) [20:52:22 CEST] <prelude2004c> lol .. well yes i guess [20:52:30 CEST] <prelude2004c> i want to just say ffmpeg -i something -i something else -i somethign else.. etc etc [20:52:34 CEST] <prelude2004c> and just play them sequentially [20:52:44 CEST] <prelude2004c> without any delay from one to the next [20:52:46 CEST] <kepstin-laptop> ffmpeg can loop a single input video, but I don't know any way to loop a list of multiple videos.... [20:53:02 CEST] <prelude2004c> yup single file is not a problem.. [20:53:06 CEST] <prelude2004c> its the multiple that is a problem for me [20:53:11 CEST] <prelude2004c> i was hopping someone knew a good way to do it [20:53:15 CEST] <kepstin-laptop> in this case, you could just combine the videos into a single file in advance, of course. [20:53:23 CEST] <prelude2004c> can some player do it ( linux ) that i can just output and | > ffmpeg [20:53:31 CEST] <c_14> you can loop an infinite concat list [20:53:35 CEST] <c_14> with some bash magic [20:53:50 CEST] <prelude2004c> how would that look like ? i don't know bash very well [20:54:09 CEST] <prelude2004c> i want to basicaly get list of files ... > output | > ffmpeg to stream local udp or something [20:54:15 CEST] <prelude2004c> or i will just output to HLS [20:54:41 CEST] <c_14> ffmpeg -f concat -i <(while true; do for file in *; do printf "file %s\n" $file; done; done) [20:54:44 CEST] <c_14> something like that [20:56:05 CEST] <prelude2004c> hum.. never used that before [20:56:18 CEST] <prelude2004c> do i pipe that out after ? [20:56:26 CEST] <kepstin-laptop> that has the caveats of using the concat input format, tho; all the files have to be in a concatenatable file format (e.g. mpeg-ts) [20:56:44 CEST] <prelude2004c> or can i just continue with -f mpegts udp://localhost:xxxx [20:56:55 CEST] <c_14> you can just continue with whatever you want [20:57:05 CEST] <prelude2004c> i dont have a problem with the same format [20:57:11 CEST] <kepstin-laptop> hmm, no, that doesn't need concatenatable formats, just codecs with matching parameters, etc. [20:57:14 CEST] <prelude2004c> would be nice if files would be mpeg4 and mkv ,etc etc [20:57:33 CEST] <kepstin-laptop> I got the different concat methods mixed up [20:57:36 CEST] <prelude2004c> so your saying i can't have one mp4 , one mkv, one wmv [20:57:59 CEST] <c_14> The video/audio settings have to be the same. [20:58:11 CEST] <prelude2004c> that may be a problem since sometimes the video may be different [20:58:19 CEST] <prelude2004c> like format or even resolution [20:58:23 CEST] <prelude2004c> some videos may be different resolutons too [21:00:21 CEST] <kepstin-laptop> then your best bet is to pre-transcode them [21:01:39 CEST] <kepstin-laptop> if you pre-transcode so everything's in matching formats, c_14's trick would work, and you might even be able to get away with -c copy instead of doing an extra live encoding. depends on the player, i guess. [21:01:48 CEST] <JEEB> loop3r: if you think that randomly pm'ing people is alright, then I will just say that no - it is not [23:05:56 CEST] <prelude2004c> hey guys.. the loop function.. any way to loop a file forever? [23:10:31 CEST] <ChocolateArmpits> Hello, Anyone familiar with parsing MXF files ? [23:42:46 CEST] <prelude2004c> anyone? i can't seem to figure this out [23:45:47 CEST] <Nolski> prelude2004c: you can just write a script that keeps looping the file. [23:46:13 CEST] <Nolski> pretty sure if it's the right format you can just keep concatting the same file upon itself [23:46:24 CEST] <Nolski> or you can just make it so it loops when it lays [23:46:28 CEST] <Nolski> s/lays/plays/ [23:49:42 CEST] <prelude2004c> hum.. dont know how to do that [23:49:48 CEST] <prelude2004c> also, maybe there is a way with cat [23:49:53 CEST] <prelude2004c> eg.. have ffmpeg listen to a port [23:50:04 CEST] <prelude2004c> and then just use a script that cat file > output [23:50:13 CEST] <prelude2004c> so i i can keep sending files to the listener [23:50:15 CEST] <prelude2004c> something like that [23:50:34 CEST] <prelude2004c> i was thinking like a local UDP port and then just cat files to that udp [23:50:42 CEST] <prelude2004c> any other ideas on how to do that ? this way i can loop any file or many files [23:51:20 CEST] <c_14> ffmpeg -f concat -i <(while true; do printf "file %s\n" /path/to/file; done) [23:52:42 CEST] <prelude2004c> what if i want to introduce multiple files [23:52:51 CEST] <prelude2004c> any way to put ffmpeg as a listner that never exists ? [23:52:57 CEST] <prelude2004c> like.. for example.. i did a mkfifo test file [23:53:04 CEST] <prelude2004c> then i did ffmpeg -i fifofile [23:53:08 CEST] <prelude2004c> and it sits there and waits [23:53:15 CEST] <prelude2004c> then i add info to that file and it loads the data... but... [23:53:21 CEST] <prelude2004c> soon as the data stops , ffmpeg exists [23:53:33 CEST] <prelude2004c> any way to keep ffmpeg listening all the time and never exist so it takes any info i give it from differnet sources [23:54:25 CEST] <c_14> You could listen on udp probably. [23:55:03 CEST] <prelude2004c> right and it never exists ? like listen 239.xx.x.x .. something liek that ? [23:55:48 CEST] <c_14> ffmpeg -i udp://ip:port?listen [23:56:17 CEST] <c_14> Worked for me (for flac anyway) [23:56:54 CEST] <c_14> That might only work for some formats though. [00:00:00 CEST] --- Thu Jul 30 2015