[Ffmpeg-devel-irc] ffmpeg.log.20180101

burek burek021 at gmail.com
Tue Jan 2 03:05:01 EET 2018

[00:16:30 CET] <Tachyon> hny
[00:16:41 CET] <kerio> HAPPY HALLOWEEN
[04:53:42 CET] <Hotpot33> well, how do I uhm.. set the video and audio to different outputs
[04:54:26 CET] <Hotpot33> like I want to extract the video into YUV files, and audio to pcm_s16le
[04:54:41 CET] <Hotpot33> BOTH in one go, since it's live
[05:02:50 CET] <DHE> ffmpeg -i ...... -map 0:v [options] video-out-file -map 0:a [options] audio-out-file
[05:03:08 CET] <Hotpot33> ah ok
[05:03:14 CET] <DHE> where "file" might be an RTMP target, literal file on disk, or whatever. usual rules
[05:16:44 CET] <utack_> relaxed yes i used that now
[05:16:47 CET] <utack_> thank you
[05:18:14 CET] <johnnny22> does ffmpeg work with libva 2.0.0 ?
[05:23:48 CET] <hiihiii> hi
[05:24:10 CET] <hiihiii> I'm trying to upload media to twitter
[05:24:25 CET] <hiihiii> my video is 45 seconds long
[05:24:54 CET] <hiihiii> this is my command
[05:26:10 CET] <hiihiii> ffmpeg -i in.flv -vsync 1 -c:v libx264 -preset slow -tune film -crf 15 -vf "fps=30,format=yuv420p" -c:a aac -ac 1 out.mp4
[05:26:52 CET] <hiihiii> but I keep getting "unable to upload" from twitter
[05:27:29 CET] <hiihiii> everything seems like it's following their encoding settings
[05:31:07 CET] <hiihiii> http://dpaste.com/228ETCP
[05:35:56 CET] <utack_> Hi atomnuker do you happen to know about aptx or were you not really part of implementing that?
[07:55:52 CET] <justswim> Hello all! Not sure if I've come to the right place, but I was wondering if anyone had any tips for adding colored emoji to a video with ffmpeg?
[07:56:27 CET] <justswim> This question on stack overflow has been unanswered for a while and I was hoping that it might be possible to solve!
[07:56:28 CET] <justswim> https://stackoverflow.com/questions/34880734/how-to-use-ffmpeg-with-drawtext-and-a-emoji-ttf
[07:59:51 CET] <justswim> This ticket has also been filed a while back without much progress: https://trac.ffmpeg.org/ticket/5777
[10:08:57 CET] <Fyr> guys, subtitles are a few seconds late. how to move them earlier when muxing with FFMPEG?
[11:07:49 CET] <atomnuker> utack_: not really part of writing it, I just reviewed it
[11:08:11 CET] <atomnuker> I don't think aptx can adjust the bitrate, its a very simple voice codec
[15:35:35 CET] <utack_> atomnuker is it really fixed 352kbit/s bitrate?
[15:36:28 CET] <utack_> that is ridiculous, using a codec with patents to make 352kbit/s stereo sound good
[15:36:49 CET] <utack_> any moron could have cooked up a patent free codec to achieve that in half a day...
[15:42:42 CET] <bencoh> :D
[15:49:00 CET] <mrsip> Hi. I am messing around with a ring doorbell, and I want to try and take a picture every 60 seconds. I have done this using ffmpeg to generate a still image for other video streams, but does ffmpeg support sip urls? or do I need to rebuild with another lib (I am using linux) or does it not support it?
[15:54:25 CET] <c_14> I don't think it does
[15:59:44 CET] <SilenceDogood> What do I do if I run ffmpeg on an avisynth script and the result has the audio out of sync?
[16:00:00 CET] <SilenceDogood> Any way I can get an exact number to use to delay the audio so it syncs back up?
[16:04:51 CET] <SilenceDogood> oh wait, using audacity, I have determined that it is 0.342 seconds. I hope that's precise enough
[16:08:22 CET] <SilenceDogood> my new command is ffmpeg -map 0:1 -itoffset 0.342 -i VTS_19_1.VOB -c:v libx264 -preset slow -crf 18 -c:a aac -b:a 192k -pix_fmt yuv420p output.mkv
[16:08:35 CET] <SilenceDogood> Anyone please tell me if anything about that looks completely stupid and wrong. I'd appreciate it.
[16:09:40 CET] <SilenceDogood> oops, typo. meant itsoffset
[16:10:28 CET] <SilenceDogood> ah, apparently that is completely wrong.
[16:17:01 CET] <SilenceDogood> new command ffmpeg -i danger.avs -itsoffset 0.342 -i danger.avs -map 0:0 -map 1:1 -c:v libx264 -preset slow -crf 18 -c:a aac -b:a 192k -pix_fmt yuv420p output.mkv
[16:35:40 CET] <furq> SilenceDogood: itsoffset only affects video streams so you probably don't need multiple inputs
[16:35:54 CET] <furq> assuming those are both supposed to be your avisynth script which seems weird if one is audio
[16:38:05 CET] <SilenceDogood> furq o rly. I wasn't aware of this. So all I'd need would be one input and no map switches?
[16:38:34 CET] <SilenceDogood> like this? ffmpeg -itsoffset 0.342 -i danger.avs -c:v libx264 -preset slow -crf 18 -c:a aac -b:a 192k -pix_fmt yuv420p output.mkv
[16:38:48 CET] <furq> probably
[16:38:49 CET] <SilenceDogood> this article here misled me then https://wjwoodrow.wordpress.com/2013/02/04/correcting-for-audiovideo-sync-issues-with-the-ffmpeg-programs-itsoffset-switch/
[16:38:52 CET] <furq> i've not used avisynth in a while
[16:39:00 CET] <furq> or itsoffset
[16:39:25 CET] <atomnuker> utack_: you're forgetting its 20 or more years old and meant to be really fast to decode in hardware (of over 20 years ago)
[16:39:34 CET] <furq> i would have thought you'd actually want -i foo.avs -i foo.vob -map ...
[16:39:48 CET] <furq> but maybe your script actually processes the audio, idk
[16:39:56 CET] <SilenceDogood> shouldn't actually hurt anything the way i did it initially though, should it?
[16:40:17 CET] <SilenceDogood> I don't believe this script processes the audio at all
[16:40:22 CET] <SilenceDogood> I mean, it just copies it AFAIK
[16:40:29 CET] <furq> that works too
[16:40:50 CET] <SilenceDogood> probably slower to encode though because of doing the script twice
[16:40:58 CET] <furq> encode a sample to check if it works
[16:41:28 CET] <SilenceDogood> the entire video's only 8 minutes so it's not such a big deal
[16:42:03 CET] <SilenceDogood> but i may want to try a clip if the encode currently running (from that two input version) isn't successful
[16:42:33 CET] <SilenceDogood> wow, ffmpeg crashed at the end of that
[16:45:51 CET] <SilenceDogood> This encodes with silent audio: ffmpeg -itsoffset 0.342 -ss 30 -i danger.avs -c:v libx264 -preset slow -crf 18 -c:a aac -b:a 192k -pix_fmt yuv420p -t 15 danger.mkv
[16:47:59 CET] <SilenceDogood> furq help? why's the audio silent
[16:49:46 CET] <quax288> „„„„„„„„„„„„„„„ we have got more than 200% of the monthly donations today, thank you all so much!(weechat devs)crnojnqsqo: Cracki RonaldsMazitis D-ion „„„„„„„„„„„„„„„„„„
[16:49:51 CET] <quax288> „„„„„„„„„„„„„„„ we have got more than 200% of the monthly donations today, thank you all so much!(weechat devs)djiiivpk: DasMoeh zcx dashcloud_ „„„„„„„„„„„„„„„„
[16:49:56 CET] <quax288> „„„„„„„„„„„„„„„„ we have got more than 200% of the monthly donations today, thank you all so much!(weechat devs)qevsbok: Yagiza aphirst zcx „„„„„„„„„„„„„„„„„„
[16:50:01 CET] <quax288> „„„„„„„„„„„„„„„„„„„„ we have got more than 200% of the monthly donations today, thank you all so much!(weechat devs)oludrvo: D-ion dashcloud_ dreamon__ „„„„„„„„„„„„„„„
[17:06:28 CET] <Llamageddon> Hello, I want to make a video file from an mp3 file(copying the stream without reencoding). Do you know if it's possible to use the cover image embedded in the file as the video's image?
[17:06:49 CET] <durandal_1707> yes its possible
[17:08:12 CET] <Llamageddon> How do I do it?
[17:09:17 CET] <SilenceDogood> LOL furq apparently all your advice is wrong.
[17:10:46 CET] <mickie> Hi All and a happy new year!
[17:12:16 CET] <mickie> Can soneome explain what ffprobe works? I'm asking about the reporting of audio streams, in a .ts recording.
[17:13:17 CET] <mickie> One recording shows aac_latm (LC) stereo, while the other which I started recording half way through the TV stream shows aac_latm (LC) 5.1
[17:13:32 CET] <mickie> They both should have 5.1 tracks on them.
[17:20:00 CET] <mickie> So, if a stream starts as stereo and then switches to 5.1 audio tracks, how can I check to see if this is so?
[17:23:24 CET] <Llamageddon> `ffmpeg -i "file.mp3" -c:a copy file.mp4` works but with issues: Seeking does not work in mplayer, and the image does not show up in VLC
[17:25:24 CET] <mickie> OK, I used -ss to start half way through and they both show 5.1
[17:25:39 CET] <mickie> Bye all
[17:33:41 CET] <utack_> atomnuker all the more reason to dump it instead of advertise and pay for it
[17:36:19 CET] <atomnuker> welp
[17:39:32 CET] <TD-Linux> well if it's more than 20 years old you don't need to pay for it anymore. but yes, it was pretty clear it only survived by marketing
[17:39:35 CET] <Llamageddon> Okay so everything I try wrt/ converting mp3 to mp4, has a broken/missing index, making seeking impossible
[17:39:48 CET] <Llamageddon> Including the examples from http://trac.ffmpeg.org/wiki/Encode/YouTube
[17:39:51 CET] <TD-Linux> they made it really hard to abx test it, in fact the ffmpeg encoder is the first real chance to compare it to sbc.
[17:50:03 CET] <SilenceDogood> I think instead of trying to use ffmpeg options to fix syncing issues, it might make more sense to just add or subtract silence manually in audacity
[17:50:33 CET] <SilenceDogood> haha that seemed to fix it
[17:51:06 CET] <SilenceDogood> so in the end, the command was just ffmpeg -i danger.avs -metadata:s:v:0 stereo_mode=1 danger.mkv
[17:51:43 CET] <SilenceDogood> but fixed the audio source in the avs to point to an external WAV which was just the right amount out of sync with the video to counter-act ffmpeg's bizarre behavior
[17:52:03 CET] <SilenceDogood> so that (hopefully) it comes back into sync on the otherend
[17:53:46 CET] <EmleyMoor> Is there an easy way to have ffmpeg overlay a block of pixels at a fixed position onto a video?
[17:54:07 CET] <SilenceDogood> EmleyMoor, it's ffmpeg. nothing's easy. but there might be a way
[17:54:11 CET] <SilenceDogood> don't know it, sorry
[17:57:55 CET] <utack_> EmleyMoor according to google "ffmpeg overlay image" yes
[17:58:04 CET] <utack_> https://video.stackexchange.com/a/12111
[17:59:26 CET] <durandal_1707> learn to google
[18:00:39 CET] <EmleyMoor> It can often be a case of finding an expression to Google that doesn't lead you down some crazy paths
[18:01:46 CET] <utack_> TIL the default video encoder is x264, and i do not have to specify it...
[18:19:56 CET] <EmleyMoor> Just doing some test runs for position now
[18:25:38 CET] <SilenceDogood> How do I extract only the red channel of a video
[18:32:22 CET] <furq> SilenceDogood: -vf extractplanes=r
[18:32:34 CET] <furq> or more likely -vf format=rgb24,extractplanes=r
[18:36:52 CET] <SilenceDogood> furq actually here's a better thing: could I reduce the video to greyscale in a way that only pays attention to red?
[18:37:09 CET] <furq> that's what extractplanes does
[18:41:06 CET] <SilenceDogood> furq OK thanks!
[18:41:13 CET] <Captain_Haddock> Hi, first time user here. I'd like to programmatically generate videos that involve (looping) animated GIFs, captions, and audio from separate sources. Can I get this done with FFmpeg?
[18:42:12 CET] <Captain_Haddock> i.e., the sources are animated GIFs; I've already been able to convert them to mp4s using FFmpeg.
[18:43:39 CET] <Captain_Haddock> I'd like to crossfade different clips from the GIFs and perhaps also the audio. I'm trying to figure out what the limits, if any, are with FFmpeg.
[18:54:36 CET] <Llamageddon> I'm trying to convert mp3 files(copying the stream) to mp4/mkv, and no matter what I do, seeking does not work for mp4 at all, and seems broken for mkv in vlc
[18:56:56 CET] <Buster> doc/Makefile:51: recipe for target 'doc/fate.txt' failed
[18:56:56 CET] <Buster> make: *** [doc/fate.txt] Error 1
[18:58:16 CET] <SilenceDogood> furq thanks for the advice on the red filter. That totally cleared up my problems with color on this video. Now the only problem left is working out how to make an exact audio sync
[19:01:20 CET] <furq> SilenceDogood: ffmpeg isn't particularly great at processing dvds
[19:01:29 CET] <furq> you might want to use another tool to demux the title set and then work on that instead
[19:02:15 CET] <furq> ffmpeg will just discard any information about the title set that's part of the ifo, it just reads the vob as an mpeg-ps stream
[19:04:17 CET] <therage3> that does explain why i was getting bad encodes with ffmpeg on this cartoon DVD
[19:04:31 CET] <therage3> I had to make mkv's first and then parse those to ffmpeg
[19:11:15 CET] <illegal> What could cause ffmpeg to be recognize by the cmdlet on windows when you use ffmpeg, but not when you add commands such as ffmpeg -i "file.mka" -a:c libopus "outfile.opus"
[19:12:28 CET] <furq> it's -c:a
[19:13:08 CET] <illegal> https://ghostbin.com/paste/ndvvn
[19:13:27 CET] <illegal> furq, still got the error
[19:13:36 CET] <furq> fffmpeg : The term 'fffmpeg' is not recognized as the name of a cmdlet, function, script file, or operable program.
[19:13:56 CET] <furq> that might have something to do with it
[19:13:59 CET] <illegal> lol
[19:14:08 CET] <illegal> yeah
[19:14:50 CET] <illegal> Thanks furq
[19:15:11 CET] <illegal> I swear I never inspect the most obvious shit
[19:25:17 CET] <SilenceDogood> I'm trying to play it by ear on syncing here which I totally hate
[19:25:26 CET] <SilenceDogood> There's gotta be some way to fix this syncing issue properly
[19:25:46 CET] <SilenceDogood> I can't get any more precise than 0.1 seconds by ear
[19:26:09 CET] <SilenceDogood> But it might still feel wrong to people watching :(
[19:29:10 CET] <therage3> you could try opening both the original stream and the new one in something like Audacity and then do the comparison
[19:29:35 CET] <therage3> with Audacity you can go down to the sampling frequency and make it that precise if you want
[19:33:10 CET] <furq> SilenceDogood: have you tried not sending the audio through avisynth
[19:33:44 CET] <furq> there might be a start delay hardcoded in the vob that isn't making it through avs
[19:33:53 CET] <furq> i forget whether that would be in the vob or ifo
[19:34:17 CET] <furq> i'm also not entirely sure how ffmpeg deals with that if you have multiple inputs but there's only one way to find out
[20:08:16 CET] <sjuxax> Hello. I am trying to transcode a video that contains an equirectangular side data stream. I can't seem to get this side data to come over. The situation is identical to https://stackoverflow.com/questions/44760588/preserving-side-data-information-for-360-video-transcoding-using-ffmpeg.
[20:09:09 CET] <sjuxax> I tried the suggested -fflags +keepside but it hasn't made a difference, as reported on that question.
[20:11:54 CET] <SilenceDogood> What the hell. my video came out in widescreen instead of 16:9
[20:13:57 CET] <SilenceDogood> look at this. Everything's right except it's been squished into widescreen https://www.youtube.com/watch?v=qQ4LPFywwOY
[20:17:54 CET] <therage3> that's more than likely because of SAR and DAR stuff. i also had the same issue with a DVD cartoon I uploaded to youtube at, presumably, 720 x 480, but YouTube always uploaded at 640 x 480
[20:18:19 CET] <therage3> also, just a fyi, "widescreen" per se doesn't mean a specific aspect ratio, just anything greater than 1.37:1
[20:18:28 CET] <therage3> which 16:9 is
[20:18:39 CET] <therage3> i understood what you meant, but remember that
[20:18:41 CET] <furq> that video is just green here
[20:20:28 CET] <SilenceDogood> furq what
[20:20:43 CET] <SilenceDogood> it's a 3D video FYI
[20:20:50 CET] <furq> yeah it's just solid green
[20:20:57 CET] <SilenceDogood> furq well that's bizarre
[20:20:59 CET] <furq> i changed it to 2d and it's still green, but a different shade
[20:21:01 CET] <SilenceDogood> try refreshing the page
[20:21:04 CET] <furq> i did
[20:21:11 CET] <SilenceDogood> what browser
[20:21:15 CET] <furq> firefox
[20:21:23 CET] <SilenceDogood> Looks OK in Chrome
[20:21:31 CET] <SilenceDogood> except for the aspect ratio issue
[20:21:34 CET] <SilenceDogood> I think I figured it out
[20:21:42 CET] <SilenceDogood> the original video is 720x480
[20:21:50 CET] <SilenceDogood> the 3D SBS is coming out as 1440x480
[20:22:15 CET] <SilenceDogood> To get the right aspect ratio without losing any quality, I need to tell this script to do 1440x960
[20:22:40 CET] <SilenceDogood> It's working on that now. hopefully that'll look better
[20:23:46 CET] <therage3> SilenceDogood: is it supposed to be in grayscale?
[20:23:49 CET] <therage3> in 2D mode I mean
[20:23:57 CET] <SilenceDogood> therage3 yes.
[20:24:17 CET] <therage3> SilenceDogood: then I see it as intended, not quite sure why furq is seeing green o.O
[20:24:32 CET] <SilenceDogood> Might be your graphics card furq
[20:25:24 CET] <kerio> SilenceDogood: are you sure you can't just set the PAR?
[20:25:32 CET] <SilenceDogood> kerio what's PAR?
[20:25:34 CET] <kerio> it seems awkward to have a vertically stretched video
[20:25:36 CET] <kerio> pixel aspect ratio
[20:25:50 CET] <SilenceDogood> kerio where's that set
[20:27:38 CET] <kerio> or just setdar i guess
[20:27:40 CET] <kerio> https://www.ffmpeg.org/ffmpeg-all.html#setdar_002c-setsar
[20:29:20 CET] <SilenceDogood> kerio that's a filter isn't it? So all that does is stretch the video the same way doesn't it?
[20:29:53 CET] <kerio> nope, that changes the aspect ratio without changing the video
[20:30:40 CET] <SilenceDogood> So all I gotta say is "setdar=dar=16/9" ?
[20:31:34 CET] <kerio> make that 16:9 i guess
[20:32:16 CET] <kerio> if your video is h264 already you can use the h264_metadata bitstream filter to change the SAR
[20:32:42 CET] <therage3> without reencoding? :/
[20:32:44 CET] <SilenceDogood> that was copy+pasted from the help page tho. it said use slash
[20:33:19 CET] <SilenceDogood> My command to make that video on youtube was ffmpeg -i danger.avs -metadata:s:v:0 stereo_mode=1 danger.mkv
[20:33:31 CET] <SilenceDogood> The avs gets the video straight from a VOB as a directshowsource
[20:33:53 CET] <SilenceDogood> which came from a DVD
[20:34:06 CET] <SilenceDogood> fortunately unencrypted
[20:34:32 CET] <SilenceDogood> the audio portion is from a lossless WAV I made in audacity
[20:34:35 CET] <kerio> ew mpeg4
[20:34:45 CET] <furq> mkv defaults to x264
[20:34:53 CET] <kerio> furq: dvd source
[20:35:03 CET] <furq> where have you seen an mpeg4 dvd
[20:35:10 CET] <kerio> mpeg2? D:
[20:35:14 CET] <furq> there you go
[20:35:20 CET] <kerio> ew mpeg2
[20:35:26 CET] <furq> ya got that right!!
[20:35:50 CET] <kerio> SilenceDogood: https://ffmpeg.org/ffmpeg-bitstream-filters.html#mpeg2_005fmetadata
[20:35:56 CET] <kerio> display_aspect_ratio
[20:36:34 CET] <furq> don't modify the source
[20:36:41 CET] <furq> if it's off a dvd then it can only be 4:3 or 16:9 anyway
[20:37:01 CET] <kerio> i imagine it's some weird 3d dvd?
[20:37:42 CET] <furq> afaik he's converted a red/blue anaglyph into stereo 3d
[20:38:17 CET] <kerio> do you really need a dvd source for homestar runner tho
[20:38:27 CET] <kerio> it exists in beautiful vector form
[20:38:33 CET] <SilenceDogood> kerio, not Dangeresque Three
[20:38:56 CET] <kerio> :o
[20:39:00 CET] <SilenceDogood> It was created in-engine from the Strong Bad's Cool Game for Attractive People episode by Telltale
[20:39:11 CET] <kerio> anyway i object to transcoding from a dvd
[20:39:21 CET] <kerio> except at the point where youtube does i
[20:39:22 CET] <kerio> t
[20:39:30 CET] <SilenceDogood> so no flash is involved here: it's 3D
[20:39:33 CET] <kerio> because you've done all you could
[20:39:35 CET] <furq> fwiw the ar in the output video is probably the result of doing two 4:3 to 16:9 conversions
[20:39:46 CET] <furq> it almost lines up with 1280 / 1.77 / 1.77
[20:40:01 CET] <furq> so presumably your avs script is needlessly doing that conversion
[20:40:31 CET] <SilenceDogood> You want to look at the script?
[20:40:37 CET] <furq> it couldn't hurt
[20:40:39 CET] <kerio> whats avs
[20:40:41 CET] <furq> i've not used avs in a long time though
[20:40:42 CET] <furq> avisynth
[20:40:44 CET] <SilenceDogood> Would this be the right command? ffmpeg -i danger.avs mpeg2_metadata=display_aspect_ratio=16/9 -metadata:s:v:0 stereo_mode=1 danger.mkv
[20:41:10 CET] <SilenceDogood> The script is from this forum thread here https://www.youtube.com/redirect?v=qQ4LPFywwOY&event=video_description&redir_token=dmtckpRIKbezj57T9_LlSxwU_vR8MTUxNDkyMTUxOUAxNTE0ODM1MTE5&q=http%3A%2F%2Fwww.dvinfo.net%2Fforum%2F3d-stereoscopic-production-delivery%2F489934-how-convert-anaglyph-side-side-5.html
[20:41:11 CET] <kerio> needs a -bsf or something
[20:41:16 CET] <SilenceDogood> oops sorry
[20:41:18 CET] <kerio> also a -c:v copy
[20:41:18 CET] <SilenceDogood> stupid redirect
[20:41:29 CET] <SilenceDogood> http://www.dvinfo.net/forum/3d-stereoscopic-production-delivery/489934-how-convert-anaglyph-side-side-5.html
[20:41:35 CET] <SilenceDogood> kerio why
[20:41:57 CET] <kerio> hold on i don't understand the avs part
[20:42:07 CET] <furq> SilenceDogood: there's a 3d conversion filter in ffmpeg
[20:42:12 CET] <furq> !filter stereo3d
[20:42:12 CET] <nfobot> furq: http://ffmpeg.org/ffmpeg-filters.html#stereo3d
[20:42:33 CET] <furq> actually nvm it doesn't support analygph as input, only as output
[20:42:43 CET] <furq> anaglyph
[20:42:53 CET] <SilenceDogood> furq yeah, already investigated that. completely useless for this.
[20:43:12 CET] <SilenceDogood> the script is from this post here. you can click on the attachment to the post to DL it. http://www.dvinfo.net/forum/3d-stereoscopic-production-delivery/489934-how-convert-anaglyph-side-side-2.html#post1615251
[20:43:43 CET] <kerio> there's no good way to do this losslessly :(
[20:44:08 CET] <SilenceDogood> Yep
[20:44:38 CET] <SilenceDogood> There are still some minor artifacts in the final product but it's pretty good overall
[20:44:45 CET] <SilenceDogood> except for that aspect ratio problem
[20:46:14 CET] <SilenceDogood> I get "mpeg2_metadata=display_aspect_ratio=16/9: Invalid argument" so that's evidently not the right way
[20:47:11 CET] <kerio> SilenceDogood: i don't understand that avisynth thing but what does it output?
[20:48:00 CET] <SilenceDogood> kerio, video?
[20:48:07 CET] <kerio> what kind
[20:48:45 CET] <SilenceDogood> kerio, don't know, that's why I linked the script
[20:50:12 CET] <SilenceDogood> kerio, the video that comes out the other end inside the mkv that the command outputs is mpeg4
[20:51:17 CET] <SilenceDogood> wait no it comes out as h264. i was reading another thing my bad
[20:54:24 CET] <SilenceDogood> OK I am trying new command, "ffmpeg -i danger.avs -vf setdar=dar=16/9 -metadata:s:v:0 stereo_mode=1 danger2.mkv"
[20:54:32 CET] <SilenceDogood> let me know if that looks wrong
[20:56:11 CET] <kerio> SilenceDogood: honestly just use -c:v copy and drop the filtering
[20:56:18 CET] <SilenceDogood> But anyways, the point is that I'd never do this for a normal Homestar cartoon, but Dangeresque Three is the only one ever released in 3D, so this is a one-of-a kind case
[20:56:25 CET] <SilenceDogood> kerio what does -c:v copy do?
[20:56:29 CET] <kerio> if it's still wrong, fix the avisynth script
[20:56:52 CET] <kerio> SilenceDogood: it doesn't do another transcode just to put the video in a mkv
[20:56:58 CET] <kerio> or something, idk
[20:58:18 CET] <SilenceDogood> kerio but it has to do that transcode because that's the format YouTube accepts
[20:58:37 CET] <SilenceDogood> Goal is that it has to work with YouTube VR, which is fussy
[20:59:09 CET] <SilenceDogood> quality was fine on that video I uploaded earlier, just wrong aspect ratio
[21:00:44 CET] <Llamageddon> Can anyone help me? I'm trying to convert mp3 files with embedded covers into mp4 but no matter what I do seeking is broken and the video is half-broken
[21:02:20 CET] <sfan5> Llamageddon: which commands have you tried so far?
[21:03:01 CET] <Llamageddon> I tried almost every single switch I could find, every example and a few different combinations
[21:03:12 CET] <Llamageddon> This has almost everything I tried: ffmpeg -i "file.mp3" -crf 21 -bf 2 -g 30 -flags +cgop -movflags faststart -framerate 2 -preset medium -c:a copy -r 1 -tune stillimage -c:v libx264 -pix_fmt yuv420p -b:v 1000k -coder 1 -loop 1 "file.mp4"
[21:03:41 CET] <Llamageddon> I tried at least 8 different commands posted online and some other switches besides
[21:06:51 CET] <furq> Llamageddon: ffmpeg -i file.mp3 -vf loop=-1 -tune stillimage out.mp4
[21:08:30 CET] <Llamageddon> furq: Trying that with added -c:a copy results in the exact same problem
[21:08:35 CET] <sfan5> can confirm
[21:08:46 CET] <Llamageddon> Is seeking somehow impossible with mp3 audio?
[21:08:55 CET] <Llamageddon> Should I just give up and convert it to aac?
[21:09:26 CET] <sfan5> that doesn't seem to help either
[21:09:36 CET] <sfan5> is there a reason why you can't just use a m4a + cover?
[21:09:50 CET] <Llamageddon> sfan5: What is m4a?
[21:10:14 CET] <sfan5> the exact same format as mp4 but only supposed to store audio dat
[21:10:16 CET] <sfan5> data*
[21:10:25 CET] <Llamageddon> I want to convert my file for uploading to youtube while preserving the original audio steam since YouTube is going to convert it anyway afaik
[21:10:46 CET] <sfan5> ah, yea so you need an actual video as output
[21:10:51 CET] <Llamageddon> Yes
[21:11:12 CET] <SilenceDogood> Specifying setdar=dar=16/9 did not change the default aspect ratio in the skybox player
[21:11:26 CET] <SilenceDogood> I guess I could try it on YouTube but it probably won't do anything
[21:13:19 CET] <sjuxax> I've uploaded a folder of sample output from a RICOH Theta V with ffprobe outs. I have also included my attempt to use ffmpeg to copy the streams whilst retaining side data (equirectangular info). See https://drive.google.com/drive/folders/1mbSFgFpDQrU52eFYQpUZttrk8jdcq9RB
[21:13:40 CET] <sjuxax> Ongoing discussion at the SO question linked above, which was https://stackoverflow.com/questions/44760588/preserving-side-data-information-for-360-video-transcoding-using-ffmpeg
[21:14:10 CET] <sjuxax> If SO's "LordNeckbeard" is present here, /me waves; we can chat on IRC if it's more convenient/useful, but I'll be in and out
[21:14:58 CET] <sfan5> Llamageddon: ffmpeg -i file.mp3 -f lavfi -i color=s=640x640 -shortest -filter_complex 'movie=file.mp3 [x]; [1:v][x] overlay [outv]' -map '[outv]' -map '0:a' -c:a copy -tune stillimage -y file.mp4
[21:15:10 CET] <sfan5> you will have to adjust the resolution at the beginning to match the cover art
[21:17:32 CET] <Llamageddon> sfan5: That seems like a pretty complicated line, where did you get it, and how does it work? And why did what I was trying not work?
[21:18:07 CET] <SilenceDogood> Nope, YouTube accepted that just fine
[21:18:10 CET] <sfan5> I made it up; it uses an empty (black) 640x640 video and then overlays the cover art on it; I have no idea
[21:18:13 CET] <SilenceDogood> I think I'm done!! :D
[21:18:26 CET] <furq> i just checked and the command i posted is almost right
[21:18:42 CET] <furq> for some reason -vf loop=-1,fps=25 is broken but -vf loop=-1 -r 25 works
[21:18:46 CET] <furq> even though they should be identical
[21:18:49 CET] <furq> anyway
[21:19:08 CET] <furq> ffmpeg -i file.mp3 -vf loop=-1 -r 25 -tune stillimage -shortest -pix_fmt yuv420p out.mp4
[21:19:26 CET] <furq> probably change -r 25 to -r 6 if this is for youtube, although it won't make a huge difference
[21:19:40 CET] <sfan5> shouldn't -r 1 work aswell?
[21:19:43 CET] <furq> sure
[21:20:02 CET] <furq> although actually potentially no
[21:20:15 CET] <furq> there used to be an issue where doing that with -c:a copy would cause the duration to get fucked p
[21:20:22 CET] <furq> the lower the output fps, the worse it was
[21:20:36 CET] <furq> that was actually with image2 and -loop, i forget if it happened with the filter
[21:20:38 CET] <Llamageddon> furq: The output of the command you posted now does not have any video when played in mplayer2 (although I added -c:a copy)
[21:20:57 CET] <Llamageddon> Only the thing sfan5 works, and perfectly so too
[21:21:08 CET] <furq> lol
[21:21:12 CET] <furq> well that's cool
[21:21:13 CET] <sfan5> it's a huge workaround but as long as it works ¯\_(Ä)_/¯
[21:21:28 CET] <furq> it does the same thing here with copy
[21:21:29 CET] <Llamageddon> It takes non-negligible time to encode the video, but it works, at least
[21:21:32 CET] <furq> that's definitely fucking broken
[21:21:45 CET] <furq> the command didn't work at all in 3.3, i had to upgrade
[21:23:46 CET] <furq> [mp4 @ 0000000002bea940] track 1: codec frame size is not set
[21:23:50 CET] <furq> i get that when copying and not otherwise
[21:24:02 CET] <furq> so yeah that's surely broken
[21:24:37 CET] <Llamageddon> furq: Well, so at the very least a bug was discovered thanks to my floundering about
[21:24:43 CET] <Llamageddon> Feel free to report it, I don't really know where
[21:24:57 CET] <furq> actually the output file has the stream order reversed so that must be the audio stream
[21:25:02 CET] <furq> so that makes sense
[21:25:11 CET] <furq> this is still broken though
[21:26:02 CET] <furq> wtf
[21:26:09 CET] <furq> i get an x264 encoding log but the stream isn't there in the output
[21:26:54 CET] <Llamageddon> furq: What do you use to check the composition of the final file?
[21:26:58 CET] <furq> ffprobe
[21:27:43 CET] <Llamageddon> Okay thanks ^^
[21:27:52 CET] <furq> i must be doing something stupid here
[21:28:19 CET] <Llamageddon> Do you know if there's a way to not have to manually specify the cover image's resolution like in sfan5's command? I'd like to make a generic bash function for doing this
[21:29:07 CET] <furq> well what i would have suggested is extracting the cover art
[21:29:22 CET] <furq> then you can just do ffmpeg -i foo.mp3 -loop -1 -i foo.jpg out.mp4
[21:29:28 CET] <furq> that all works as expected
[21:29:47 CET] <Llamageddon> Ahh
[21:29:55 CET] <furq> -loop only works with image inputs
[21:30:12 CET] <furq> cover art is treated as a muxed video stream
[21:30:21 CET] <furq> which might explain why this is being such a pain in the arse
[21:31:25 CET] <Llamageddon> furq: That is also broken with -c:a copy
[21:31:33 CET] <Llamageddon> >_>
[21:31:49 CET] <furq> what's wrong this time
[21:31:54 CET] <furq> oh
[21:32:01 CET] <furq> add -shortest?
[21:32:08 CET] <furq> otherwise it'll run forever
[21:32:19 CET] <furq> seeking should definitely work with that though
[21:32:34 CET] <sfan5> inb4 it doesn't
[21:32:42 CET] <Llamageddon> Adding -shortest just makes the file 0.01s or so long
[21:32:52 CET] <furq> oh ugh
[21:32:53 CET] <furq> -loop 1
[21:32:59 CET] <furq> typo
[21:33:23 CET] <Llamageddon> I tried that too, same thing
[21:33:37 CET] <furq> paste the command
[21:33:46 CET] <Llamageddon> Dunno what but something is completely fucked when it comes to -c:a I'd say
[21:33:53 CET] <Llamageddon> ffmpeg -i in.mp3 -i out.jpg -c:a copy -loop 1 -shortest out3.mp4
[21:34:01 CET] <furq> -loop 1 goes before -i out.jpg
[21:34:15 CET] <Llamageddon> Wait that matters?
[21:34:17 CET] <furq> yes
[21:34:19 CET] <Llamageddon> Shit x.x
[21:34:49 CET] <furq> that's why you need to extract the image
[21:34:56 CET] <furq> that option is specifically an input option for image files
[21:35:03 CET] <Llamageddon> I wish any of the resources I found online said that the order fucking matters
[21:35:04 CET] <Llamageddon> X_X
[21:35:34 CET] <furq> it's a bit late to say this but 99% of online resources for ffmpeg are outdated or wrong or just stupid
[21:35:40 CET] <furq> other than ffmpeg.org
[21:35:43 CET] <Llamageddon> Yeah I noticed that
[21:35:53 CET] <furq> it's better to just ask in here
[21:35:59 CET] <Llamageddon> Okay the thing works
[21:36:22 CET] <furq> i still can't get the loop filter to work so that's my best suggestion thus far
[21:36:26 CET] <Llamageddon> Could I also reduce the framerate/bitrate/something since the video is a static image?
[21:36:32 CET] <furq> add -tune stillimage again
[21:36:44 CET] <furq> and you can adjust the framerate (-r 6 or whatever)
[21:36:47 CET] <furq> those won't make a huge difference though
[21:36:50 CET] <Llamageddon> Anywhere in the line?
[21:36:57 CET] <furq> as an output option
[21:37:12 CET] <furq> [input 1 options] -i input1 [input 2 options] -i input2 [output optionsp output
[21:37:14 CET] <furq> [input 1 options] -i input1 [input 2 options] -i input2 [output options] output
[21:38:41 CET] <Llamageddon> Okay huge thanks, it all works perfect now, and stillimage and -r 6 did speed the encoding up considerably
[21:38:51 CET] <Llamageddon> Not like it takes a long time, but since it makes no difference anyway
[21:38:52 CET] <furq> you can add -preset superfast as well if you just want to speed up encoding
[21:38:58 CET] <furq> (output option)
[21:39:10 CET] <furq> it's a still image so the video stream will be tiny anyway
[21:39:37 CET] <furq> also if this is for youtube you probably want to scale the image to 720p so that you get the highest quality audio
[21:39:43 CET] <furq> if it isn't already
[21:40:01 CET] <Llamageddon> That'll stretch the cover though, hm
[21:40:12 CET] <furq> you can keep the ar, just as long as the height is 720
[21:40:20 CET] <furq> -vf scale=-2:720 as an output option
[21:40:28 CET] <Llamageddon> What does the -2 stand for?
[21:40:51 CET] <furq> -1 scales to the correct value for the AR and height, -2 does the same thing but ensures it's an even number
[21:41:45 CET] <furq> your output is yuv420p which means the width and height need to be even numbers
[21:42:03 CET] <Llamageddon> I see I see. Really huge thanks man.
[21:42:20 CET] <Llamageddon> I've gotta note this down for future reference if I ever end up wanting to transcode videos
[22:11:00 CET] <Llamageddon> furq: One more question: What switch and where should I use if i want a less verbose input(e.g. only 1-4 lines)?
[22:13:44 CET] <mike_check> -hide_banner
[22:15:08 CET] <metalhead33> Hello everyone
[22:16:38 CET] <metalhead33> Not exactly sure about the newest FFmpeg, but as a user of FFmpeg 3.3, I looked into the source code, and found out that there can not be muxing to xwm, only demuxing fromk it.
[22:17:43 CET] <Llamageddon> mike_check: Hmm maybe I'm adding it in the wrong spot but it doesn't seem to reduce the amount of stdout ffmpeg outputs substantially
[22:21:05 CET] <mike_check> Llamageddon: can you post the command/output in pastbin?
[22:21:21 CET] <Llamageddon> mike_check: I could but anyway, I found what I needed, which is `-loglevel panic`
[22:22:59 CET] <furq> i'm still here trying to get this to work without extracting the image
[22:23:03 CET] <furq> i don't think it's possible any more
[22:25:32 CET] <Llamageddon> The workaround with overlaying the image posted by the other guy worked, dunno if that counts
[22:26:01 CET] <SilenceDogood> Thanks for your help furq. Final product is here: https://youtu.be/1LSadYPuB3A
[22:26:03 CET] <furq> it does not
[22:26:48 CET] <SilenceDogood> OK I gotta go, cya! :D
[22:26:59 CET] <furq> this is really annoying
[22:27:04 CET] <furq> the duration is two seconds out and i have no idea why
[22:30:08 CET] <furq> oh wtf
[22:30:27 CET] <furq> it actually works with -c:a copy even though ffmpeg and x264 both report the wrong number of frames written
[22:35:00 CET] <furq> http://vpaste.net/dwIOQ
[22:35:08 CET] <furq> i have no idea what's going on there
[22:35:12 CET] <furq> and also i think i'm done with this shit now
[22:35:33 CET] <Llamageddon> furq: Report the bug perhaps, if ffmpeg has a bugtracker/mailing list?
[22:35:39 CET] <furq> there is
[22:35:44 CET] <furq> idk if this is a bug or if i'm using it wrong
[22:35:50 CET] <furq> it's normally the latter
[22:47:58 CET] <tM^caoX> damnit, bashing my head around using ffmpeg with jpeg's, trying to preserve original exif metadata, is this truly impossible, to date? say simple 'ffmpeg -i img_0001.jpg output.jpg" where i want the exif to be copied as-is into the output..?
[22:48:27 CET] <furq> does -map_metadata 0 work
[22:48:50 CET] <tM^caoX> no, tried it. also tried "-map_metadata -1" ... :/
[22:49:28 CET] <tM^caoX> right now i'm using exiftool (the -tagsFromFile switch) causing an extra step in my process, would be awesome finding a way to eliminate the need for this.
[22:52:17 CET] <Llamageddon> furq: I ended up using `ffmpeg -i "$FILE_IN" -loop 1 -i "$FILE_IMG" -c:a copy -c:v libx264 -shortest -tune stillimage -r 6 -preset superfast -vf scale=-2:720 "$FILE_OUT"`, which works perfectly, although on my weak rig, inside a Linux VM, it takes almost 90s for a 4m50s long song, do you have any quick ideas about speeding it up?
[22:52:47 CET] <sfan5> you can pick faster presets but I'm not sure if that's a good idea
[22:52:49 CET] <sfan5> also less fps
[22:54:01 CET] <Llamageddon> sfan5: I tried -r 1 but that breaks seeking... which is probably not important because YT likely converts it anyway, but ah well. How do I try faster presets, and why is it unlikely to be a good idea?
[22:54:35 CET] <sfan5> -preset fast / -preset faster / -preset ultrafast
[22:54:43 CET] <sfan5> because it can reduce quality
[22:54:50 CET] <furq> he's already using superfast
[22:55:07 CET] <furq> you can try ultrafast
[22:55:11 CET] <Llamageddon> I'm copying the audio stream and the video is just a single image, I don't think quality matters :P
[22:55:51 CET] <Llamageddon> furq: Is that an output option, or should I add it to the video stream part?
[22:57:07 CET] <metalhead33> Okay, now I know why muxing to xwm is not possible
[22:57:40 CET] <furq> Llamageddon: change superfast to ultrafast
[22:58:49 CET] <Llamageddon> Hmm, don't think it makes a difference with what I'm doing. Still, thanks again for all the help!
[23:29:43 CET] <FishPencil> On a bluray disk are only the m2ts files encrypted and not the other files such as the mpls playlists or the java menus?
[23:42:55 CET] <foul_owl> Having an issue trying to cut videos. The start time is off by several seconds
[23:43:23 CET] <foul_owl> What is the most accurate way to cut a video? ie, the closest I can get to my timestamps?
[23:45:27 CET] <klaxa> if you are using -c copy you are limited to cutting at i-frames at the beginning (and under some circumstances at the end as well afaik)
[23:45:30 CET] <kepstin> foul_owl: if you're copying the video (-c:v copy), the closest you can get is the nearest keyframe, so it depends on how the original video was encoded.
[23:45:41 CET] <kepstin> foul_owl: if you re-encode the video, ffmpeg's seeking is frame-accurate
[23:46:46 CET] <foul_owl> How often do keyframes happen? Like every few seconds or so?
[23:48:42 CET] <klaxa> depends on the codec and the encoding settings
[23:48:52 CET] <foul_owl> Gotcha
[23:48:59 CET] <foul_owl> Ok so sounds like reencoding is my best option here
[23:49:23 CET] <foul_owl> What is the best way to reencode that doesn't cause any quality loss from the original?
[23:49:29 CET] <foul_owl> Or at least minimizes it
[23:50:24 CET] <kepstin> depends what you plan to do with the video afterwards, really
[23:50:53 CET] <foul_owl> I basically just want to keep the bitrate / codec / quality of the original clip
[23:50:59 CET] <foul_owl> But cut it to frame accuracy
[23:51:19 CET] <foul_owl> Is there a flag to tell ffmpeg "keep the original codec but reencode because I need frame accurate cuts" ?
[23:51:22 CET] <klaxa> but do you want to keep editing it or do you want to share it directly?
[23:51:30 CET] <foul_owl> I don't need to edit it further
[23:51:30 CET] <klaxa> there is no such flag
[23:51:34 CET] <foul_owl> I just need to cut
[23:52:07 CET] <klaxa> your best bet is probably just using -c:v libx264 and experiment with -crf values
[23:52:12 CET] <klaxa> afaik the default is 23
[23:52:16 CET] <klaxa> lower means higher quality
[23:52:22 CET] <foul_owl> Gotcha!
[23:52:36 CET] <foul_owl> That seems like a pretty good standard, I just have no idea :)
[23:52:50 CET] <klaxa> so: ffmpeg -i input.mp4 -ss 01:23:45.67 -c:v libx264 -crf 23 -t 10 output.mp4
[23:52:52 CET] <kepstin> hmm. there are some tools out there iirc which can cut h264 videos by only re-encoding the first (&last?) gops to cut it frame accurate, but copy the middle bit
[23:53:00 CET] Action: kepstin can't remember offhand what could do that
[23:53:20 CET] <foul_owl> -ss after -i ?
[23:53:47 CET] <klaxa> since a few years (?) it doesn't matter
[23:54:14 CET] <klaxa> https://trac.ffmpeg.org/wiki/Seeking
[23:54:34 CET] <foul_owl> Hmm, it seems to make a large difference in the amount of time it takes to encode on my system
[23:54:41 CET] <klaxa> > As of FFmpeg 2.1, combined seeking is still possible but I have yet to find a valid use case for it since -ss as an input option is now both fast and accurate.
[23:54:43 CET] <kepstin> if you put -ss after -i it'll be a lot slower, yes
[23:54:51 CET] <foul_owl> But I'm on Debian also, so the packages are probably old as hell
[23:54:56 CET] <kepstin> since in that case it has to decode then throw out the entire video
[23:55:13 CET] <klaxa> oh, i'm a dumbo, nevermind me
[23:55:46 CET] <foul_owl> Thanks everyone!
[00:00:00 CET] --- Tue Jan  2 2018

More information about the Ffmpeg-devel-irc mailing list