[Ffmpeg-devel-irc] ffmpeg.log.20190502

Fri May 3 03:05:02 EEST 2019

[00:15:15 CEST] <TikityTik> for -sws_flags lanczos, do they need to be before the filter with the scale option? Or can it be after?
[00:15:45 CEST] <JEEB> option order only tends to matter if it's an input or output option
[00:15:50 CEST] <TikityTik> Also, I am noticing a lot of artifacting when encoding with libvpx at a low resolution like 240p
[00:16:07 CEST] <TikityTik> where line moving horizontally produce noticable interlacing effects (kind of interlacing)
[00:16:10 CEST] <JEEB> also that it is applied to the output or input file that you meant to apply to it
[00:16:24 CEST] <xxxphotography> can you people easily tell the difference between 30 fps video and 60 fps video ?
[00:16:37 CEST] <DHE> heck yeah
[00:16:38 CEST] <JEEB> proper content, yes
[00:16:46 CEST] <TikityTik> i can if the video was truly 60 fps and not artificially put as 60 fps
[00:17:08 CEST] <xxxphotography> TikityTik articfial 60 fps : what is that exactly?
[00:17:14 CEST] <JEEB> motion interpolation
[00:17:14 CEST] <DHE> frame interpolation
[00:17:18 CEST] <JEEB> or otherwise
[00:17:19 CEST] <TikityTik> where you take a 30 fps input and encode it as 60 fps
[00:17:34 CEST] <xxxphotography> oh, why would somebody do such a thing
[00:17:41 CEST] <DHE> because framez!
[00:18:00 CEST] <xxxphotography> does ffmpeg allow doing "30 fps input and encode it as 60 fps" ?
[00:18:02 CEST] <JEEB> the results are anywhere from passable to hilariously bad
[00:18:24 CEST] <JEEB> xxxphotography: in both ways. just speeding it up and then there's the minterpolate filter for motion interpolation
[00:18:44 CEST] <TheAMM> and duplicating frames works too
[00:18:48 CEST] <JEEB> sure
[00:18:51 CEST] <JEEB> but less interesting 8)
[00:18:58 CEST] <xxxphotography> does ffmpeg allow doing "30 fps input and encode it as 60 fps" ?
[00:19:09 CEST] <JEEB> you were just replied to
[00:19:25 CEST] <xxxphotography> okay, i didn't know it was answer to me
[00:19:39 CEST] <TheAMM> You were mentioned by name...
[00:20:52 CEST] <xxxphotography> can you people easily tell the difference between 1080i video and 1080p video ?
[00:21:18 CEST] <JEEB> one has combing and the other doesn't?
[00:21:27 CEST] <JEEB> (depends on the content within those interlaced fields of course)
[00:21:35 CEST] <kepstin> xxxphotography: assuming the 1080i was deinterlaced with a high quality deinterlacer, the 1080p video is 1080p60, and i'm sitting a reasonable distance away from the tv, it's hard.
[00:22:56 CEST] <kepstin> keep in mind that 1080i is (in ntsc territories) 60 fields per second, which a good deinterlacer would turn into 60fps, so motion will be smoother than e.g. 1080p 30fps
[00:22:59 CEST] <DHE> I must know of this deinterlacer
[00:23:49 CEST] <kepstin> hey, if you're far enough away from the tv and watching in realtime, you just need something that does motion compensation and bobbing and it'll look good enough. at least at the bitrates broadcast stuff tends to use :/
[00:23:53 CEST] <xxxphotography> there is no such thing as 1080i  30fps ??
[00:24:28 CEST] <DHE> broadcast here in north america will be either 720p 60fps, or 1080i at 30 frames (60 fields) per second
[00:24:34 CEST] <DHE> and by 30, I of course mean 29.97
[00:24:54 CEST] <kepstin> xxxphotography: 1080i is either 50 or 60 fields per second (depending on region), and each field is half the vertical resolution of the full frame
[00:25:31 CEST] <kepstin> xxxphotography: 1080i video is often stored with both fields combined into a single "frame", giving a 25 or 30 frame per second video with combing
[00:26:00 CEST] <cehoyos> xxxphotography: Would you like to tell us about your usecase? Interpolation is possible, it is possible (lol) to recognize interlaced content but all these questions depend on what your "original" is. If you record with 30, 25 or 24 frames per second (progressive), you will never see combing artefacts, no matter if you distribute "1080i" or "1080p"
[00:26:54 CEST] <xxxphotography> cehoyos look at my name
[00:27:09 CEST] <DHE> so, single still images and not actually movies?
[00:27:25 CEST] <kepstin> xxxphotography: assuming you're doing web distribution, you cannot do interlaced media (web browsers don't handle it)
[00:27:26 CEST] <xxxphotography> both
[00:27:51 CEST] <xxxphotography> webbrowsers does not support  interlaced?
[00:27:55 CEST] <xxxphotography> didn't know that
[00:28:10 CEST] <kepstin> xxxphotography: so... distribute video at its original native frame in progressive frames. no reason to do anything else
[00:28:21 CEST] <DHE> web browsers run on monitors that are progressive and the web video specs don't call for deinterlace support
[00:28:36 CEST] <cehoyos> You should use the lowest framerate that the intended players support (could be 10fps) to increase quality at same bitrate.
[00:28:46 CEST] <kepstin> xxxphotography: if you want 60fps result, then film on cameras supporting 60fps progressive recording.
[00:29:14 CEST] <cehoyos> (for still images that end up in a video, that is)
[00:42:42 CEST] <Guest93628> https://mp4repair.org  can repair  mp4 file with no "moov atom" .  can ffmpeg do the same?  big reward
[00:44:00 CEST] Last message repeated 1 time(s).
[00:47:30 CEST] <JEEB> there's an open source thing that utilizes FFmpeg's libraries for that. requires a working file with the exact same way of creation though
[00:47:46 CEST] <JEEB> https://github.com/ponchio/untrunc
[01:08:04 CEST] <Guest93628> jeeb why can't ffmpeg do it
[01:10:15 CEST] <JEEB> Guest93628: because that's a very specific low-level function. that's why another app that utilizes the FFmpeg libraries can do it. so FFmpeg plays a role in how it's done, but I'm not sure if it makes sense for the ffmpeg.c application to be tailored for that use case?
[01:11:33 CEST] <Guest93628> it says it needs libav
[01:16:00 CEST] <JEEB> it uses the libraries so either should be fine
[01:16:18 CEST] <JEEB> at least I've seen its dev here at some point
[03:48:13 CEST] <another> (just when i got the popcorn... ;)
[03:50:06 CEST] <kepstin> i assume the way those moov repair sites work is that they just have a db of various collected moov atoms, and they have some heuristics to guess one that matches the uploaded video.
[03:52:26 CEST] <genko321> kepstin i see:  how does  https://github.com/ponchio/untrunc  work then
[03:53:00 CEST] <kepstin> the way untrunc works is that you take one broken video and one complete video from the same encoder using same settings. and it uses the complete video to repair the broken one
[03:53:17 CEST] <genko321> kepstin and if that is true,  i am guessing ffmpeg can never support doing that
[03:53:33 CEST] <genko321> sounds way too complicated for ffmpeg to do such a thing
[03:53:35 CEST] <another> to quote from the description: "Restore a damaged (truncated) mp4, m4v, mov, 3gp video. Provided you have a similar not broken video. "
[03:54:03 CEST] <genko321> another oh so that untrunc is pointless
[03:54:21 CEST] <kepstin> yeah, untrunc works best when you're able to make another video using the same encoder - e.g. if it's a broken cell camera video, just take a new short clip on the same phone, and then that can be used to fix the broken video.
[03:54:39 CEST] <genko321> https://mp4repair.org people  are genius then
[03:54:53 CEST] <wallace_mu> hi, all i can't make fontconfig work with ffmpeg, no matter what font i specify in the ass subtitle file,  i'll get error  `Glyph 0x66FE not found, selecting one more font for (serif, 200, 0)`, and it ended up selecting by itself, but even when i specify what it's automatically selected in the ass, it still say not found..
[03:54:56 CEST] <genko321> then were able to do something nobody else can
[03:55:12 CEST] <genko321> they*
[03:55:25 CEST] <genko321> kesptin  https://mp4repair.org people  are genius then  : would you agree
[03:56:18 CEST] <another> as kepstin wrote, they probably have a huge db
[03:56:18 CEST] <kepstin> genko321: it's not really that nobody else can do it, it's that it's a) a lot of effort to collect all the various moov atoms for different devices and videos, and b) you do have to have someone clever to figure out how to choose the correct one to use.
[03:56:51 CEST] <genko321> and that site works very fast too
[03:57:45 CEST] <another> good for you then
[04:03:45 CEST] <genko321> another why you acting so bitter?
[04:07:27 CEST] <another> genko321: not trying to be. just trying to stay polite
[05:47:49 CEST] <TikityTik> I am noticing a lot of artifacts when encoding with libvpx at a low resolution like 240p
[05:48:05 CEST] <TikityTik> where lines moving horizontally produce noticable artifacts (kind of interlacing)
[07:16:55 CEST] <alimiracle> hi
[07:16:58 CEST] <alimiracle> why ffmpeg concat ''demuxer'' add mor so smol time in Video.
[07:17:49 CEST] <alimiracle> for me now in this test its add 0.12 time
[08:46:07 CEST] <genko321> what is this?  Codec ID/Info                            : Microsoft MPEG-4 v3 (pre-standard)
[10:15:38 CEST] <xqo> hello, does anyone know what relation addresses in WEBM/matroska files have? A Cue Point for a track will point to a cluster containing audio/video data, and the address ALWAYS points to an offset 48 bytes earlier than the actual cluster position
[10:15:51 CEST] <xqo> what is this address relative to? What are the 48 extra bytes?
[13:10:46 CEST] <DaniloPereira> :O ffmpeg on IRC?
[13:10:47 CEST] <DaniloPereira> duuuuude i love ffmpeg
[15:23:09 CEST] <another> genko321: it's mpeg4 part 2 video in a format before it was fully standardised
[15:23:41 CEST] <genko321> another i see,  and h264 =  mpeg4 part 4 ?
[15:27:34 CEST] <another> part 10
[15:27:55 CEST] <another> https://en.wikipedia.org/wiki/MPEG-4
[15:28:19 CEST] <genko321> okay
[15:28:33 CEST] <genko321> Microsoft MPEG-4 v3 (pre-standard)  looks horrible with so many blocks
[15:32:56 CEST] <DaniloPereira> can't use -preset in h265 vaapi?
[15:37:11 CEST] <another> DaniloPereira: https://ffmpeg.org/ffmpeg-all.html#VAAPI-encoders
[15:37:22 CEST] <another> preset seems not be an option
[15:40:07 CEST] <another> for what it's worth: preset is an encoder-specific option. encoders that have it are e.g. libx264, libx265, libwebp, ...
[15:40:28 CEST] <hendry> i'm having issues processing a file from GH5 "Conversion failed" https://s.natalian.org/2019-05-02/P1077291.mp4.log
[15:40:44 CEST] <hendry> Should I report this has a bug ?
[15:40:50 CEST] <DaniloPereira> :( that sad
[15:40:50 CEST] <DaniloPereira> i liked to use vaapi
[15:40:50 CEST] <DaniloPereira> you think its a good idea to enconde with vaapi and use libx265 to reduce the sie?
[15:41:01 CEST] <DaniloPereira> size*
[15:42:20 CEST] <DaniloPereira> also
[15:42:27 CEST] <DaniloPereira> how to show subtitles in ffplay?
[15:46:29 CEST] <another> DaniloPereira: ? you mean encode with vaapi and then encode again with libx265? no, that is not a good idea as you will loose quality
[15:46:55 CEST] <another> hendry: as it says near the end: Hardware does not support encoding at size 4992x3744 (constraints: width 32-4096 height 32-4096).
[15:48:44 CEST] <DaniloPereira> what if i set the crf to 0? will i still lose quality?
[16:09:07 CEST] <another> i'm not sure if crf 0 maps to lossless in x265
[16:09:54 CEST] <another> but i still don't understand what you hope to achieve here
[16:11:48 CEST] <DaniloPereira> same
[16:11:50 CEST] <DaniloPereira> XD
[19:31:05 CEST] <ChocolateArmpits> hey, when using formfeed characters in text input for the drawtext filter, they sometimes cause blank rectangle characters to be placed across the frame in random locations. The placement changes each time the command line is run. Is this to be expected?
[19:32:31 CEST] <ChocolateArmpits> I'm using the form feed character for newline rendering, but will probably move to using a text file which shouldn't be as troubling
[21:54:18 CEST] <friendofafriend> I'm using a webcam with V4L2, it outputs either yuyv422 or mjpeg.  How do I switch between them with ffmpeg?
[21:55:18 CEST] <friendofafriend> Read over the docs, it doesn't really make that explicit.  https://trac.ffmpeg.org/wiki/Capture/Webcam
[21:56:27 CEST] <friendofafriend> Ah, I guess it's -input_format, says https://superuser.com/questions/494575/ffmpeg-open-webcam-using-yuyv-but-i-want-mjpeg.  That would be nice to add in the doc.
[22:12:52 CEST] <ChocolateArmpits> friendofafriend, that's not "documentation" as in "documentation". Here's the documentation: https://www.ffmpeg.org/ffmpeg-devices.html#Options-19
[22:13:06 CEST] <ChocolateArmpits> oops https://www.ffmpeg.org/ffmpeg-devices.html#Options-18
[22:13:10 CEST] <ChocolateArmpits> ^ that's the correct link
[22:13:30 CEST] <ChocolateArmpits> the trac page is more like a tutorial
[22:13:46 CEST] <kepstin> there's even an example specifically for selecting mjpeg on that page
[22:14:39 CEST] <friendofafriend> Ah, I see.   https://www.ffmpeg.org/ffmpeg-devices.html#video4linux2_002c-v4l2
[22:14:44 CEST] <friendofafriend> Victim of Google rank, thanks.
[22:15:51 CEST] <friendofafriend> I still think it's odd for the trac to mention listing formats for a V4L2 device, and then not mention how to use them.
[00:00:00 CEST] --- Fri May  3 2019