Mpv, madvr & co - all about video rendering

Well sure, but I take it we all agree that we're tuning these things for even the slightest differences (or placebo), right? Even if it wouldn't be necessarily visible if you watch a regular video, I would still pick the "better" quality option, as long as I'm not dropping frames.

opengl:scaler-resizes-only:dither-depth=auto:interpolation:tscale=mitchell:waitvsync:es=no

This works for me.

what's the fucking point of these stupid threads? just ask on dumb questions thread

I dont see a difference in this shit.
Also how much are you paid to shill a free program all day every day? Youre worse than the shillary posters.

Because we argue about autistic shit here as well, not just answer stupid questions.

It's probably a bot.

am I the only one that doesnt see the difference?
Am I using it wrong? It just goes in any .conf file in /etc/mpv/lua-settings.. right?

Lanczos3 with SuperRes or jinc without SuperRes?

Since when does Cred Forums promote closed source software?

I am disappoint

If no upscaling, that's normal dude!

Jinc is enough.

Fuck off cia nigger

What? how does that answer the question?

better than what?

ewa_lanczos.

Still no presets with auto-profiles?
Does the windows build support it?

Sorry, forgot to delete the quote.

ewa_lanczossharp is almost the same as ewa_lanczos for luma.
difference will be like 0.01% for chroma.

>difference will be like 0.01% for chroma.
And it's not for luma? Why the difference?

How can I use these in my configs? I see them posted every time, yet every time I try to use them they fail.

Difference in what?

You need a recent mpv build.

ewa_lanczos vs ewa_lanczossharp

The build needs to be less than a month old.

Windows builds are on sourceforge.
OS X can build the git master using Homebrew.
Linux users can compile it using the build-mpv helping scripts on their repo.

is the insane madVR really the highest?

What's the correct approach to installing mpv on Windows?

Nah but it's already messing the picture badly enough for crazy people.

There are no correct way to install something on windows.
You should know it little pajeet.

You can find builds here: sourceforge.net/projects/mpv-player-windows/files/64bit/

To set it up as the default application for opening video and such: github.com/rossy/mpv-install/blob/master/README.md

Also get youtube-dl from: rg3.github.io/youtube-dl/download.html and put it in the same folder as mpv.

To setup a config like the ones in OPs post, hold the Windows key and press R, then write %appdata% and make a folder (if it doesn't exist already) named mpv. Then create a text document called "mpv.conf" in that folder with the settings you want inside it.

I think you should not using madVR anymore.
All its filters are deprecated, everything else is dummy as hell.

What's a good value for anti-ringing in mpv?

Antiringing is a shit!

Thanks.

>sourceforge
Off yourself.

I'm currently running (pic related). Is this recent enough? Or are those config options on their own not enough. Currently running Intel NUC i3, HD5000 on Arch.

What part of "The build needs to be less than a month old." you can't understand?

i gues sit depends on where he is.

Does mpv stand for anything?

>If no upscaling, that's normal dude!
wtf does that mean? upscaling? so if my movies are not shit copies none of this autist ricing is going to even be noticeable?
Do I have to download a 90's vhs recorded from tv then ripped to a computer inorder to see these improvements?

Jeez, don't speak broken English and be a dick at the same time, it doesn't look good. Just relax, and thanks for the reminder.

> am I the only one that doesnt see the difference?
yes
> It just goes in any .conf file in /etc/mpv/lua-settings.. right?
lol, what a retard, rtfm

smfb

If video is 1080p
And you screen is 1080p

The only difference you're going to see is chroma upscaling because it's most likely 4:2:0, otherwise no matter the player it's going to look the same (as long as every player is color corrected)

Get something that's below your native res, even if it's a 1600x900 video

Obviously you can add shaders like sharpning and other crap and those are immediately noticeable

oh ok, thanks

I'm having an issue where unless I either have fullscreen exlusive mode or use Direct3D 11 for presentation, my video hangs when I go full screen, and a memory leak happens.

Anyone have a similar issue and find a way to fix it? And in the mean time which would be better to use of the two options?

I'm on Windows 10 using the latest madVR with an AMD R9 380 on it's latest drivers.

> madvr problems

sourceforge.net/projects/mpv-player-windows/

I already have mpv downloaded, I just prefer madVR as I've used it longer.

There's no need for antiringing when scaling in sigmoidal light and using a ewa scaler
Prolly a Win10 issue, you should open a ticket on the madVR issue tracker and ask on the madVR doom9 thread
iirc the memory leak is an unresolved bug since late 2015

Just use Direct3D 11 for presentation what is the problem, it probably would be the best option.

I noticed graphical corruptions when using it whilst showing video info on screen, so I was assuming it wouldn't be good to use.
I would, but I'm too lazy to sign up to report it.

>sigmoidal light
>ewa scaler

Can you explain what these are and why they don't need anti ring?

I meant the corruptions occured whilst paused, so I guess it's not a big issue. If I'm using Direct3D 11 should I not use seperate device for presentation and DXVA?

Is haasn or somebody else smart here? Was wondering how I could use mpv as an audio player and image viewer.

they still need ar though, but ar is crap.

>but ar is crap.
hence why I asked for a good value. I was wondering if a value like 0.3 would help a bit without causing any artifacts

input.conf for keybindings in the same folder as mpv.conf?

idk, never used those.

nvm, it worked

it will help a bit, but also cause a bit of artifacts...

i think haasn's mpv.conf had some image viewer stuff in it, it's on his github

I've never seen any artefacts with madvr antiringing. This is the only top feature of madvr now since mpv reverse engineered everything else from madvr.

Is frame interpolation still possible in mpv? What config would I need?

it works differently in madvr, but seems like no body cares about ar.
another option is to use SSimSuperRes, if you don't like that it's sharp, reduce oversharp parameter.

So why not extracting the antiring part and create another shader?

From what?

From SSimSuperRes.

SSimSuperRes is ar + sharper

I know it but why not extracting the AR code from SSimSuperRes and create a new AR only shader?

because it still will be SSimSuperRes

Do you mean it will be as slow as the original? No way to optimize it and turn it as AR only shader?

nope

Is there a way to fix thumnails of files with mpv? With it set as the default player, some files correctly show a thumbnail of the video, whereas others just have a mpv logo as the thumbnail. Using WIndows.

Install gentoo

Use icaros!
videohelp.com/software/Icaros

Thanks, that worked.

tell me about ytdl-format option /mpv/...
I currently have this:
ytdl-format=18
Very happy with this option since it downloads fast and AVC is hardware accelerated which matters to me.

How do I add options for other sites that ytdl supports? For example, on Vimeo I prefer http-360p format. How do I add it to options so when I do mpv vimeo/.... it plays that format?

Using video-sync=display-resample on the Judgement release of Ghost in the Shell causes weird behaviour. Looking at stats script, the AV correction goes nuts. It jumps erratically from 0.017 to -0.006 to 0.016 to -0.017 and you get the idea. It fills the video with judder and what looks like frame skipping or something. It's the only movie/video on my library that has this issue.

You can download the torrent here bakabt.me/torrent/172528/koukaku-kidoutai-ghost-in-the-shell-1995-hdtv-1080p-judgment
Just needs a free account and it's freeleech

The niizk release works fine. I'm wondering if it's an mpv bug or encoding issue - and btw, the movie plays fine without video-sync

>>low mpv preset:
this might be a stupid question but are these presets listed in OP built-in or are you suppose to add them to mpv.conf?

>wmv
Kek

Looking at this site, the guy who made that release is actually one of the mpv devs. Maybe ask him in IRC or something

Are these stats good or not?

Your computer is too shit, upgrade or get the fuck out.

Better than 99% of people in this thread so fuck off.

Trying to swap from MPC to MPV but even with the insane MPV preset it can't even compare to MPC am I doing something wrong?

Current MadVR settings are just slightly higher than the insane preset.

mpv.conf is in %appdata%/mpv and I copied and pasted the preset.
added the shaders to ~~/scripts/shadername.glsl

also how to stretching? I'm using a 21:9 monitor

What about latest version?

ewa_lanczos is a very sharp scaler. You should read some of his explanations on this page for pros/cons of using every scaler. There's not one best scaler for anything. You always trade something for something else.

imagemagick.org/Usage/filter/nicolas/#upsampling_examples

>Your computer is too shit, upgrade or get the fuck out.
You know even if I use potato preset it still has the same judder problems

windows?
try
opengl-backend=dxinterop
or
opengl-backend=win
angle, which is default, doesn't work well with display-resample

I use dxinterop by default. Haven't tried win yet but I doubit it solvers the issue. Also never had major problems with display-resample using angle to be honest. Yes, the frame timings are huge and shit but there's no significant judder like what I'm getting here. I'll ask on mpv's irc

What scale do you use for upscaling 360p- 480p videos??

>insane mpv preset
>medium madvr preset

>ewa_lanczossharp
>chroma upscaling: catmull-rom + ar
How do i set it to use that with upscaling only?
Also madvr?

Chroma mostly matters for non-upscaled videos. OP's preset is generalized for all video content.

Still has chroma bleeding.

Do you mind posting your conf? I am gonna try to reproduce it.

Anyway, you don't need an account to download from a bakabt torrent link, also why are you choosing judgements version instead of nikkz?

>tfw no new update for sourceforge builds
>nightlys never
I just want my bitbucket builds back

sourceforge builds is more likely weekly as you can see the date.

Why is your VSync Ratio a clean 5 but my VSync Ratio is like a weird uneven number all the time?

Because you need a better graphics card/better drivers. Stable V-Sync is required for you not to drop frames like crazy like you're doing. 1080p essentially not scaled should not be so expensive unless you're running nnedi3.

>Do you mind posting your conf? I am gonna try to reproduce it.
I'm currently not at home but from the top of my head it's more or less like this
profile=opengl-hq
opengl-backend=dxinterop
interpolation
tscale=oversample
scale=ewa_lanczossharp
cscale=ewa_lanczossharp
dscale=catmull_rom
video-sync=display-resample
hwdec=no
3dlut-size=256x256x256
scaler-lut-size=10

> you don't need an account to download from a bakabt torrent link
Oh I thought you did, honestly.

> why are you choosing judgements version instead of nikkz?
Because the judgment release has better cropping, more vivd colors, less grain and it's half the size. I did watch the niizk version instead though but this issue is bothering me because it's literally the only video/movie that has this problem. Watching 1440p60fps videos on youtube is smooth as butter and 4K 30 fps videos without hwdec are also perfect without any dropped frames or judder, etc.

I don't see any obvious errors now. However, I would still say ewa_lanczossharp looks better in my opinion.
In case anybody wants to know, the file is from bakabt. I'm just using it because it's the lowest resolution file I had.

Prove that antiringing in mpv is shit, I for one see only its advantages:
w.screenshotcomparison.com/comparison/186079

Look at the ringing in the subs, I rendered them at video resolution and scaled the video with ewa_lanczossharp and scale-radius=8 to make the difference obvious.

Antiring manages to remove most of the ringing around and inside the subs without adding additional artifacts. At least I don't see them.

>w.screenshotcomparison.com/comparison/186079

How do I enable antiring?

scale-antiring=0.7
in mpv.conf

Well sure, I realize there is no perfect solution and that it's all subjective. I'm just basing it on the fact that the OP recommends ewa_lanczossharp for normal scaling, and that the manual says "If your hardware can run it, this is probably what you should use by default." for ewa_lanczossharp. If one is "better" for luma, then there is no reason to assume it isn't "better" for chroma as well, even if you're just judging subjectively.
That can't be true, right? Even if you're upscaling a low res video, you still need to chroma as well. And you also need to scale chroma even if you're using yuv444.

it has weird interlace like artifacts, probably because this video was ripped from a DVD using some crappy deinterlacer.

When would you even have scale-radius at 8? And are you using blend-subtitles?

I've never had that problem, even when upscaling 480p subs to 1440p. But then again, I don't set scale-radius to 8.

That might be true, but if that's the case then you'll see those artifacts on all screenshots. I'm just comparing screenshot to screenshot, which leads me to conclude that I prefer ewa_lanczossharp.

i think that was just for demonstration purposes

>When would you even have scale-radius at 8?

Obviously never, the point was to prove that antiring in mpv is not adding a ton of artifacts.

How can you even tell which one is better without original 1440p image?
Take some 1440p image, downscale it with mitchell and then compare upscaling to original.

Thanks, there's nothing else I need to enable right?

would berry

Again, the point was to prove that antiring does not add additional artifacts compared to scaling the video without antiring.
To compare that you only need the two upscales.

>Because the judgment release has better cropping, more vivd colors, less grain and it's half the size. I did watch the niizk version instead though but this issue is bothering me because it's literally the only video/movie that has this problem. Watching 1440p60fps videos on youtube is smooth as butter and 4K 30 fps videos without hwdec are also perfect without any dropped frames or judder, etc.
>less grain
I though more grain was better for older released shows?

Anyway, I'll just wait to you get home to test, but either way it's just this one file maybe try opening an issue on github?

A lot of people are also reporting minor problems with the file, so maybe is the file?

Actually nevermind after seeing the screenshots, I can see why you say is better.

but I can see a lot of additional artifacts there

I can't see shit, so tell me exactly where is the anri-ringing most noticeable

Not him, but can you point where? Because I can't see any either.
The subs mate. There's an extra slightly darker line (also known as ringing) in the white part without anti-ringing.

Alright, yeah I see it.

Are you using blend-subtitles? how about you turn it off?

I am, but I'm not him like I said. Also, he said it's only an example to show that anti-ringing is working, and that there are no artifacts. It's irrelevant if he's using it, it's only to showcase.

>but I can see a lot of additional artifacts there
Please tell me where to look then, I don't care being proven wrong, but if every argument is only "it looks shit" or something similar, I have to assume it's just shitposting.

Anti-ringing is a smoother, so its artifacts would include smoothing out grain and fine detail. Even the default deband=yes has enough artifacts where I can go from seeing a difference on an imagemagick compare to being well below what I would notice because deband wiped out enough of the high-pass detail.

While I'm waiting for a test example to download, here's what happened with your example. The antiringing filter was good here - almost entirely isolated to the subtitles.

>so its artifacts would include smoothing out grain and fine detail.
And I'm not seeing any of that.

But it also affects edges. I picked a 480p show with extremely soft edges.

And noisy images:

0x0.st/QD9.png

>Anti-ringing is a smoother, so its artifacts would include smoothing out grain and fine detail.
Well I assume it's masked, so only things that get detected as ringing are smoothed, the question is, how reliable that detection is. At least in my test image it does appear to work just fine.

>Even the default deband=yes has enough artifacts where I can go from seeing a difference on an imagemagick compare to being well below what I would notice because deband wiped out enough of the high-pass detail.
Of course you will see a very clear difference when using deband. It's sole purpose is to alter the source and remove banding artifacts. Using debanding is always a tradeoff between smoothing gradients and preserving detail. Antiring on the other hand is designed to prevent adding additional artifacts, namely ringing, to the source.

If you would know how ar is implemented in mpv you would know it can be shit, look at source code.

I'm not really a programmer, so it would be nice if you could explain how mpv's method works, or at least what it does badly in your opinion.

To be honest, it's a bit pointless to just post an image that just displays the differences. That way nobody can tell if the things antiring altered were just ringing artifacts or really detail.

Why is madVR being supported?

>anti-ringing does things!
wew lad, thanks for the info.

>And you also need to scale chroma even if you're using yuv444
But in this case scale is used to scale chroma too. Keep in mind that cscale is only used to scale chroma to the luma resolution of the source. If additional scaling is required scale is used.

That I didn't know.
But what if you're using separate luma scalers like that nnedi3 prescaler?

At upscaling matters Luma, when there is no upscale - Chroma.

It replaces pixel of upscaled image with the darkest/brightest pixel from the GROUP of nearby pixels of the source image if there is ringing, but that's not always the case, it can also damage pixels that have not been affected by ringing.

At upscaling matters Luma and Chroma

I'm fairly certain this isn't English. Also, how would chroma not matter if you're playing a non-native resolution yuv420 video? You still need to scale the chroma.

Or, you know, you could use a softer shader if you're going to smooth it out anyway. There's plenty of other upscalers in filter_kernels.c.

And there's this nice picture to help you pick between some of them.

imagemagick.org/Usage/img_diagrams/cubic_survey.gif

>Linking an image on an imageboard
Also, I seriously doubt anti-ringing smooths out everything like a smoother shader would.

Take for example a 720p video with 420 subsampling that is displayed in fullscreen at 1080p. Luma is initially 720p and chroma is initially 360p. First chroma is scaled to 720p with the scaler specified in cscale, then the whole image is scaled to 1080p with the scaler specified in scale.

You should look at the other jinc scalers. They're really not that different from each other, but using one or the other pretty much solves your need of using anti-ringing. If you really don't want artifacts, use haasnsoft. ewa_lanczosharp is definitely going to cause some, while ewa_lanczossoft is in the middle of the two subjectively.

Okay sure. But the discussion was about how anti-ringing introduces a bunch of artifacts, but none of those artifacts have been posted. Personally I don't recognize any ringing in my videos, but if anti-ringing doesn't introduce any artifacts, then why not use it?

Except it does according to how it's implemented:

The artifacts might not be perceptual, but that's pretty much the case with any reasonable scaler (even those that have blurry "artifacts") beyond bilinear assuming a limited upscaling factor. Both accomplish the intended goal of reducing without causing perceptual artifacts.

*smoothening

Theory =/= practice. If there are artifacts, just post an example. The artifacts aren't perceptual because they don't exist. I'd rather have a filter that only elimates ringing artifacts when there's ringing, as opposed to choosing a more smooth scaler to just smooth everything out.

is vsync-jitter at 0.007 too high?

Except I did up here.
The 0x0.st one was originally:
0x0.st/QDW.png (no AR)
0x0.st/QD4.png (AR)

And this one here didn't really do anything with the subtitles but changed a bunch of pixels.

0x0.st/QDy.png (no AR)
0x0.st/QDJ.png (AR)

>Except I did up here.
That's not an example of the artifacts, that's an example of what anti-ringing does.
But thanks for finally posting an example, I'll try and see if I can actually notice them or not.

Sorry, the first 2 are paired with diff8b, and the second 2 are paired with the first 0x0.st one.

How's that not an example of artifacts? AR shouldn't be doing anything to those images because there's pretty much nothing but grain, but the differences prove there are changes. And as I said before, these aren't perceptual.

what ar value are you using?

>How's that not an example of artifacts?
Because it only shows what the anti-ringing does. Without reference you can't tell if it's working there when there is actual ringing (in which case there wouldn't be artifacts), or if it's working there when there isn't actual ringing (in which case it would be artifacts). Also, the actual artifacts that show up don't appear as colour coded pixels.

>AR shouldn't be doing anything to those images because there's pretty much nothing but grain
How were we supposed to know that? At the time of posting obviously, not now because now we can compare the actual artifacts.

>And as I said before, these aren't perceptual.
You didn't actually say these weren't perceptual, you said artifacts in general might not be perceptual, which is obvious. If these aren't perceptual, then I don't think they're good examples of artifacts, because how bad are artifacts if you literally can't see them? Talking in direct comparison of screenshots obviously. And I don't see any reason to believe these aren't good examples until I'd see better examples.

And as a similar example. This is the diff between deband=yes and deband=no.

I can't even remember which one is which of these, but I remember I turned deband on for just one of them.

Can you ABX the difference between these 2?

0x0.st/QDw.png
0x0.st/QDv.png

Pretty much, our eyes suck at detecting artifacts, even if there's a ton of it.

Chroma doesnt matter much.
Chroma layer has little to no role in quality when you upscale something. Thats why its generally recommended to use higher Luma settings instead of chroma when upscaling something.

Not much =/= not at all.
I'm saying upscaling chroma matters at least as much for a native resolution yuv420 video, as it does for a non-native resolution yuv420 video. You or at least the previous posts seem to disagree with that.

The difference here is, although computable, imperceptible. In fact the changes are so slight that it would be very hard to prove that changing those pixels wasn't the "correct" thing to do.

On the contrary, the positive aspects of of antiring are easily visible on very sharp edges.

>How's that not an example of artifacts? AR shouldn't be doing anything to those images because there's pretty much nothing but grain, but the differences prove there are changes.
Not every artifact that sharp scalers cause that very distinct ringing pattern. There will a lot of pixels that are too dark or too bright. And it is hard to say whether antiring does the right or wrong thing in your examples.

Im a simple man, i know the basics. Chroma channel applies to all videos and matters the most when there is no upscaling. Luma channel is most important when upscaling is happening. Those are the facts. Im not arguing with anybody.

>Chroma channel ... matters the most when there is no upscaling.
Compared to what? Because I assumed compared to chroma channel when there is upscaling, is this right or did you mean compared to luma with not upscaling?

>Can you ABX the difference between these 2?
I believe so yes, I think the first one is banded, the second on is debanded. There are slight differences mainly in the background structure.

Is antiring positive though? It's essentially a low-pass filter targeted at high frequency data.

People see removing ringing as positive, debanding as positive, sharper images as positive, smoothing as bad, and value chroma information far less than luma information because all but the last one are preferences, not objective improvements. Even when many of these subjective improvements substantially alters the original image objectively ( ) it is generally imperceptible to people in general.

It is hard to say if debanding did the right or wrong thing in the example I provided earlier. Either way, it still did something, though imperceptible. Antiring also does something imperceptible to the image, even when it's difficult to say if it should have done it at all.

Finally, if you were the guy who was worrying about using targeted AR with a sharper shader was worse or better than using it with a naive shader than smooths all the pixels, it's imperceptible as well. And the differences of using the naive shader and a sharper one with AR are all clustered around edges only, which means unsurprisingly that the naive shader only changes edges to reduce ringing.

0x0.st/QDI.png (haasnsoft)
0x0.st/QD6.png (ewa_lanczos+AR 0.7)

>People see removing ringing as positive, debanding as positive, sharper images as positive, smoothing as bad, and value chroma information far less than luma information because all but the last one are preferences, not objective improvements.
Not me, I see it as an improvement because it goes against artifacts, which are by definition unwanted extra visual data.

Been having problems with streams stuttering, using youtube-dl.
This wasn't happening when I streamed to firefox, so I reinstalled livestreamer to see how it would handle it. To my surprise it streams perfectly, just like firefox.

Is this a known issue with youtube-dl, or have I just broken something?

youtube-dl wasn't designed for streaming

>(haasnsoft)
stop this right now

Alright, so it's not on my end at least. I can live with that.

Chroma is only scaled 2x even if you're watching a 360p video on a 2160p monitor (whereas luma would be scaled 6x)

>People see removing ringing as positive, debanding as positive, sharper images as positive, smoothing as bad, and value chroma information far less than luma information because all but the last one are preferences, not objective improvements.

No, I don't like sharpening. I dislike it. I personally use haasnsoft, debanding and antiringing.

Glad to see all this autistic mpv only club shitposting is finally in one circlejerk

you guys need a pastebin tho

>haasnsoft
>antiringing

:))

I think you're wrong, or are saying something you don't actually mean. I'm assuming chroma will be doubled to bring it in line with luma, but then it would still be scaled just with the same luma scaler.
And not a lot of videos are going to be a quarter of your monitor's resolution, like chroma is in most videos (compared to luma).

>A friend can betray you, but an enemy will always stay the same.
Free and opensource software has nothing to do with this madshi's monstrosity.

Let demolish this stupid big castle!

MPV is not even close to be a threat to madvr yet. It is just linuxtards have no choice but to use it because madvr isnt there.

>MPV is not even close to be a threat to madvr yet

Optimized code, daily releases, proper color management and more shows that mpv is already more sophisticated than any other gay video player. You can even use it as an image viewer or music player thanks to the functionality.

Are you fucking retarded?
Most of mpv users are winfags.

>This is the only top feature of madvr now since mpv reverse engineered everything else from madvr.
What did mpv reverse engineer from madVR?

Pretty sure he's talking about the old NNEDI3 and Superxbr shaders that used to be included with mpv, but the developers didn't reverse engineer those lol, not even close. They removed those several months ago though because separate user shaders are superior and allow more flexibility.

>it works differently in madvr, but seems like no body cares about ar.
Funny thing is madshi actually described the madVR AR algorithm to me in an e-mail, I just didn't find it made any visible difference compared to the more naive algorithm mpv uses.

>wmv
Sounds like it's either an encoding issue or an issue with how ffmpeg decodes wmv

Can you upload a 10-second clip (make one with ffmpeg -c copy or something)

Also, it would help if you played for 10 seconds or so (or however long it takes to observe the issue happening) using --dump-stats=FILE and uploaded the resulting file somewhere.

Ignore him, pretty sure he's just parroting bullshit. Your issue is clearly file-specific, and has nothing to do with your PC's specs

Looks good

>And you also need to scale chroma even if you're using yuv444.
False, 4:4:4 has the luma and chrome planes the same size and aligned.

I assumed (wrongly) that it would scale scale the chroma separately still. Technically it still needs to be scaled, but I guess it just scales together with luma then.

scale scales RGB

How the fuck did you get your Vsync Jitter and Frame Timings that low.

You have to be using the worst settings I swear to fucking god.

>Take some 1440p image, downscale it with mitchell and then compare upscaling to original.
Won't that just give you the upscaler that best inverses the effect of mitchell?

Link to stream

I see absolutely no difference between the two

>Can you ABX the difference between these 2?
They look absolutely identical to me

I again see absolutely no difference between those two images. Shouldn't haasnsoft cause a far more noticeable difference? On my end it's significantly blurrier

>I think you're wrong, or are saying something you don't actually mean.
Maybe you're misreading my post?

>I'm assuming chroma will be doubled to bring it in line with luma, but then it would still be scaled just with the same luma scaler.
That's pretty much what happens, yes.

For a 360p video, the chroma plane would only be 180. So the chroma scaler would first bring this up to 360p (2x scaling), and then the combined video would be scaled to 2160p using the general scaler (6x scaling).

Which is exactly what I said, so I'm not sure how we're disagreeing

It's to high >0.5

Do people not see a difference? Look at the edges. Haasnsoft is obviously softer and doesn't have as much artifacting.

>That's pretty much what happens, yes.
In which case technically chroma is still scaled together with the luma.

>For a 360p video, the chroma plane would only be 180. So the chroma scaler would first bring this up to 360p (2x scaling), and then the combined video would be scaled to 2160p using the general scaler (6x scaling).
>
>Which is exactly what I said
No it isn't. You said previously that chroma would only be scaled 2x, not 2x and then 6x.
But this is all semantics and very autistic, so I'm not going to argue about this anymore. I guess it's very possible that what I'm saying is actually technically incorrect as well.

Neither NNEDI3 nor superxbr were reverse engineered from madVR as you say,

NNEDI3 was made in C(?) by tritical, ported to OpenCL by SEt, then ported to HLSL by the MPDN people, and finally ported to GLSL by bjin

superxbr was made in HLSL by Hylian, ported to GLSL by bjin and then later reimplemented in GLSL by haasn

Neither were taken from madVR or madshi's code. (Although I think madshi did contribute to the MPDN version of NNEDI3)

Nice, I got a 10-second clip from the author of that file

Can confirm that it happens, but it seems like an issue with the audio track, not the video track. (Also, playing with --no-audio fixes it)

>No it isn't. You said previously that chroma would only be scaled 2x, not 2x and then 6x.
The point is that the scaler you specified using ‘cscale’ only does 2x scaling

thanks for doing that, was about to try it. So it's definitely an issue with the release, correct?

Can anyone share a config that replaces the current OSD with a bar on the bottom? I've seen some screenshots of people using this around here before

osc-layout=bottombar

Just any chaturbate stream. It will play for a couple seconds, then stop for a couple seconds, repeat.
Twitch streams are fine, and I haven't tested any other livestreams.

Wait, youtube-dl can play chaturbate streams? Do I just need to pass the URL?

>Optimized code, daily releases
You did not see madvr code and from what we seen its not any slower at the same settings than mpv. Madvr is pretty much perfect except of very few bugs maybe, it doesnt need a new release every day.
>proper color management
Madvr does not have it?
>any other gay video player.
Madvr is not a video player, dumbass.
Are you fucking retarded?
Most of mpv users are freetards.

Yep. You can also let ytdl output to a file, and mpv will happily play the file too.
Livestreamer supports chaturbate too.

it's playing fine here, no noticeable stutter

I thought it was only me doing this. I think they changed something on their end because this started like a week ago out of nowhere.

Just tried chaterbate and worked fine.

I used mpv then the url not youtube-dl btw

btw it seems a lot better now than last week, for a few days I could barely watch anything on there because it was so fucked up.

Also if anyone knows how to watch myfreecams with mpv please let me know.

Hmm. Guess I won't be sleeping tonight then. Thanks tho

That's what I thought, because this only started happening the other day. But livestreamer works despite not being updated in over a year.

Both ways give me the save problem, unfortunately

I thought youtube-dl was for actually downloading and the mpv command streams content. Thats how its always worked for me. And I just tried it 1 min ago.

>But what if you're using separate luma scalers like that nnedi3 prescaler?
I decided to make this image of how images flow through vo_opengl to help explain what gets scaled, where, by what, and why

Suggestions welcome

Are you for real? How about you educate yourself first then make retarded images like that...?

...

Why did you crop that ass?

The scaler used for this step depends on the type of plane. (Luma planes get scaled by --scale, chroma planes get scaled by --cscale), so I put it on the arrow

Maybe there would be a better way to represent that?

>Luma planes get scaled by --scale
What?

>Luma planes get scaled by --scale
You wot m8.

github.com/mpv-player/mpv/blob/master/video/out/opengl/video.c#L1836

That's for cases when you want to watch only Luma channel or what?
Why would you need to scale luma?

>I thought youtube-dl was for actually downloading and the mpv command streams content. Thats how its always worked for me. And I just tried it 1 min ago.
youtube-dl supports more and more livestreams/streaming websites

Is there any reason to use MPV over MPC-HC on Wangblows?

Why dont you do some research yourself? Not only will you find the right answer because you know what you want out of a player, but you will also learn how to use your own mind ;)

the answer is "no" then
thanks

You're welcome

Good luck in life

> comparing life to asking a question about media players on 4 leaves

It's for hypothetical cases. For example, you could imagine a situation where the chroma plane is actually bigger than the luma plane (which would be possible to accomplish for a user hook).

I actually thought mpv always used the biggest plane as the reference, but it actually uses the first “rgb or xyz or luma” plane as reference. My mistake on that part, I've corrected it. (And also added some more explanations where appropriate)

>wanting everything to be spoonfed in life

Got to grow up sometime

I mean, MPV is not really the most popular player out there, and this is a thread for it, and he is asking about it.

Are you pretending to fit in?

So give him your unbiased answer

Is it worth it to dering the source before scaling?

Only if your source has unwatchably bad deringing