What are some top video renderers?

[Yog_Sothoth]

I know that madvr is considered the best, but I need some alternatives, since my hardware does not always allow to use it with proper performance.

I used to use Enhanced Video Renderer (custom preset). But now I noticed that in direct comparision, its quality is much worse than madvr's. At the same time, I've tried several other renderers and some of them, I think, look better than EVR and maybe even no worse than madvr. I've tried madvr, evr, Direct3D 11 Video Renderer, Direct3D9 Video Renderer, Haali Video Renderer (although, it seems that when I choose it, instead, the player loads Direct3D9 Video Renderer for some reason), MPC Video Renderer. I'm not sure if I'm able to determine which one of them gives the best quality, so I need your help.

I'm trying to switch from MPC Home Cinema to Pot Player since it has auto switching profile options.

I also noticed that sometimes its built-in decoder works faster than LAV decoder, and sometimes vice versa. And also I noticed that sometimes built-in decoder drops frames when used with madvr, while LAV decoder doesn't, with the same file. So I'm interested in that aspect as well.

[aufkrawall]

When doing Jinc 2D scaling, mpv is much (e.g. ~3x on Polaris) faster than madVR and quality is the same. It's also much faster than EVR when using bilinear scaling, I can clearly see that on my mobile Gemini Lake SoC.
Though other aspects like the deband filter or madVR's smart separation of chroma up- and image downscaling are not directly comparable.
mpv's D3D11 renderer afaik also doesn't rely on bad APIs like D3D9 at all, so it should be less prone to driver breakage.

[Yog_Sothoth]

Originally Posted by aufkrawall

When doing Jinc 2D scaling, mpv is much (e.g. ~3x on Polaris) faster than madVR and quality is the same. It's also much faster than EVR when using bilinear scaling, I can clearly see that on my mobile Gemini Lake SoC.
Though other aspects like the deband filter or madVR's smart separation of chroma up- and image downscaling are not directly comparable.
mpv's D3D11 renderer afaik also doesn't rely on bad APIs like D3D9 at all, so it should be less prone to driver breakage.

I didn't understand half of what you said, but as far as I understand, you suggest me using mpv, the player, and it has its own renderer? I've heard about it, but never used it, but, wow! it really is super fast! But are you sure about the quality? I tried comparing through screenshots, here's an example:
potplayer with dx11 renderer:
https://imgur.com/XnOM9Dc
mpv:
https://imgur.com/W9bsRLC
I think the first one is noticeably clearer. These are downscales from 4k film.

[stax76]

Defaults are not HQ.

[DanilaZabiaka]

As for me its better to limit levels add noise - as for me I dont like when image in same time is very dark and bright, its not natural

my shader pack for potplayer (debanding, levels - in potplayer and MPC it wrong, noise)
https://www.upload.ee/files/11303646/PxShader.7z.html

[pirlouy]

In your mpv.conf file, try these settings (HQ):

gpu-context=d3d11
gpu-api=d3d11
profile=gpu-hq
scale=ewa_lanczossharp
cscale=ewa_lanczossharp

I guess there are better settings, but none can be as good as madshi's NGU algorithm for those searching perfection, but if your source is good enough, it is very hard to see differences when watching your movie or TV show. Plus you don't have coil whine.

If you want madVR smooth motion (not good for your screenshot comparison):
video-sync=display-resample
interpolation=yes

If you want to use hardware decoding (even d3d11 deinterlacing works, compared to madVR):
hwdec=d3d11va
vf=d3d11vpp=deint=yes:interlaced-only=yes

mpv powerful key bindings + powerful configuration file + rapidity + quality = default player (for me).
Unfortunately, I use 23Hz on my TV (in order to use Samsung 120 Hz interpolation, even if 4:2:2) and mpv can't change refresh rate, so I launch my files with madVR then use mpv.

[DanilaZabiaka]

>In your mpv.conf file, try these settings (HQ):
Its bullshit and fake of coders. Everything could have GUI. Text config is needless, if you must change critical settings with text, there is something wrong

WTF is HQ? High quality? What kind of quality? Compressed video has problems cause it does not have enough details like 'noise'

You dont need just decode compressed video, you also should add details which are lost. Correct levels and colors

Even old CRT TV have sliders, wtf why do you think you need to change something with text config in 2020 manually?

There is Apple, Android, >95% of users stay on smartphones with graphic interface. These people who force user to fuck with configs are crazy. Software could be ready out of box, but PC developers are crazy

[stax76]

Nobody forces you to use it, you are simply not educated enough for it and probably also not smart enough.

[DanilaZabiaka]

Haha, ur stupid its not an argument, why not discuss something?
If normal guy not agree with some statement, he can disapprove some arguments

I know how modern technologies works - all you need is just show that you are smart. If you can write something in text config - you are smart and professional.
Everybody can move sliders or switch checkbox. Smart guy can write something like hackerman

What is the difference between d3d9 and d3d11 for video? May be d3d11 'faster' but how is it better? Video/audio compression based on psychoacoustic models

Does 'hackerman' knows psychoacoustic models, psychology?

No

He just can type some text. All he can is to provide some screenshots and 'do you see - another screenshot is more detailed have more colors more resolution'

He does not know history or how something works - if something have words HQ, UltraHD, 10bit so its good, more is better

But actually not. Coders and hackermens are very bad in ergonomic, psychology, they cant think. There is a lot of bad things in modern video, and its not render problem. Its possible to fix that, but it has nothing common with 'just write HQ in config file'

[DanilaZabiaka]

Look at Mario for example
There is 'original' png lossless png Mario 194 KB

Yeah, its 'ideal', but you cant see this is real world

Its gif mario, 63.3KB. Reduced pallete
But it uses DITHERING and so it looks more natural and not so smooth like 'ideal' png


What else? It's 653KB Mario, still PNG, but its size is 3x bigger


Why? I added noise and now surfaces looks more natural. Also I changed levels, black point is 16, white is 236

So it looks better than 'original' Mario, but it takes 3x bigger file size with noise, I need it to make surfaces more realistic. Of course png cant compress that well. Especially moving compressors. Nobody cared about it in analog/crt time, but with hi-res display its actual question

The irony is that 'old lossy' gif looks more natural than 'lossless png'

Video Compression operation must ignore small details like noise, but you can restore it. Renders do not restore it, refresh rate does not restore it, ultra-hd makes picture worse for some reason, HDR does not cares about it. Yes, you can use huge bitrate and force to keep some details, but anyway compressors destroys them even on bluray

[DanilaZabiaka]

Let's see how video compressors works
I use this https://www.office-converter.com/PNG-to-MP4
After conversation of noisy image I get 10.8 KB file. Okay, great, but how does it looks like?


In PNG its 339KB (if save in paint like previous images, its 273 kb, whatever), but it has no noise, compressor of only 1 frame destroys it. What about 24FPS with k-frames?

Which HQ render fix that?

[DanilaZabiaka]

Lets take mp4 image and apply noise (and debanding) shader again (without levels correction now)
It's 615KB image


I wanna say that shaders could help image. Also, its 5:39, i still dont wanna sleep so i made this
Also, these MadVR are just a fake, cause I dont know software or technologies with cares about natural view, they should be able at least to add some noise, make debanding, correct levels. As I know technologies now are opposite and paranoic about noise removal and how to give to user more synthetic image with crazy colors like HDR oled 0-255 range on lcd

[Yog_Sothoth]

Ooff. I see you were having fun while I was away. Anyway, I made a significant discovery which settled the matter. One particular hw-decoder - d3d11 - works much faster than any other, and with it, I can actually play some files, like 4k x264, using madvr, which I couldn't do before. So it's almost any file, I can play with these settings. For more "heavy" ones, I can use the same decoder, but with Direct3D 11 Video Renderer. POTplayer allows to create corresponding profiles for different files playing and automatically switch between them while opening a file.

So, while I find mpv pretty interesting and fast player, I don't see any significant advantages over my setup. It doesn't seem there's any quality difference between it and Direct3D 11 Video Renderer.

Btf, there is though one problem with POTplayer. I don't know how to set proper subtitles switching, LAV splitter doesn't do it with POTplayer for some reason.

[QBhd]

Just what are you trying to do with subtitles? Subtitles work just fine with PotPlayer and LAV. I have mine setup up to only display forced subs.

QB

[Yog_Sothoth]

You probably have a simple option to only display forced subs, but I need advanced settings to work. They don't.

[QBhd]

Again, what are you trying to do? I have used PotPlayer forever and LAV even longer.

QB

[Yog_Sothoth]

Again, I'm trying to use my advanced subtitles settings from LAV splitter and they don't work.

[clsid]

You must disable the internal codecs of Potplayer if you want to use LAV Splitter. The player may also be overriding the language choices, so disable any language settings in the player itself.

[Yog_Sothoth]

Yea, I've tried it all.

[QBhd]

Clearly you have it all figured out and don't actually need any help... I am sorry I thought otherwise.

QB