New CRT shader from Guest + CRT Guest Advanced updates

RetroArch Additions Shaders
#3936

OFC I was asking with such a platform in mind. :sweat_smile: On my Surface laptop, the difference between fast and fastest is apparently a lot of warmth and the cooler needing to do substantial work, doing stuff with NTSC seems out of the question etc. I suppose part of it is that the intel drivers just suck, I must also suspect surprises and differences (negative ones) depending on shader/core.

But mask zoom is really the only feature I miss from the fastest. So if it’s like only in the 10% range of a drop, it might be worth it.

1 Like
#3937

New Release Version (2023-11-10-r1):

Notable changes:

  • Mask Zoom added to the fastest version.
  • Clip Saturated Color Beams added to the fastest version.
  • Shadow Mask transformation adjusted to be like in other versions.
  • Only small performance hit with default settings (like 1.5%).

Download link:

https://mega.nz/file/JxpBBYqD#hmVTxuAgsUEAOwyIdmZF1rB-tGg7wAd-k8DbtUrJu68

16 Likes
#3938

:+1:I’m going to test it later today and report back later.

1 Like
#3939

I think it’s a very nice addition to the fastest version. It might require a slightly different tweaking for mask mitigation, though.

androdun-231110-192247
androdun-231110-1922471920×1440 1.79 MB

9 Likes
#3940

Any recommendations for 1080p users?

1 Like
#3941

Just use an existing preset and zoom the mask down until it looks good, if the mask is too coarse. Should work in most cases.

If the scanlines are uneven, then integer scaling options come quite handy.

2 Likes
#3942

New Release Version (2023-11-11-r1):

Notable changes:

  • Dotty line bug fixed with the fastest version.
  • Preserve mask functionality removed to keep up with other versions.
  • Small performance bump with the fastest version.

Download link:

https://mega.nz/file/0tAVhLJa#DpNZ91maRC4UzD9wzvVfa6hNzhYRhHQA1BbUkH9nnTQ

11 Likes
#3943

I’ve tested the older update for fastest and it works fine and no notable performance drop on the Surface. I had only used the old mask shift option though until now to get shadow mask variants. Can you elaborate a bit on current the option, does it equal positive values only or something? With the old option, I could shift e.g. mask 12 with -4. This looks kinda funky when you view it without scanlines, but seemed alright with. With the new version it doesn’t look the same, making me wonder if we miss anything with the newer option.

2 Likes
#3944

I want(ed) to avoid historical incorrectness and possible confusion with this, since in general shadow mask is “historically correct”, other mixtures between aperture and shadow masks didn’t exist in form of crt displays. New code is also simpler and faster.

Now it’s very easy in general to get a shadow mask, just set the option to 0.5 or 1.0 for double or single line shifting. If the total mask size is odd, then just use mask zoom of 1.0 or -1.0 to get it even. Works very well.

Mask 12 size is 7.0 (odd), so it’s best to use mask zoom of -1.0.

I know this is a new and a bit different set of circumstances, but I started to appreciate it right away, because shadow masks and slot masks are basically on auto-adjust now and the user can just pick a mask and zooms, everything else gets adjusted by itself.

4 Likes
#3945

New Release Version (2023-11-11-r2):

Notable changes:

  • More nice work on the fastest version. Previously, the image was garbled when using Fast Forward. In this release, the issue is fixed.
  • Bench/Game mode select parameter added to the fastest version.
  • Fastest version is (now) best used with FF frame skipping unless benchmarking.

Download link:

https://mega.nz/file/s9AA1QoI#qPtWFgix42BZZmji5QE8s18ZfuLKkZALgSn4il1-4gI

12 Likes
#3947

WoW!!

How do I achieve this image? Could you post the parameters you modified?

Also, do these presets have dither corrections? I don’t think I saw parameters for that.

4 Likes
#3948

Here are the parameters for the fastest version:

vigstr = "0.400000"
h_sharp = "3.700001"
s_sharp = "1.500000"
warpX = "0.050000"
warpY = "0.060000"
csize = "0.080000"
bloom = "0.700000"
mask_bloom = "0.150000"
bloom_dist = "1.200000"
halation = "0.250000"
gamma_c = "1.150000"
clips = "0.700000"
gsl = "-1.000000"
scans = "1.400000"
shadowMask = "6.000000"
maskstr = "1.000000"
maskboost = "1.100000"
mask_zoom = "1.000000"
mzoom_sh = "1.000000"
slotmask = "1.000000"
slotmask1 = "1.000000"

Classic de-dithering requires special code and passes, best you prepend a dedithering preset if you want this effect. NTSC version can also dedither.

4 Likes
#3949

New Release Version (2023-11-12-r1):

Notable changes:

  • More nice work on the fastest version. Previously, the image was garbled when using Fast Forward. In this release, the issue is fixed.
  • Bench/Game mode select parameter added to the fastest version.
  • Fastest version is (now) best used with FF frame skipping unless benchmarking.
  • Edit: smoother frame flow, optimizations (fastest version).

Download link:

https://mega.nz/file/54pUmKIQ#keAzSCiWJGdzDHA3eQzkhLNikgCOg1F9Y2xx0-TiJh0

13 Likes
#3950

Or you can also use Cyberlab Custom Blargg NTSC Video Filter Presets!

3 Likes
#3951

Some more information on recent changes regarding Fast Forwarding (FF) and the fastest version:

  • In benchmark mode, legacy behavior is present. Best used to test shader performance on a device.
  • In game mode, shader runs “efficient” while not Fast Forwarding (like always). A gpu is taxed more lightly doing ordinary frame rates.
  • For extra FF performance, FF mode using Frame Skipping is to be used.

I guess garbled image while FF could be a deal-breaker, so it was ‘fixed’ in recent versions.

4 Likes
#3952

Thanks @guest.r !!

Doubt: I was wondering: do these parameters already belong to the Fastest version or should they be used to further modify the Fatest?

What is your suggestion for each variant of your shader? Which systems are best with each of them?

Also, what parameter do I change for 1080p displays?

About prepend, I haven’t used it yet. How do I do that?

1 Like
#3953

They all exist in the fastest version. I tweaked the fastest version, saved the preset and copied the parameters from it to the forum. Only parameters which were changed from their default values get listed in the preset. It is the same situation every time when you tweak a shader and save the preset.

Really hard to tell, because some variants are here for less capable GPUs. Otherwise, all can do pretty well with most content. In general, the HD version is good for cores with higher native resolution input, standard and lite versions can do vertical content with MAME (TATE mode) and the NTSC version is special regarding image processing.

1080p displays are better off with less mask width, so it’s best you select the mask you prefer to look at this resolution. The best options are masks 6, 9 and 10 with mask zoom set to 0.0 (in other words, you don’t use mask zoom).

This option is a part of the Retroarch shader menu. It allows you to combine presets once you already loaded a shader preset. “Prepend” means you want the new selected preset to be before the current preset in the shader chain.

2 Likes
#3954

Thanks a lot @guest.r !!

2 Likes
#3955

Thanks @Cyber ! This works too!

1 Like
#3956

Is the “Mask 1” suitable for 1080p? If so, do I need to adjust anything to use this?

1 Like