Jump to content

SOL: High Fidelity Graphics for KSRSS - Wen Release?!?


G'th

Recommended Posts

4 hours ago, G'th said:

Cloud patterns will be isolated into smaller individual textures and will populate the body, and will fade in and out and change shapes as they cross the world. 

4 hours ago, G'th said:

would be easier to control for both clarity and consistency, and in turn it allows for me as the modder to actually make use of the source texture

This is very a very exciting prospect. I've found in the packs I've been using recently some clouds look fantastic and some look  how should I say this, exceptionally angular :P, from what I've gathered from what you've said this would potentially eliminate this as you could use the full texture on the smaller cloud systems which would be joined by other smaller systems to form the major cloud systems. This questions is based in my lack of understanding on the subject but would this eliminate the texture seams from traditional cube maps?  Overall I think this would be a fantastic avenue to peruse. I would guess you're working on the cutting edge of what EVE can do as I've never seen this before. 

I'm not a modder so I don't really know if I fully understand the meaning of the paragraph but I don't need to be one to think it sounds very exciting! 

Edited by Kuiper_Belt
Page Three?!
Link to comment
Share on other sites

Just now, Kuiper_Belt said:

This is very a very exciting prospect. I've found in the packs I've been using recently some clouds look fantastic and some look  how should I say this, exceptionally angular :P, from what I've gathered from what you've said this would potentially eliminate this as you could use the full texture on the smaller cloud systems which would be joined by other smaller systems to form the major cloud systems. This questions is based in my lack of understanding on the subject but would this eliminate the texture seams from traditional cube maps?  Overall I think this would be a fantastic avenue to peruse. I would guess you're working on the cutting edge of what EVE can do as I've never seen this before. 

I'm not a modder so I don't really know if I fully understand the meaning of the paragraph but I don't need to be one to think it sounds very exciting! 

Yes, it absolutely would kill seams as a problem. Seams with cubemaps aren't terribly hard to deal with, with the simplest solutions imply being that each face of the cube just doesn't have anything that goes towards the edge, which is probably what I would have went for if I what I want to do instead wasn't an option, but they are still a pain and exceptionally small ones are easy to miss and even harder to track down. 

But with this method, each cloud pattern would be inclusive to the textures edges, making it impossible for seams to occur. (unless they're already was seams in my source map, but from what I've seen that shouldn't be the case)

Link to comment
Share on other sites

6 hours ago, G'th said:

So I feel considerably better now, particularly after getting inspired by certain things, so we're back on.

But with that, I want to get some feedback on what Im going to be doing.

 

Now, the main idea that inspired the mod was to eventually culminate in my vision for a dynamic Jupiter that recreates, to what extent EVE and Scatterer allows, how the real Jupiter looks and behaves, as seen in the video and pictures linked a few posts ago. 

And Im pretty confident in my ability to do this efficiently. Unlike conventional cloud packs, that use a single texture to depict the entire atmosphere and cloud bands, my method would involve using a number of textures, dedicated to each individual element that run independently from one another. While that sounds like a lot, nature is on our side given that much of Jupiter, and the other giants that are done in the same way, often have densely repeated features. One set of cloud band textures (there would be two for each element) could easily cover almost every band in Jupiter, with EVE settings enabling the necessary variations in those bands. The ones not repeated meanwhile, would only represent just the additional sets added, like the Red Spot or the polar hexagonal storms. 

So while a planet like Jupiter could nominally have, lets say, 20 individual elements, the number of textures required is may be half of that, if not less. So more than the typical method, but thats the trade off for the dynamics that I'm shooting for. 

Now, I explained all that to contexualize what Im actually asking for feedback on: Extending this concept to Earth, Mars, Venus, and Titan. 

Obviously my efforts thus far have been on Earth, and through my experimenting and tweaking something Im fundamentally taking issue with is the clarity and consistency of the clouds. While my source map for the Earth clouds hits these marks adequately, when scaled down, the consistency drops dramatically with some parts looking fantastic while others are rather disappointing. 

 

Ive noted before that I dont want to go nuts with 64k textures and all that other stuff that has to sacrifice gameplay fidelity to accomplish, and given I can't even process textures that big anyway, the solution for me is, as said, in using the same method Im intending for the gas giants to govern the other cloudy worlds. 

How it would work, to use Earth as an example, is more or less similarly to how the typhoon does. Cloud patterns will be isolated into smaller individual textures and will populate the body, and will fade in and out and change shapes as they cross the world. 

The benefits are rather obvious even beyond the dynamic clouds. Each individual cloud pattern (which could be one giant system, individual spurts, or some mix) would be easier to control for both clarity and consistency, and in turn it allows for me as the modder to actually make use of the source texture I have that my computer just simply refuses to break down, which is understandable and I don't blame it, shes trying her best. And naturally, having clouds that are no longer static and fixed across the entire world meets a personal want for me, as while I enjoy the first time seeing a specifically nice section of clouds, I kind of don't like it when Ive seen the exact same thing for the 193785th time.

The cons, though, are that each texture level (from 2k to 16k) will be progressively more punishing to a users memory requirements beyond that of cloud packs of a similar resolution. 

My philosophy is is that I don't like single serving visuals that only function in one specific scene in the game. This is why I'm not a fan of hyper exposed nonsense that are so common with TUFX profiles, and why when I did Ad Astra that getting good looking volumetrics (within reasonable performance) was so important, because the transition from launchpad, to flight, to orbit, to deep space for me all too often loses a lot of consistency, even if in a few places it looks really really good.

Thats why when I post screenshots I try to pick ones from a variety of orbits and situations, as that reflects what Im trying to synchronize and going for in general. 

So with all that said, does this sound like the right direction? I dont actually need reassurance per say, as I'm not the self-doubting type even in the throes of depression and toxicity induced defeatism, but even so, I think Id like to just hear some opinions. 

While it would be a heavier memory footprint, the alternatives just don't sit right. Higher res maps are difficult to work with, static, and ultimately the variation tools just don't work as well on global maps as they do on smaller ones without distorting them in unpleasant ways, nullifying the point of higher res. 

And plus, as part of what inspired me today, I believe I have a good method to enable something that the above technically loses me out on. Dynamic clouds in EVE all revolve around the detail texture setting, which basically takes a second texture and overlays it on top of the main texture, literally texturing the first one. 

This has the benefit of making clouds appear to have more depth to them, and is half of what makes the clouds as they are now look like they do. But, to do dynamics, you lose this. 

However, I've found that this doesn't necessarily mean you lose the depth, and the typhoon is an example of that. The detail texture for it is essentially just the exact same texture, but what makes it look good even while next to the other clouds is the quality of the texture itself. It isn't flat and while I still think I could do with a second pass at isolating it, the textures own quality shines through.

So, taking that knowledge, I think we can get the best of both worlds if I front load the texturing that EVE does on the fly into the detail texture itself, allowing for much of the same visual effect, but while also still allowing for dynamic clouds.

I don't know what all EVE does as part of detail textures work, but if it works as a relatively simple overlay then if anything I should actually be able to get even better quality out of it, just because I can leverage gimp to do a more comprehensive job at it, and directly touch up where it falls short.

 

But anyways, thats my long rant. Let me know what you think, and I appreciate the words of encouragement from you all. It was definitely needed and I have since all but abandoned discord, and I'll just have to deal if any more garbage tries to filter through Reddit. 

Currently in GGE I'm 6 (I think maybe 5 now) 16k textures for Jupiter's bands, these have a substantial impact on frames and that's something you should keep in mind. Maybe you could use RGBA cubemaps to lower the amount of textures? For example, 4 16k (98304) textures turns into 6 4k (24576) textures since you can use the RGBA channels individually.  Minor issue is you'll have to color them manually, but from what I've seen here that may end up better in the long run. Also highly suggest checking out my video on UV flow maps as I think it could help here if you're going for dynamics.

As for detailtex idea, probably could work at a scale of 1 or so, you'll just lose what the band originally looks like (if that makes any sense).

Anyways keep up the good work :), lemme know if you want me clarify any of this in dm.

Link to comment
Share on other sites

33 minutes ago, ballisticfox0 said:

Currently in GGE I'm 6 (I think maybe 5 now) 16k textures for Jupiter's bands, these have a substantial impact on frames and that's something you should keep in mind.

Theres a trick to this, I think. Maintex from what Ive seen tends to not benefit as much from higher resolution textures as detail does, unless the specific content (ie, clouds) also must be as detailed as the texture texturing it. 

A flat circle isn't going to really change whether its 1k or 16k,  and thats what most cloud shapes on a map tend to be. Unless you have really high quality captures that actually picks up real cloud textures without washing out to pure white, then using higher resolution maintexs I think should only be necessary to the point that the shapes involved are being conveyed appropriately. A complex shape, like a typhoon, shouldn't be blown up or distorted, and pinning down the minimum resolution needed is where you can buy back some performance with no clarity loss.

I especially think this is the case because even in those cases where quality may drop, it might prove beneficial simply for naturally smoothing out the transitions. Ie, instead of the typhoon just abrubtly cutting off, it somewhat smoothly fades out along the blurrier, lower resolution edges of the maintex.

But thats just for dynamics. 

 

For gas giant bands, which don't really fade like this, I question whether the maintex bands are genuinely benefitting from the higher resolution if you're also using a detail texture to overlay it. Something complex like the Great Red Spot you'd probably want to have both, but a basic cloud band, probably doesn't need that complex of a base shape if its detail texture ends up doing more of the work at getting it to look good.

49 minutes ago, ballisticfox0 said:

Maybe you could use RGBA cubemaps to lower the amount of textures?

While I get how the RGBA cubemaps work and how to make them, are they really saving on the memory footprint? Ie, is the total memory cost dropping by using this or is it still the same? 

Because if it does happen to drop, then yeah, that would be brilliant for what would basically be lossless compression, but if it stays the same then I don't see much use in it for what I'm going to do.

Like I talk about above, I'm thinking not every texture is going to actually have to be at 16k or whatever texture size. 

53 minutes ago, ballisticfox0 said:

As for detailtex idea, probably could work at a scale of 1 or so, you'll just lose what the band originally looks like (if that makes any sense).

Yeah itd have to be at a scale of 1. Going above 1 shrinks the texture and dramatically speeds up the rate at which it fades in and out which can't really be compensated for by detailspeed. 

And, as Im sure you've probably tried at least once, less than one is basically just broken and shouldn't even be a valid input lol. 

Link to comment
Share on other sites

54 minutes ago, G'th said:

While I get how the RGBA cubemaps work and how to make them, are they really saving on the memory footprint? Ie, is the total memory cost dropping by using this or is it still the same? 

Because if it does happen to drop, then yeah, that would be brilliant for what would basically be lossless compression, but if it stays the same then I don't see much use in it for what I'm going to do.

Like I talk about above, I'm thinking not every texture is going to actually have to be at 16k or whatever texture size. 

I’m fairly certain there is at least some benefit as the textures a technically grayscale in game instead of full RGBA, and most clouds are white so you don’t really have to worry about that. And plus, if you need it to be colored, EVE has a covered for that.

So not only are you saving on raw files sizes but you should be saving in game RAM too. (Of course you could just use a standard alpha map to the same effect)

 

54 minutes ago, G'th said:

Theres a trick to this, I think. Maintex from what Ive seen tends to not benefit as much from higher resolution textures as detail does, unless the specific content (ie, clouds) also must be as detailed as the texture texturing it. 

A flat circle isn't going to really change whether its 1k or 16k,  and thats what most cloud shapes on a map tend to be. Unless you have really high quality captures that actually picks up real cloud textures without washing out to pure white, then using higher resolution maintexs I think should only be necessary to the point that the shapes involved are being conveyed appropriately. A complex shape, like a typhoon, shouldn't be blown up or distorted, and pinning down the minimum resolution needed is where you can buy back some performance with no clarity loss.

I especially think this is the case because even in those cases where quality may drop, it might prove beneficial simply for naturally smoothing out the transitions. Ie, instead of the typhoon just abrubtly cutting off, it somewhat smoothly fades out along the blurrier, lower resolution edges of the maintex.

But thats just for dynamics. 

While would say that is true of planets like Earth and Mars, maintex res doesn’t matter, I would say that’s a bit different for gas giants.

On earth you basically just use the maintex to place the detail tex, which is fine. But on gas giants the maintex is your full texture because it covers the whole body, there’s not really a way to make it more detailed from afar with the detail tex, hence why I used 16k textures on the gas giants. (I apologize for poorly wording this)

That being said, 16k textures are probably overkill anyways and 8k would probably be fine.

Edited by ballisticfox0
Link to comment
Share on other sites

Here's my (literal) first attempt at dynamic Earth clouds, and despite the texture actually being distorted they came out, uh, well take a look.

Gxmu8XG.png

jtw7yDN.png

wuqSed1.png

 

And the dynamics are working as expected, so its basically all just a matter of getting the visuals tidy (and finding a realistic speed for which the clouds to change as they traverse the world):

 

qDw7org.png

2eh42D2.png

 

Link to comment
Share on other sites

This is incredible! Wanted to ask, would it be possible for the true sight variant to have a thinner atmosphere like IRL? idk if mods like this can do that lol but it would be good for people like me who like the most realism possible :) 

See the source image

Link to comment
Share on other sites

Just now, Jacktical said:

This is incredible! Wanted to ask, would it be possible for the true sight variant to have a thinner atmosphere like IRL? idk if mods like this can do that lol but it would be good for people like me who like the most realism possible :) 

See the source image

Oh they absolutely can and thats going to be something I'll be working on. 

Won't be exactly as perfect as IRL as theres only so much to be done at 2.5x scale, but there is room for improvement for sure.

Link to comment
Share on other sites

21 minutes ago, G'th said:

Much progress, very wow

This is looking very promising. Great work so far!

2 hours ago, G'th said:

And the dynamics are working as expected, so its basically all just a matter of getting the visuals tidy (and finding a realistic speed for which the clouds to change as they traverse the world):

NOAA GOES has a lot of interesting time lapse data to look at for studying cloud evolution:

https://www.star.nesdis.noaa.gov/GOES/fulldisk_band.php?sat=G17&band=02&length=150&dim=1

1 minute ago, G'th said:

Won't be exactly as perfect as IRL as theres only so much to be done at 2.5x scale, but there is room for improvement for sure.

This is good to hear, I was going to bring this up too. I have noticed that the newer versions of Scatterer in particular seem to extend the atmosphere way too high up, which ruins the illusion that the planet is large when in low orbit.

I think a good example of what it should look like can be found in the opening of this incredible short film, where the atmosphere is a nice thin line on the horizon (sequence starts at 0:35):

Spoiler

You might also find some inspiration in there for some of the outer planets and their moons. ;) 

Link to comment
Share on other sites

13 minutes ago, fleventeen said:

NOAA GOES has a lot of interesting time lapse data to look at for studying cloud evolution:

You know whats great about that link? 

It made me realize theres no reason clouds can't run opposite one another. 

14 minutes ago, fleventeen said:

I have noticed that the newer versions of Scatterer in particular seem to extend the atmosphere way too high up, which ruins the illusion that the planet is large when in low orbit.

Indeed. With old scatterer atmos were generally defined by what it calls config points, which basically adjusts the atmosphere at certain altitudes, and because of how old scatterer worked, I think a lot of configs are overcompensating for something that isn't much of a factor anymore. 

But using those same points is how I could get the atmo to be consistently realistic looking despite the planet not being in the right scale. Right now I have it set up at four points from sea level up to 7500km, and it looks best at higher altitudes where I have it set to more or less match real life. 

Its just the low orbit altitudes that I still need to tweak in tandem with cloud height to get it right. 

Link to comment
Share on other sites

31 minutes ago, fleventeen said:

This is looking very promising. Great work so far!

NOAA GOES has a lot of interesting time lapse data to look at for studying cloud evolution:

https://www.star.nesdis.noaa.gov/GOES/fulldisk_band.php?sat=G17&band=02&length=150&dim=1

This is good to hear, I was going to bring this up too. I have noticed that the newer versions of Scatterer in particular seem to extend the atmosphere way too high up, which ruins the illusion that the planet is large when in low orbit.

I think a good example of what it should look like can be found in the opening of this incredible short film, where the atmosphere is a nice thin line on the horizon (sequence starts at 0:35):

  Hide contents

You might also find some inspiration in there for some of the outer planets and their moons. ;) 

About the scatterer thing, that’s due to broken configs, not scatterer itself

Link to comment
Share on other sites

On 5/5/2022 at 11:53 AM, ballisticfox0 said:

About the scatterer thing, that’s due to broken configs, not scatterer itself

Yeah, I am aware of the broken config issues. It's probably off topic for this thread, but does anyone know if the scatterer default configs have been adjusted to compensate for the newer changes? I did a couple quick tests, and it seems that v0.0835 balloons the atmosphere up higher than v0.0772, which is odd because you would think the default configs would have been updated to reflect the new system.

Spoiler

v0.0772

VO42wgL.png

v0.0835

dH9fPez.png

 

Edit: I'll pose this question in the main Scatterer thread for anyone interested.

Edited by fleventeen
Link to comment
Share on other sites

10 hours ago, G'th said:

You know whats great about that link? 

It made me realize theres no reason clouds can't run opposite one another. 

It turns out atmospheres have layers, much like onions :D

Edited by fleventeen
Link to comment
Share on other sites

Hour and a half to go until Im off work, and then gird your loins folks. My weekend plans are to steamroll Earth to a more or less final state, as I'm pretty much done experimenting. 

Link to comment
Share on other sites

Alrighty folks, some hella progress. No updates on the dynamic clouds today as I needed a fully clouded Earth to work on stuff, but this will give an idea. 

Oh, and we also have some much better City Lights and new Earth Textures. While the land may not seem all that different, it is much more consistent in quality and responds much better to saturation effects from Scatterer and TUFX. The TUFX profile is also going to be a perpetual pain in the ass to tweak, as I'm trying to nail down a consistent quality, so as per usual don't expect it to be final. 

mUnlSqr.png

iiziHf9.png

cESLJJi.png

As5MWyh.png

1mdBDU5.png

34SzBXP.png

zZIhYYk.png

cFajXMb.png

 

AND, we also have some daytime textures for Cities, which you can see occasionally in the other shots, but here's some more dedicated shots of them:

DKgMrz8.png

c8HueiE.png

Link to comment
Share on other sites

1 hour ago, MashAndBangers said:

Earth never looked better!  Real Earth, move aside.

For Scatterer, are you using the latest update?  (and has this question been asked before?)

Yes and Yes :)

Link to comment
Share on other sites

To demonstrate the pretty insane quality we're getting here, in addition to the dynamics shown in the GIF, here's a series of shots taken at 150km in altitude with no zoom other than what you'd be at for actually playing the game. 

En3pcWi.png

hnWv0nP.png

MOGk9JB.png

VGgNwqN.png

bMlMR9U.png

UeYrLOA.png

zY6cWmV.png

zHw39Li.png

6ysjWi2.png

aeoRUyV.png

 

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
×
×
  • Create New...