How hard would it be to recreate these visual effects?
polycounter lvl 3
Hello everyone, hopefully I'm asking in the right forum section. This texture here contains a lot of the major visual and lighting effects for a game I'm working on. It's 1024x1024 and I'm looking to recreate it as accurately as possible in 4096x4096. My question is, more specifically than asked in the title, are there any tools for recreating effects like these easily and accurately? I use GIMP and I know it has lots of tools and filters, but I'm not super experienced with them quite yet. I see some things I could recreate using tools but I don't know about most of them. I wanted to see if there was anyone here that was more experienced with GIMP or other programs/tools that knew if I could achieve my goal with tools. One of the reasons I'm emphasizing using tools, is that besides the fact I'm pretty adamant that I want my project to remain faithful to the game's original design anyway, these visual effects are really prominent, so I need them to turn out especially well.
If anyone wants to take a closer look or wants to help, I can share the .dds file. For whatever reason, GIMP blacks out the alpha channel when exporting as a PNG.
If anyone wants to take a closer look or wants to help, I can share the .dds file. For whatever reason, GIMP blacks out the alpha channel when exporting as a PNG.
Replies
The procedural workflow in Designer also allows you to easily tweak things after creation, including resolution, without loss of quality.
I would make each effect in its own graph and then have one graph that brings all individual graphs together into one image.
Since you're concerned about matching the originals well, I'd start by simply AI upscaling your map to 4k. Most of the effects would scale just fine. Some, like gradients and normal maps, will get artifacts, but then you only have to worry about fixing those (with blurs and smudging) or recreating those specific ones.
There are lots of free AI upscaling tools and models, with even more youtube tutorials on how to use them.
The smallest textures look to be 128x128 which should be plenty of resolution. Unless they're being scaled super gigantic.
Also, 4k is super wasteful of texture memory, even compressed. Curious if you could get away with a 2k instead. Really depends on how bad the screenshots look currently.
I'd suggest trying a test of just Bilinear upscaling the texture to 2k, and adding some noise. See how big the pixels of noise get in-game. That should give you an indication if you really need to go to 4k or not.
Some things look fine, others look bad because they're pretty blown up. The game also uses BC1/DXT1 and BC3/DXT5 type compression, BC3/DXT5 in the case of this texture. I'm just a bit of a resolution junkie haha, I really like things to be super sharp and detailed, and I like taking very zoomed in shots of those things.
Low res things not only catch my eye a lot because I like to play in 3840x2160 on a 50 inch monitor, but because the game has a lot of instances where you get close up shots of things, and I can't help but focus on the game's low quality assets at times.
Also, in a game usually the resolution of each texture is carefully balanced against how it appears in game, not just blanket rezzed up.
I've been working on it long enough that I'm fully aware of what the game can do and what it can't, at this point I'm in the process of remaking all my crappy AI upscales by hand so they're actually good.
So in short, I'd love if anyone could give some tips/tricks/advice on how to recreate some of these effects using tools, brushes, and filters etc. so they're nice and precise. Thanks in advance!
Note: I put a black layer behind everything to make things easier to see, the texture normally has a lot of transparency.
EDIT: I did ask about this before, I forgot I could go into my post history on my profile page. I got a little advice but would love any more that could be offered as I still haven't quite gotten the effects to where I want them, and haven't been able to use Substance Designer yet as one person suggested.
I think what would help is for you to post your work in progress, what you’ve managed to recreate yourself, and ask for feedback on it.
I unfortunately don't have any work saved. I never got anything I was satisfied with, and just kept restarting and playing around more with the brushes and filters but could never quite get it right. Not too long before and after this post I played around with it some more. In GIMP, several of the images look like they can be recreated with just the right soft brush, which is what I've been trying but haven't quite got the right one exactly for the very first one, the next 3 look like they'll be a bit easier, but I was trying to finish the first part of the effect before moving on.
As for the bright light effects with the lines of light going out in different directions, I was playing around with the Supernova filter in GIMP and while I got some that were close none of them quite lined up the way I wanted them to.
The bottom line of images are for lens flares, and they're a bit of a mixed bag, the hexagons I think I can work out how to do, but I wasn't entirely sure how to get the circle with the gradient or the rainbow effects quite right, so I didn't quite know where to start there.
Lastly, the colorful squares look like normal maps, and I think they shouldn't be too hard to make in blender, but I wasn't 100% sure if the first one with a hole in it was made by hand or if there was a specific process to bake a normal map with a hole like that. The second one I was just trying to get some more eyes on because I was struggling to tell if it was a very soft sphere or cone shape. And the last one, the one that looks like a World Space normal map rather than a Tangent Space normal was kind of the same situation as the first, there's a little hole in it and I didn't know if there was a specific method for baking it precisely like that or if that was just hand made. EDIT: Removing the transparency does show the World Space Normal Square as more of a cone shape so I think the other ones are probably spheres.
As for the rest, I'm pretty sure I'll just have to hand draw them to the best of my ability. I just wanted to reach out to other artists and see if anyone knew how to do certain effects or if maybe some of them were common knowledge that I'd missed out on. I've gotten a little more experienced with art since I last posted but I'm no expert haha. I've been working on a different project for some time now, so textures took a back seat, but I recently discovered how to fix an issue with how the game renders these effects and they look much better now. So, I no longer remove the texture entirely, and really want to take another crack at shining them up.
Thanks for your time!
Or ... not at all. These are all quite soft to begin with, hence woud barely be affected by scaling them up 4x. And for the ones with a bit of artefacting left, just some simple blurring and perhaps some sharpening would do.
There are online services that allows you image upscaling. Also free ones. It must not be Topaz Labs Gigapixel. Google will lead you.
I still have Real ESRgan installed as a command line tool. It's one of the older AI upscalers. And does still a great job in upscaling by factor 2 or 4. This one could already suit your needs. Not much hassle to setup, quick results. Have a look at the portable versions that are linked here: https://github.com/xinntao/Real-ESRGAN?tab=readme-ov-file
In case you want to dive deeper into AI, ComfyUI is free, open source, works offline. It is the full toy to create AI image and videos. But also allows simple tasks like image upscaling, with and without AI. It is node based. https://github.com/comfyanonymous/ComfyUI/releases
And a simple AI upscaling workflow could look like this. Upscaling by an AI model:
AI upscaling models can be found here. My favourite is still ESRgan. But there are other models for various needs: https://openmodeldb.info
When you have ComfyUI installed, then you could also test if AI creates the images that you need.
You could even use Krita nowadays to upscale images with AI. It uses the diffusion method though, which alters / recreates the image. And i have never toyed around with it. Since doing diffusion upscaling is easier done in ComfyUI when you have it already installed. There is a dedicated node for that, called Ultimate SD Upscale.
Just that i mentioned it, the SD WebUI from Automatic1111 allows image to image generation. With a high res fix feature. That's basically upscaling. That's the second best known offline tool besides ComfyUI to generate AI images or movies. But it has fallen behind ComfyUI, featurewise. And the development speed has slowed down in general. Except this area. Image to image is still a bit complicated in ComfyUI.
One word as a warning. ESRgan as a standalone is not really memory hungry. But when you use ComfyUI or Krita for AI stuff, then you need a good graphics card. I recommend 12 gb nowadays. And more is better. I ran too often in an OOM with my old 8 gb card.
To me these look like clever use of non destructive effects (outer glow, and so on) as well as a bunch of oldschool PS filter combinaisons like the ones people used to share 15ish years ago. Perhaps also some video effects from old school rendering software, like the "video post" in 3DSMax, or cheapo After Effects VFX plugins.
I second Pior's wish here. Give us an example and tell us why you are unhappy with the result.