Beyond CRT: Is It Time for a New Generation of Retro Shaders?

CRT shaders have come a long way, and the community deserves credit for the work it has done. For many players, recreating the look of an old CRT is an essential part of the retro experience. But what if we could take things further?

As someone who grew up with these consoles and experienced the original hardware firsthand, I have to admit that I’ve reached a point of CRT saturation. I still appreciate the nostalgia, but I’m ready for something new. And I suspect there are plenty of other players who feel the same way.

With AI-assisted shader development, could we move beyond simply recreating old displays and start reimagining how retro games look? Imagine advanced lighting, realistic depth, professional color grading with cinematic LUTs, and visual effects that give classic games a fresh look without losing their original artistic identity.

To make this possible, emulator cores may need to evolve too. Where technically feasible, cores could expose additional information about sprites and background layers, including their positions, depth, and rendering priority. This could allow shaders to process different game elements individually rather than treating the entire frame as one flat image.

I’m not suggesting we replace CRT shaders. They should remain available for everyone who loves the authentic look. The goal is to expand our options and explore what modern technology can offer beyond traditional post-processing.

Retro gaming shouldn’t be limited to nostalgia for those who grew up with these games. New visual techniques could also help younger generations discover and appreciate them in fresh ways.

Could Libretro and its emulator cores pave the way for a new generation of shaders featuring AI-assisted development, per-layer effects, advanced lighting, and real depth? What would it take to make this possible?

2 Likes

Why Ai plus shaders?

If AI has to stay in the equation, it is easier to pay tokens and ask it to do a copy of the game with editable assets.

(Not a fan of that, but still…)

1 Like

I think there’s a slight misunderstanding! I’m definitely not talking about generative AI replacing game assets or remaking games—I agree that would take away from the original art.

I mean using AI purely on the development side—to write shader code faster, optimize complex algorithms, and even assist core developers in mapping and exposing sprite and background layers from emulator PPU logic. AI is moving fast in coding and core development, and I genuinely believe it’s going to play a huge role in the future of emulation specs. The original pixel art stays 100% intact; the goal is just to apply smarter, layer-aware lighting and depth to those same pixels.

Fair point regarding generative AI, but that’s not what I’m proposing here.

I’m strictly talking about using AI as a tool for writing and optimizing shader code, as well as helping speed up the work needed on emulator cores to expose individual rendering layers. AI coding tools are becoming the future of software development, and they could make this kind of complex core evolution much more feasible. The original graphics remain untouched, but layer-aware rendering allows modern lighting and depth effects to be applied dynamically.

Where you do you propose things can improve the fidelity and accuracy of the developer intended look past CRT shaders?

Well for shaders I think a useful information could be the layers of the game. Sprites could have slightly different shader settings than background or menu or hud elements, or text. But then we are talking about core specific shaders.

1 Like

Even though developers optimized their art specifically for CRT blendings, they were also constantly striving to create atmosphere, separation, and scale despite severe hardware limits. They used clever tricks like palette cycling, scanline interrupts, and parallax scrolling to fake dynamic scenes.

If emulator cores can expose layers and depth data, we can build a whole new class of smart shaders that take those original artistic intentions further.

Imagine modern deferred lighting where light sources in a level (like torches, laser fire, or magic spells) cast real-time soft light and cast subtle shadows specifically on background layers while highlighting the foreground sprites. Imagine scene-aware color grading that dynamically adjusts contrast or atmospheric tint between foreground action and distant backdrops, or adaptive depth that gives sprawling 2D environments a true sense of scale and atmosphere.

It’s not about replacing the nostalgia of CRT shaders, but rather giving retro games a modern ‘remastered’ pipeline in real time keeping 100% of the original developer art while lighting and rendering the scene with modern spatial awareness

2 Likes

That’s exactly the core idea! Layer-aware shaders processing sprites, backgrounds, and HUD separately would completely change the game. And you’re right—it would require cores to evolve and expose those layers, but the visual potential makes that shift worth exploring.

1 Like

I like the idea of doing MORE than CRT Shaders in concept. As I think indeed we are capable of doing things to make games look even better. RGB looks with Composite blending. etc. hiding masks on solid colors? Conceptually, I’m bankrupt.
But I’m afraid I wander away from this when we wander away from making the games look as high fidelity as developer intended. Personally.

I’m happy to see thinking outside the box!

1 Like

Since you appreciate out of the box ideas, take a look at the Shader artistic-styles section in my repository. You’ll find some really unconventional concepts some great, some decent, and others that ideally need sprite coordinates which emulator cores unfortunately don’t support yet. Still, I’m really happy with the results, especially with precise parameter tuning:

2 Likes

AI definitely helps with shader development, but I would say its biggest flaw is it has no sense in keeping itself grounded. When working with AI it will bring in ideas and techniques from virtually any signal processing domain: RF, audio, communications, video, 3D graphics, etc. Its knowledge of analog video is limited (probably because it’s an obsolete subject). So if you don’t know how to understand what the AI is doing, it’s easy to add ‘hallucinations’ to the shaders.

This is fine for prototyping or if you care more about the end result than the accuracy of the model.

With those caveats in mind, I am more or less satisfied with what AI can do for me. It also helps tremendously with maintanence, allowing solo developers to manage large code bases without too much time.

As for your idea of expanding beyond CRT, I guess that’s a matter of trends. Before fast and accurate CRT simulation became feasible, stuff like this was all the rage. I won’t try to predict future trends, but I will say that decompilation seems to be very popular, and that relates to some of things you’re alluding to: smoother FPS, modern rendering pipelines, sprite layer manipulation, etc.

I still use CRT presets for most 2D games, but remaster-style shaders would be fun to try on a few titles now and then.