AI converts ASCII game to real-time AI-rendered graphics – Thunder Lizard ASCII visuals transformed, but latency and consistency need improvement

Software engineer Jeff Schomay has been experimenting with techniques to transform game visuals using artificial intelligence-powered rendering. In a recent blog post, he shared screens from his Thunder Lizard “old-school ASCII RPG style game” and you can see various AI models reconstructing the title’s basic character-based visuals into far plusher full-motion graphics. However, there are quite a few compromises needed to achieve a ‘playable’ 10fps and 1ms latency AI visual enhanced game.

The current game

You can play Thunder Lizard, using the current ASCII rogue-like engine visuals, right now, via that link. In brief, you use the cursor keys to navigate around the prehistoric landscape while attempting to catch and east smaller dinosaurs, while avoiding larger ones. You grow as you eat smaller prey, but you have to flourish enough to become the most powerful dinosaur before the volcano destroys the entire landscape.

Wouldn’t this game be more fun with snazzy pixel art visuals? Schomay must have thought so, too. And you can see various samples of AI-rendered alternatives to the ASCII throughout his blog post.

Cake and eat it – not possible

As Schomay put it in his blog post, there is a “need for speed” in gaming, that isn’t yet met by many AI renderers. At least concerning what is available to indie devs.

We get to see some fantastic looking prehistoric landscape renderers, which would really add something to Thunder Lizard. However, the software engineer couldn’t coax better than “a four-second latency, it was unplayable” from the most handsome AI game graphics renderer.

The compromise

You can read much more about Schomay’s efforts with various AI models in the linked blog, but for actual playability the choice was ultimately Fal.ai. This model beat the other contenders for its “very fast generation times, source image adherence, and decent look and feel,” according to the software engineer.

Though what you see is an (AI) engine capable of just 10fps and 1ms latency, it’s the best option, according to Schomay’s efforts, in this case.

Pixel peepers might feel a little let down by the chosen AI visuals, which eventually won out. However, Schomay ends his blog on an upbeat note, stating that “it is amazing to see AI image generation happen in real-time.” He is also excited about the future, but hopes some strides can be made in frame to frame consistency.

We've previously reported on the use of AI in developing and enhancing games and graphics. Nvidia shared a lot of RTX technology reliant AI-enhanced development tool information earlier this year.

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button!