The veteran artist and developer created a point-and-click adventure tech demo with AI-generated art, an impressive but unnerving achievement.
, a veteran game dev based out of Finland, has recently started sharing early demos of a 2.5D point-and-click adventure game with art and assets sourced from Midjourney and Stable Diffusion, two machine learning text-to-image tools. The project convincingly shows how an individual artist with limited resources can use these controversial, questionably ethical tools to punch above their weight when developing a game.
In his first blog post discussing the prototype, Kemppainen describes wanting to see how far he could push these tools in the creation of a videogame prototype. To that end, he chose to make a 2.5D point and click adventure, allowing easy generation of 2D background art with a more challenging experiment in turning AI-generated assets into textures for 3D models.
Kemppainen used AI tools to generate a number of character concept art sheets, the likes of which you may have seen in a collector's edition artbook. He ultimately settled on a weird old bearded man he got out of Stable Diffusion—the character basically looks like a stock"old coot but make it cyberpunk," someone I can imagine doing a sidequest for in Borderlands. Kemppainen then roughly aligned this art with a 3D model he made and touched it up by hand.
To turn AI-generated location art into a playable space, Kemppainen created 3D collision data in modelling software Blender, overlaid that on the 2D background, and lit the scene in Unity. The result is impressive, but in a discomfiting way. There's still some AI image gen weirdness—our dirty Borderlands Santa's favorite convenience store seems to be named"PAICKTPPARRCTK: MIATIIKK PICAIKANKT," and Stable Diffusion birthed him with a rather incongruous set of pigtails.
It helps the project's credibility that Kemppainen isn't a fabulist producing images of dead-eyed anime ladies with eleven fingers on each hand and claiming that these tools will replace human artists,with experience on the art side of things and credits on multiple games, including Remedy's Quantum Break. All together, he thinks his use of generative tools saved him at least five days of work vs creating similar background and character art by hand.
日本 最新ニュース, 日本 見出し
Similar News:他のニュース ソースから収集した、これに似たニュース記事を読むこともできます。
Hideo Kojima had an idea for a Policenauts sequel 'but it never happened'The original's a great adventure game that almost no one in the west has played.
続きを読む »
Matthew Slater Fights Back Tears in Emotional Presser After Patriots' Season EndsMatthew Slater was very emotional in his postgame press conference following the New England Patriots' loss to the Buffalo Bills in Sunday's Week 18 regular season finale. So, is this it for Slater?
続きを読む »
How to watch Vinland Saga season 2 online for freeVinland Saga season 2 - How to watch online for free
続きを読む »
Game Pass loses We Happy Few later this weekAs confirmed by the Xbox Game Pass app, Compulsion Games' We Happy Few will leave Game Pass and PC Game Pass at the end of this week.
続きを読む »
A Belgian palace remains puzzlingly closedBrussels celebrates its Art Nouveau architecture this year—but Stoclet Palace is hidden from public eyes
続きを読む »
OpenAI developing software to detect ChatGPT textPlus: Apple using fake AI voices to help indie publishers release audiobooks
続きを読む »