Hey everyone!
Just popping in to ask for some advice from the community. I’m a newbie in game development and I’m curious about how you all go about getting your game assets. I’ve been messing around with Luma and Meshy for making 3D models, and they’ve been surprisingly not bad. But I’m kinda torn on whether diving into AI tools would make things easier or more of a headache. If I want to quickly build my small game, what tools do you all recommend?

theish said:
But I’m kinda torn on whether diving into AI tools would make things easier or more of a headache.

At work, we have an arist that uses AI-tools to assist him in developing assets, mostly for images but occasionally modesl. These tools in their current state require a lot of supervision, and still knowledge of the processes of 3d modelling. AI might give you assets that have problems - be it with weird surfaces, normals, coordinates; maybe they are too high poly, too low poly - or it could just produce results that are unrealistic and has certain errors that humans won't typically make. Based on their experience, I'd say that using AI to fully make art for your game is not yet a valid option. It can be of help, though you still need your own artistic eye, and a lot of patience, to get the results you want (talking long run - it's easy to get something initially that looks good, but an entire game needs a lot of different assets and problems might show later on).

I think “AI Tools” is too vaguely defined. People have redefined AI again and again and again over the decades, currently trending in use as generative systems based on machine learning. That's certainly not the only meaning of the term, and it can mean (and does mean) many other things as well.

Art tools are filled with technologies that have historically been considered “AI” over the years such as intelligent scissors algorithms that follow contours in an image for rapid precise selection, adaptive color balancing, and automatic adjustment of levels. Quick select, automatic select and mask, object aware refinements, all AI. Dynamic scaling, such as resizing an image where it squishes areas with few features rather than a uniform bilinear filter have been considered “AI”. Adaptive level of detail systems are AI, adaptive mesh refinement is AI, adaptive sampling is AI, many skinning algorithms automatically fit a mesh to a skeleton using an AI system. Vectorization models are an AI, image registration is an AI, backfill is AI-driven, content-aware fill and patch and move, edge enhancements, automatic image stitching, vanishing point detection, smart sharpen, all are AI tools.

Same with programming tools, software analysis tools, packaging tools, compression algorithms, a tremendous number of them have been labeled as “AI” over the years. Search trees are AI, pathfinding is AI, NPC behavior is AI, auto-targeting weapons are an AI. Animations and inverse kinematics to work out object paths are AI. We are surrounded on all sides by the term “AI”, so much it is almost meaningless.

In general if a tool helps you create assets that look better, work better, fit your game, or otherwise improve your life as a game developer, and if you've got proper legal rights to use it, go for it.