AI as Base for Tweaking

One antidote to the loss of precision might be to use AI to generate things, and then human artists fine-tune and tweak them to match their creative vision. When I experiment with having AI generate some imagined world, I find the result is typically uninteresting, lacking originality, and doesn't quite align with what I see in my mind's eye. It fails to match creative imagination. However, crafting art by hand is a time and resource-heavy endeavour.

Here's where I see AI being useful in creating a base layer. Imagine you can generate characters or other game assets for a virtual world in seconds, creating a rough starting point and then use hard-won skills to delicately mould this template into exactly what you want. I'm inspired by open models like Hunyuan3D to generate 3D assets. However, when you have a creative vision, it's very difficult to articulate that vision exactly in words for a generative model to get exactly right. Prompting alone is not precise enough to make art. Therefore, humans will still need valuable crafting skills.

These models will save hours. If you want background items to fill your game world, you will no longer need to painstakingly craft each item. But if you have a design for a beautiful vase in mind, you can easily generate a foundation mesh and add flourishing touches. Therefore, we should focus on learning (and building) the right tools to edit the work of AI and creating compatible interfaces between prompting and tweaking. Personally, I don't care if AI was used to make art; what I care about is if the imaginator's vision is accurately reflected in the work of art, and it can make me think and feel something.