Iterating 3D models and animation with Opus 5.5 raises questions about LLM taste
The author spent several days tuning a 3D model's textures, animation, effects and layout with Opus 5.5, and was struck by how quickly the model grasped their personal taste.
Added a few more interaction with Opus 5.5. Finishing this makes me wonder if we actually are getting closer & closer to the point where LLMs can pick up the delicate nuance of human taste? I mean to be fair I spent a few days tuning the 3d model, giving feedback about texture, animation, effect, and overall layout. But I've noticed the LLMs ( well opus 5.5 in this case but this can be applied with other frontier models as well) understands me extremely quick. So quick that sometimes I didn't even have time to think and take a step back between sessions (this by itself needs another post tbh). And I have no doubt it'd get even faster in the future. Would I then have a clone of my artist self, only waiting for me to kickstart a project? Would I then be giving only a single prompt to it to know everything I want? I'm not sure... I guess we'll see.
This post quotes another post: “Opus 5.5 adds page-flip interactions and texture to a typography book design”