Generating game assets end to end
How far an automated pipeline can get toward usable text and art for a game character.
Recent work shows models can produce individual game artifacts; that is some distance from helping a non-expert developer assemble a coherent character. We are building a pipeline that goes from structured prompts to both written and visual assets, and asking an uncomfortable question about the results: how much of this is genuinely new, and how much is the training data showing through?