A new study from Google AI Research analyzes how diffusion models generate creative imagery. Researchers mapped the latent space to identify where novelty emerges during the denoising process. This technical deep-dive reveals the specific mathematical triggers for visual variation. The findings help developers tune model temperature for more predictable artistic outputs.