As a test of the new workflow, I created a two-stage configuration consisting of Stage 1: Anima Turbo and Stage 2: Pony Realism. The key point is to perform i2i in Stage 2 with a denoise value of 0.25.

By using Anima—which is based on the Qwen architecture—for Stage 1, we leverage ViT for scene understanding. Stage 2, Pony Realism, is used to add artistic style.