An efficient image generation foundation model with a Single-Stream Diffusion Transformer. This demo uses the few-step Turbo variant (guidance_scale=0.0, ~9 steps).