Fast 0.6B diffusion model aligned with HyperNoise
Generate images from text prompts
Reward-based Noise Optimization for 1-step t2i models