Stable Diffusion XL Turbo has just been released, and it’s capable of generating AI images as fast as you can type the text prompts, thanks to a new distillation technology. The latter enables single-step image generation with incredible quality by reducing the required step count from fifty to just one.
By utilizing Adversarial Diffusion Distillation, SDXL Turbo gains many advantages shared with GANs (Generative Adversarial Networks), while avoiding artifacts or blurriness often found in other distillation methods. In more technical terms, SDXL Turbo can generate a 512×512 image in 207ms (prompt encoding + a single denoising step + decoding, fp16).
Our analyses show that our model clearly outperforms existing few-step methods (GANs, Latent Consistency Models) in a single step and reaches the performance of state-of-the-art diffusion models (SDXL) in only four steps. ADD is the first method to unlock single-step, real-time image synthesis with foundation models,” said the team.