AI Turbocharges Image and Video Generation with Stable Diffusion
Originally Published 2 years ago — by Ars Technica
Stability AI has launched Stable Diffusion XL Turbo, an AI image-synthesis model that can generate images in real-time based on a written prompt or source images. The model uses a technique called Adversarial Diffusion Distillation (ADD) to produce image outputs in a single step, significantly reducing the time required compared to its predecessor. While the images generated by SDXL Turbo may not be as detailed as those produced by the previous model, the speed savings are impressive. Stability AI claims that the model can generate a 3-step 1024x1024 image in about 4 seconds, compared to 26.4 seconds for a 20-step image with similar detail. The model's fast generation speed opens up possibilities for real-time generative AI video filters and experimental video game graphics generation. SDXL Turbo is currently available for non-commercial research purposes, but Stability AI is open to exploring commercial applications.