Stability AI has today announced the release of SDXL Turbo, a revolutionary text-to-image model that sets a new standard for image generation.
Unlike Bing Image Creator that relies on OpenAI’s DALL-E 3, SDXL Turbo uses a technique called Adversarial Diffusion Distillation (ADD) to generate high-quality images in a single step, enabling real-time text-to-image generation.
“SDXL Turbo achieves state-of-the-art performance with a new distillation technology, enabling single-step image generation with unprecedented quality, reducing the required step count from 50 to just one,” the AI tool maker, Stability AI, boasts in the announcement.
Take a peep at how it works in real time:
Quick test with SDXL Turbo, another amazing, super-fast diffusion model. It works right out of the box with @diffuserslib
unofficial demo txt2img and img2img: https://t.co/xdYgZpkVN4 pic.twitter.com/abMET8RuvN— Radamés Ajna (@radamar) November 29, 2023
The tech maker said that they’d tested different image-generating models, namely StyleGAN-T++, OpenMUSE, IF-XL, SDXL, and LCM-XL, and found that SDXL Turbo was the best. It’s faster and just as good as other models that take many steps to generate an image.
Besides, they also rolled out Clipdrop, Stability AI’s image editing platform, that integrates the power of SDXL Turbo’s real-time image generation firsthand. Now live as a beta demo, Clipdrop is currently available for free trial and is compatible with most browsers.
You can take SDXL Turbo out for a spin exclusively on Clipdrop.