Stability AI Unveils Advanced Image-Generating Model to Lead the Industry

Stability AI Unveils Stable Cascade: The Next Generation of Image Generation

Stability AI has introduced Stable Cascade, a cutting-edge model for image generation that surpasses its industry-leading predecessor, Stable Diffusion. This new model promises enhanced speed and capabilities, serving as the foundation for numerous text-to-image generation AI tools.

Versatile Features and Capabilities

Stable Cascade offers a wide array of features, including the ability to generate photos, provide variations of the exact image it creates, and enhance the resolution of existing pictures. Users can also take advantage of text-to-image editing features such as inpainting, outpainting, and canny edge. Inpainting and outpainting allow the model to selectively edit specific parts of an image, while canny edge enables users to create new photos using only the edges of an existing picture.

Cinematic photo of an anthropomorphic penguin sitting in a cafe reading a book and having a coffee.

Image: Stability AI

Availability and Competition

Stable Cascade is currently accessible on GitHub for researchers but not for commercial use. Its release comes amidst a growing landscape of image generation models from companies like Google and Apple.

Innovative Architecture and Performance

Unlike Stability’s flagship Stable Diffusion models, Stable Cascade comprises three distinct models that leverage the Würstchen architecture. The first stage, stage C, compresses text prompts into latents, which are then passed to stages A and B for decoding the request.

Comparison of inference time Stable Cascade v other models

By breaking requests into smaller bits, Stable Cascade compresses the request, requiring less memory and fewer hours of training on hard-to-find GPUs. This results in faster performance while delivering better results in both prompt alignment and aesthetic quality. Stable Cascade takes approximately 10 seconds to create an image, compared to 22 seconds for the currently used SDXL model.

Legal Challenges and Funding

Stability AI, known for popularizing the stable diffusion method, has faced several lawsuits alleging that Stable Diffusion trained on copyrighted data without permission from rights holders. A UK lawsuit filed by Getty Images against Stability AI is set to go to trial in December. To help fund its research, the company began offering commercial licenses through a subscription in December.