Stability AI attempts to stay ahead with new image generation AI model

In the ever-evolving world of artificial intelligence (AI), Stability AI has just taken a new step with the launch of Stable Waterfall, a significant advance compared to its precursor model, Stable Diffusion. This latest generation tool claims to redefine market standards with superior performance and speed, marking a turning point in AI image generation.

Stable Cascade stands out for its ability to generate images with unprecedented precision and variation. It can not only create images from scratch but also improve the resolution of existing images. More so, it introduces advanced editing features likeinpainting and theoutpainting, allowing you to edit specific sections of an image, as well as Canny Edge, which transforms the outlines of a photo into a completely new image.

A fascinating example of his ability is creating a captivating image on demand: « Cinematic photo of an anthropomorphic penguin sitting in a cafe, reading a book while having a coffee. » For researchers and AI enthusiasts, this achievement is accessible on GitHub, although its use remains limited to non-commercial purposes for the moment. This limitation comes at a time when large companies such as Google and Apple are also developing their AI image creation tools.

At the heart of Stable Cascade is a technical innovation: instead of employing a single large language model, this approach uses three distinct architecture-based models Würstchen. This method involves compressing text prompts into smaller latent codes, which are then processed to generate the final image. This strategy not only saves memory but also reduces processing times, from 22 seconds with the previous SDXL model to a remarkable 10 seconds to generate an image, providing improved quality and efficiency.

However, Stability AI’s journey has not been without controversy, including lawsuits related to the use of copyrighted data without the approval of rights holders. One notable case is the one brought by Getty Images, scheduled for decision in December. Despite these challenges, Stability AI initiated the marketing of licenses via a subscription model in December, an approach intended to financially support its research activities.

Through these developments, Stability AI continues to make its mark on the AI ​​industry, popularizing the stable delivery method and pushing the boundaries of image generation technology. Stable Cascade not only embodies continued innovation in the field but also raises critical questions about the ethics of data collection and the use of AI.