Stability AI has announced Stable Diffusion 3, the latest and most powerful version of its artificial intelligence model for image generation. SD3 is based on a new architecture and will work on various types of hardware, although like the current models it will be necessary to have a powerful PC in order not to suffer biblical waiting times, since the processing will take place as always locally.
Stable Diffusion 3 promises significant improvements in performance, especially in the processing of complex requests with multiple subjects, ensuring exceptional image quality and improving the creation of texts to be inserted into images, even being able to correct any spelling errors. The model uses an updated “diffusion transformer”, a technique that was pioneering in 2022 but is now used by many models (including OpenAI’s Sora). The new version is now scalable and will combine the “diffusion transformer” with “flow matching”, a new technique that improves quality without excessive hardware demands.
The model suite ranges from 800 million parameters (less than the commonly used SD 1.5) to 8 billion parameters (more than SD XL), with the intent to run on a variety of hardware. Nevertheless, those who rely on current versions know that it is /and will be) necessary to rely on a fairly recent GPU with a good amount of VRAM memory, i.e. a configuration designed for machine learning.
It is impossible to compare these models directly, as none have really been released and all we have are the words of the press release, however the announcement of the new version of Stable Diffusion is already causing some hype among insiders. To date, it is considered as the preferred model for doing any type of image generation, especially since it has few limitations in content creation. Interestingly, the company put security at the heart of its announcement.
Stability AI reiterates its commitment to safe and responsible AI, stating that numerous precautionary measures have been implemented to prevent misuse of the model by bad actors. The company emphasizes that safety is a primary focus at every stage of development, from training to deployment, and that collaboration with experts and the community is critical to responsible innovation. It’s not entirely clear what the company means, but we’ll know more once it’s released in public.
Despite not being available yet, you can already sign up for the waitlist for an exclusive preview.
