KINOMOTO.MAG

Stable Diffusion 3 API Unleashed

╱

Stability AI proudly introduces its highly anticipated Stable Diffusion 3 generative AI model after a brief preview period. Accompanying this release is a preview of the “Stable Assistant” chatbot technology. Stable Diffusion 3, initially previewed in February, now officially joins Stability AI’s developer platform as an API.

This integration allows Stable Diffusion 3 to seamlessly integrate into various applications, empowering developers with its advanced text-to-image generation capabilities. Alongside the base model, Stability AI introduces the Stable Diffusion 3 Turbo model, enhancing AI-generated image creation.

With Stable Diffusion 3, Stability AI implements innovative machine learning and AI techniques, enhancing image generation quality and typography. Emphasizing user-centricity, Stability AI has implemented safeguards within the API to ensure responsible usage, continuously refining them based on user feedback.

While Stable Diffusion 3 is now available via API, Stability AI is actively developing an open model release to democratize access to generative AI technology. Christian Laforte, Stability AI’s CTO and interim co-CEO, reaffirms the company’s commitment to openness and transparency, promising the imminent availability of model weights for self-hosting with a Stability AI membership.

*^{_{Created by Kinomoto.Mag with Mid-journey}}*

Some feedback from today on LinkedIn:

“Firstly, the model is only accessible through a developer platform, though ZHO has provided a convenient node for image creation in ComfyUI.

Secondly, the model isn’t free, which was somewhat expected given the recent drama.

However, the disappointment doesn’t end there. Images shared on Reddit suggest that the model was rushed in development, and currently, the quality is not impressive. In fact, the base SDXL model surpasses SD3 in terms of output quality.

While prompt consistency is maintained, the overall quality, especially human anatomy, is subpar. Additionally, this model is the most censored version to date.

I had high expectations and this release, hyped as it was, feels underwhelming. Here’s hoping that better SD3 models will emerge from the community, but this initial release doesn’t inspire much optimism.”