Unlocking the Stable Diffusion 3 API: Your Gateway to Enhanced Capabilities

The company Stability AI, behind the development of the Stable Diffusion generative network, has released an API for the next-generation neural network, Stable Diffusion 3. The new version of the neural network was announced in February. The publication of the API suggests that only a few weeks remain until the official release of the base model of the new generation.

The new version of Stable Diffusion significantly differs from the previous version, currently represented by Stable Diffusion XL. First, for commercial use, SD 3.0 will be paid; the monthly subscription starts at $20 for individual artists, illustrators, and entrepreneurs; for large companies, the price will be higher. According to the developers, in the new generation of the neural network, they have finally learned to embed text captions and generate realistic objects, indicating that the new neural network may have been initially optimized for generating marketing materials.

The second difference is censorship. Stability AI materials place a lot of emphasis on the “safety” of the new model, which, according to the creators of the neural network, does not mean the impossibility of creating deepfakes or photorealistic images of weapons. Early leaks mentioned licensing mechanisms that will restrict developers’ legal capabilities to bypass built-in limitations. Similar restrictions were already introduced in the unsuccessful Stable Diffusion 2.0 model.

Finally, from a technical standpoint, not just one, but several base models of different quality and resolutions will be available. It is claimed that the largest and highest quality model with an output resolution of 1024×1024 will require 24 GB of video memory, while the lightest one, with an output resolution of 512×512, is positioned as a replacement for SD 1.5 at a new quality level.

It is too early to judge the capabilities of the new model at this point. Let’s wait for the official release of the base models and trained models from third-party developers, which usually appear one to two weeks after the base model is released.

0 / 5

Your page rank:


Subscribe: YouTube page opens in new windowLinkedin page opens in new windowTelegram page opens in new window

Leave a Reply

Your email address will not be published. Required fields are marked *

Post comment