Stability AI releases Stable Diffusion 3 Medium, improved hands and faces in generated content

Stability AI has released Stable Diffusion 3 Medium (SD3 Medium). The company positions SD3 Medium as its most advanced open-source model to date, even outperforming Midjourney 6.

According to Stability AI, SD3 Medium is capable of generating photorealistic images from textual descriptions. The model is designed to address challenges previously faced by almost all AI image generators with hands and faces in generated content.

SD3 Medium also incorporates text elements with high accuracy, thanks to its underlying Diffusion Transformer architecture.

The main aspect of SD3 Medium is its accessibility. Unlike some resource-intensive AI models, SD3 Medium is made to run on standard consumer graphics cards. This can help it in broader user adoption. The model also offers customization capabilities through the use of small datasets, as revealed by Stability AI.

Stability AI has also indicated plans to expand its offerings into video and audio generation in the future, as reported by The Next Web.

Stability AI offers SD3 Medium under a non-commercial license for free use. For commercial applications, a Creator License is available for artists, designers, and developers.

If you’re a large-scale commercial user, you can contact Stability AI directly for licensing details. This comes after developers received API access for Stable Diffusion 3 a couple of months back.

In summary, Stable Diffusion 3 Medium represents a step forward in accessible and powerful text-to-image AI generation, especially when it comes to generating hands and faces.