Stability AI, the company behind the text-to-image model Stable Diffusion, is now offering a subscription service that standardizes the commercial use of its model for customers and has made changes to it.

The company states that the membership "redefines" the way it grants users the right to use the model for commercial purposes, with the aim of finding a balance between profitability and openness. Stability AI's membership is divided into three tiers: a free tier for personal and research purposes, a $20 per month subscription tier for "creators, developers, and startups with annual revenue under $1 million, funding under $1 million, and less than one million active users," and an enterprise plan. All three tiers provide early access to the new AI models, but only members of the two paid tiers are allowed to use them for commercial purposes.

In a blog post, the company stated, "As our company continues to grow and scale, Stability AI members will play a crucial role in funding the research and development of our future core models."

So far, Stability has been known for its openness. Stability's CEO Emad Mostaque stated that the company is still committed to releasing "open models" that anyone can access the code and weights of. He compared this announcement to Meta's release of Llama 2, calling it open-source because it can be used for free on different platforms, but it does not fully meet the open-source definition of the open-source initiative.

However, posts on the Stable Diffusion forum express confusion about what exactly is covered by commercial use - whether it is only for those who use the model to provide their own generative AI services or if anyone using images generated by Stable Diffusion for commercial purposes must pay a fee. This has become more complicated because AI-generated works have been denied copyright protection so far, making their value less certain.

Mostaque admitted on X that the company is "still working on pricing to keep it simple."

Stability AI has already allowed users to pay for some of its models. It announced the pricing structure for Stable Audio, its text-to-audio AI model, in September.