Stable Diffusion 3 is now available via API, providing access to developers

Open source generative artificial intelligence startup Stability AI Ltd. offers developers its most advanced next-generation text-to-image artificial intelligence model, Stable Diffusion 3, through an application programming interface.

Today's decision comes after the stable version of Diffusion 3 has been in preview for just two months since its release in mid-February. Its API availability will allow developers to integrate it into applications and access powerful image generation capabilities. The company also announced that the Stable Diffusion 3 Turbo model, a fast version of SD3, will also be available to developers via API.

Stability has built SD3 on a new architecture that aims to improve the accuracy of word generation and spelling in generated images. One of the problems many models face is that they tend to generate gibberish when asked to spell words or phrases in scenes. Developers of text-to-image conversion models have struggled with this.

To solve this problem, Stability developed the Multimodal Diffusion Transformer, or MMDiT, architecture, which uses a separate set of model weights for images and language. According to Stability, this has greatly improved the model's ability to produce clear and accurate writing on rendered images.

While the model is available through the API, it is not yet available to developers in an open release, Stability reported. "We are constantly working to improve the model in anticipation of its open release," the company said. No timeline has been given for when it will become available for self-hosting with a Stability AI membership, but Stability said it will happen soon.

To ensure that Stable Diffusion 3 and Stable Diffusion 3 Turbo are delivered via APIs with the best performance, Stability has partnered with Fireworks AI. Fireworks is a high-performance API platform that delivers enterprise-grade service with 99.9 percent uptime.

Beta version of friendly chatbot Stable Assistant

Stability has also announced that it is starting to invite a limited number of users to participate in an early beta release of its Stable Assistant, which utilizes Stable Diffusion 3. The company describes the assistant as a "friendly chatbot" powered by text and image generation technology, as well as SD3 and Stable LM 2 12B, a language model released earlier this month.

It works similarly to how OpenAI's ChatGPT Plus integrates with DALL-E 3, and is capable of generating images during a conversation. As a result, users can ask it to generate images and then refine them by simply talking to the chatbot as a creative assistant, suggesting a new way to create images rather than giving one prompt and trying to refine it to create the desired image.

The chatbot opens up other possibilities for users, such as providing images for writing projects, helping to create character portraits, slides, and other visuals to enhance content.