Stability AI has announced that its developer platform API now supports Stable Diffusion 3 and its enhanced version, Stable Diffusion 3 Turbo.
The official pitch is that the SD3 model meets or exceeds the performance of industry-leading models such as DALL-E 3 and Midjourney v6 in the area of text-to-image generation, especially in terms of font and cue adherence, as assessed by human preference. The new Multi-Modal Diffusion Transformer (MMDiT) architecture improves weight separation processing for image and language representations, enhancing text comprehension and spelling compared to previous versions.
Additionally, while the model is currently available via API, Stability AI continues to improve the model and plans to soon open up model weights for self-hosting by users with a Stability AI membership. The company has also partnered with Fireworks AI to deliver the service via the fastest and most reliable API platform available, and has committed to 99.9% service availability to ensure stability and reliability for enterprise users when using the developer platform for critical generative AI work.
API Documentation: Stability AI Developer Platform API.
發佈留言