Stability AI's new Stable Diffusion 3.5 series offers enhanced customizability and performance, boasting three new models designed for varied applications ranging from smartphones to powerful servers.
The most powerful model, Stable Diffusion 3.5 Large, features 8 billion parameters, allowing it to produce high-resolution images with remarkable detail, improving overall problem-solving capabilities.
Hanno Basse emphasized that training involves multiple captions for each image, which prioritizes shorter prompts to promote a wider distribution of diverse image outputs.
Unlike previous approaches that faced criticism, these models aim to generate images with varied skin tones and features without needing extensive prompting.
Collection
[
|
...
]