huggingface
A small model that separates language understanding from image rendering to generate pictures quickly.
SITE CAPTUREBased on information published on the service's website. Selection and review policy
WHAT IT DOES
What does it do?
Agate is a 260M parameter text to image model. A small transformer reads the prompt and creates a 16×16 region plan, which a convolutional renderer follows to generate the image. A ComfyUI node is available, and the code and model weights are marked with an MIT license.
KEY FEATURES
Key features
- Generate images from text prompts
- Use a transformer to plan a 16×16 region layout
- Use a convolutional renderer to draw the image according to the plan
- Use the model through a ComfyUI node