Retro Diffusion models are designed specifically for grid aligned, style consistent, limited color pixel art generation. Developed by and trained in collaboration with experienced pixel artists, this family of models delivers beautiful pixel art across a wide range of styles, resolutions, and modalities.
RD Plus
This model is the third generation variant of Retro Diffusion “text to image” models, trained at a native pixel resolution of 256x256 or lower across over a dozen styles ranging from full scenes, to video game maps, to UI icons.
Settings overview
- Text and/or image inputs with support for “strength” value representing the denoise strength of the generation.
- Width and height parameters for precise sizes. Some styles may only support a subset of the visible range
- Input palette for strict color adherence.
- Background removal
- Tiling options for image seam wrapping.
For faster image generation, animations, or tileset generation, see these other Retro Diffusion models: RD Fast RD Animation RD Tile