Optimized Diffusion of Olfactory Representations

ODOR Is the world's first text-to-scent model, intended to be able to synthesis any particulate down to a size of 0.0004 μm. Any gas or fine particle mist is theoretically capable of being prompted for.

Trained on over 1 billion smells, ODOR excels not only at creating known smells, but by combining natural language prompting ODOR can be used to design scents never before whiffed.

ODOR was designed to run in the same WebUIs as SD with minor tweaking and can utilize any standard commercially available consumer-grade matter printer available in most homes.

Open source.

ODOR is intended to be expanded upon, and can utilize standard LoRAs to create specialized scents that otherwise may not be in the base model,



We look forward to seeing what the community creates, so that we may all breathe better together!