FLUFFX - the fluffy style LoRa experiment

This model is a basic attempt in creating a LoRa with a small set of custom images on a local GPU. It is done by using Kohya and 20 image of white fluffy carpets. Almost default Kohya training setup.

The LoRa needs to be injected with pretty high weight to perform well in the diffusion. Around 1.3 to 1.8 seems best.

This model is also part of a small tutorial series for designers and artist to train custom LoRas. Feel free to have a look at the guide over HERE

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Technical or conceptual suggestions or training setup tips are highly welcome!

Released under love peace and happiness licence :)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Donate for coolness at my PATREON