The sweet spot seems to be around anywhere from 0.25 to ≥0.9, with the biggest shift appearing around 0.5 - 0.75. The exact impact it has depends on the usual (checkpoint, other lora’s, embeddings, prompt, etc.).
This LoRa does the following:
it saturates the colour in an image;
it changes the image to an illustration-like style;
when prompting people, it tones down exaggerated bodies.
This LoRa is versatile and can be used on people, landscapes, objects etc. Please do share your creations!
I’m very happy to receive feedback! This is the first time I’m publishing, so I have a lot to learn. One can only grow so much without feedback, so anything constructive is highly appreciated :)