Works best at 1.0 but kinda blurry, use around 0.7 for most balanced results.

Use word "liminal" to activate LoRa

Trained mostly on real pictures of liminal spaces. Some of them was upscaled using SUPIR to avoid too much grain and get most liminal pictures.

This is a trial version of LoRa made on a small but high-quality dataset. If you like it, let me know and I'll make a more advanced version. Enjoy!