• V1: Trained with Adamw8bit optimizer. This version is not an automatic quickfix; you need to prompt what you want to see. Flux is your friend, just tell it what to do

• V2: This version requires less specific prompting; even with simpler prompts things should happen. Same dataset but using Prodigy optimizer

All V2 example images also use the aiai-Glimra stylizer LoRA

LoRA trained on more or less weird anime-style images generated with Dall-E in Bing.

Works best with weirdness, horror and similar themes.

Trigger is aiai-B1ngLucid, but works without it.

If you like what I do, feel free to tag me on social media; I'd love to see your creations.