This is the first LoRA I've ever published on CivitAI and it is intended to reflect the art style, design and atmosphere of the PC game Darkest Dungeon by Red Hook Studios.

For the best performance use the trigger "DarkestDungeon style" or "in the style of DarkestDungeon" (or try something similar - but beware that Flux is case-sensitiv!).
I've rendered all preview images in WebUI Forge using Euler/Simple, 20 - 25 Steps, Distilled CFG Scale 5.0. The LoRA performed best at a weight of 0.5 - 1.0.
If you have problems with text rendering and/or prompt following, please lower the LoRA weight, increase the image size or tweak your prompt!

Have fun - and please share some of your images! :-)

+ AvaCadava +

v2.0 : This LoRA is a wild mix & merge of numerous failed training attempts, that finally lead to something usable. It is a bit "softer" than v1, but much more stable in its results.

v1.0 : Initial release. 10h training on local machine of 150 images on Flux Dev in a resolution of 640. 3000 steps in 20 epochs with batch size 1. Network Rank Dim 4. Only a single keyword was used as caption: "DarkestDungeon"

Search AI Models & Addons

very few results