A style LoRA trained on World of Warcraft screenshots covering a variety of subjects.
Please note that the LoRA does not know all in-game races - I will be improving the dataset to try cover them properly.
📜 Recommendations
Begin each prompt with the phrase:
wow screenshot ofIf you enjoy the model, please share your generations and leave a rating - I’d love to see what you make with it!
Any BUZZ tips would greatly help with training more models too.
Feel free to send me ideas for more LoRAs.
Thank You ❤
Description
First version
FAQ
Comments (2)
Looks good. Would you mind sharing your training params / what software you used?
Thank you, this was trained on AI Toolkit with the defaults for Z Image Base, except for toggling on Do Differential Guidance and Cache Text Embeddings. The epoch I used as the final was for step 4000. The dataset itself was around 350 screenshots with captions generated by Gemini Flash 3 (mostly environments screenshots which is likely why it's tougher to get characters looking correct).