Model description

SDXL LoRA trained upon the Shoshi Zohar costume catalog

While the model was trained with the trigger word "ShoshiZohar", you should also use the word "costume" in your prompt as it was present in all of the BLIP2 generated captions and must be used to produce costume results. I got the best results with cfg scale of 6, LoRA weight scale of 1.0 and Euler A scheduler

WARNING

This model has been trained upon a costume catalog intended for children and their parents. Therefore it tends to generate (fully clothed) photo-realistic images of minors. I've learned that this goes against CivitAI's guidelines and might even get your work unpublished by their moderation team. If you choose to use this LoRA for generating images and you wish to post your results on CivitAI, please make sure it complies with the site's guidelines.

Trigger words

You should use ShoshiZohar to trigger the image generation and also have costume appearing in your prompt.

Huggingface model card

This model is also available in Huggingface as Norod78/SDXL-ShoshiZohar-Lora