This is the "Illustrious XL personal merge" model (specifically: v30Noob10based), merged with an additional model of my own creation (Illustrious 0.1 trained on aesthetic background images, at 512 resolution. (PM_Aes3k_512)
The raw trained Illustrious model is also provided for your merging needs. (Aes3k_512)
Add 'scenery' to your prompt for max effect.
Having been unsatisfied with backgrounds in models since the aesthetic masterclass that was SD1.5 and it's derivates I wondered, what if the constraints of the model were actually what caused it to be so good? So I decided to try a previous experiment of mine (aesthetic backgrounds) at 512 training resolution instead of 1024. To my surprise, it actually worked better.
I encourage other more experienced model creators to try this technique, this is just a proof of concept I did quickly.
描述:
This model is not very good on its own. Merge with other models to improve backgrounds
训练词语:
名称: illustriousXLPersonal_aes3k512.safetensors
大小 (KB): 6775430
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
