(Changes In Updated Version)

• New dataset and training that won't always put your character's hands on their face like before

• Removal of trigger word, it tends to be much more consistent when there is no trigger word

Sitting Across From concept for Pony & Animagine 3.1 XL

I found this concept was very hard to consistently generate in Pony & Animagine consistently, so I made a LoRA to make it easier. Usually, these models generate with a side-view of the character. This is trained to generate a direct, frontal, eye-level view.

Important: Be careful not to describe legs, pants, feet, shoes, etc... in the prompt, otherwise it doesn't really know what to do. Upper body / arms only.

Trigger words: no triggers words! Just add the LoRA!

Recommended words:

• table

  • if you want the table included

• negative prompt: hands on face

  • if you dont want the model to frequently output with the character's hands on their face