I've had good generation consistency with Auto/Kerras/Polyexpo on the SDEs and Euler A at LoRA: 0.8 / trigger: 1.1. When it comes to Clip, both 1 and 2 produce good results, though I would recommend 2. Overall, I'm finding Euler A gives the best results.

Prompting -- you can edit the metadata, since A1111 or Forge save it as an external JSON, to put the LoRA name in the activation text field and adjust the weights to suit the model checkpoint you're using.

For the most part, while I do use other models if I'm not seeing what I want, I will mainly use my checkpoints: UberBoop (SD), XLBoopV2 (SDXL), Equus(Pony), and Fluxzilla (Flux).

Mira Awad

Born June 11, 1975, Rameh, Israel

Mira Awad was born on June 11, 1975, in Rameh, Israel. She is an actress and writer, known for Zaytoun (2012), The Bubble (2006), and Arab Labor (2007).