Okay so this didn't turn out exactly how I planned 😭 but I ended up loving the results anyway lol.

This LoRA is VERY prompt and style-tag dependent, so don't be surprised if you get different results depending on how you prompt it. I highly recommend mentioning the style you're aiming for and using tags like semi realism and soft airbrush shading if you want that softer illustrated look.

If you're trying to avoid realistic generations, stay away from words like realistic or photorealistic. I've also noticed that overly technical prompts can sometimes push the results in that direction.

You'll probably notice I throw these tags into most of my prompts:

*aesthetic, ultra detailed rendering, model-like features, expressive detailed eyes, natural facial harmony, semi realism, soft airbrush shading, nini-tone, sharp focus, high detail, 8K resolution*

I mainly use them to keep the style in that sweet spot without becoming too realistic or too painterly.

You can use the combos from the pinned post and experiment from there! So far this LoRA seems to blend well with other LoRAs and works nicely as a base.

Recommended settings:

• Weight: 0.8–0.9 (you may go lower when combining with other LoRAs)

• CFG: 1–1.2

• Scheduler: sgm_uniform

• Samplers: DPM++ 2M or Euler (tho depending on the version you're using euler may give realistc results)

One last tip: keep your prompts simple and be specific about what you want. In my testing, it responds much better to clear simple descriptions than overly complicated prompts.

I'm still testing and will update things as I learn more. Have fun experimenting!💗