I trained Sandra in her look of the 80s as LoRA model.

Prompt is simple:
Something with a woman and add the LoRA: <lora:SandraCretuXL:0.85>

Values between 0.75 and 1.0 work best.

I strongly recommend:
- Upscaler 4x-UltraSharp
- GFPGAN Face restoration (Yes, better keep ADetailer disabled)

A video clip I made, using this LoRA:

Another music video: