As a fan of Star Srek and Seven as a character, I decided to do some testing on this LoRA.

I wanted to use as few positive prompts as possible while maintaining some consistency in my testing. My negative prompts were mostly a mash of embeddings. I did some tests with and without the negative prompts and I got consistently better results with the embedding.

With this LoRA:

A LoRA strength of 0.6 keeps the images consistent with reduced artifacts. Higher strengths get too many artifacts, and lower strengths look like some lookalike cosplaying as 7of9.

A lower CFG seems to perform a bit better between 4-7.

The HighResFix is needed as it produces significantly better results. The initial number of Sampler steps affects the composition as a whole, but it is less important than the HighResFix steps.

I usually started with 10-25 steps as it didn't seem to improve significantly at higher numbers. I mostly used the DPM++ 2S a sampler, then 30-70 steps (yeah) using HRF. A higher Denoising Strength is beneficial in this case, usually between 0,5 and 0.8 for HRF.

Adding prompts for medical implants and the like slightly reduces the chances of getting a good image, but you will miss out on those nostalgic details without them.

I have quite a few checkpoints (60 or so?), after testing all of my SD 1.5 checkpoints I found that objectiveReality_v20 seems to produce the best results on average for a look closest to the show.

Overall this LoRA is a little unstable, but it produces good results.