I find significant variation on the results of this lora dependent on the base model so ymmv.

basic prompt:

• view through window of woman in apartment.

• {her face and shoulders visible|her upper body visible|her side profile visible|her back partly visible|only part of her figure visible|}, {,lower body hidden below the window sill|}.

add variations for window size, style, camera angle as you like. training images also tagged with 2 women, open windows, couples etc but tbh, MOSS seems happy with whatever you through at it.

if you want a more long distance view, add some detail about the building structure, eg:

• {multi story|first floor|upper floor|lower floor|second floor|third floor|}, {building struts|}, {brick|} {steel|} {concrete|} exterior {with { glass panels and|} wooden beam accents|}.

• {street in foreground|branches and excessive foliage in the foreground obscuring the view|}.

have a look at the examples and have a play. MOSS prompt adherence is incredible, which makes these kinds of scene so much easier to produce.

Yes, I know, it's frustrating seeing a lora on early access. Just wait a few days and it's free. Or, if you like this Lora and want to support me creating more, please buy-into early access. It's not much buzz and it helps me train more.

I dont make any money from any of this. I dont have a paid-account on Civitai.

This is a hobby for me and the training is paid for in buzz. The same buzz I get from early access downloads, tips, etc... mainly early access downloads. Without that, I have no buzz. No buzz, means no training. No training... well, no loras.

I do this because I enjoy it. I have a highly paid day job and this is just the stuff I enjoy doing on the side, lol. The moment it stops being fun is the moment I stop bothering with it.