EXPERIMENT:BUILDINGS -indoors版本FLW_indoors (ID: 589972)

This one is pretty simple again, its been trained on interior images from architecture made by Frank Lloyd Wright, because all images trained were real it will lean into realism on any model it is applied to, and because the dataset is almost devoid of humans you will want to weight up prompts involving people or characters unless they are well known by the model you are using, weighting down the lora will also work. The rooms are mostly from the late 40s through the 60s but it is flexible enough and the checkpoint will bleed in to change the stylization a bit.

triggers:
flwindoors: this is the general trigger, because it isn't a known tag it seems to be weaker and can probably allow more function flexibility when used alone.

room: this is a known trigger and when paired with flwindoors makes the lora work more strongly in its effect on images.

negatives to consider:
fireplace: lots of fireplaces they must have been hip and cool back in the day

painting: seems some checkpoints love filling the walls with paintings, sometimes it is a nice accent but other times theres just way too many of them

when prompting be sure to include furniture for the best effect, chairs tables lamps plants, whatever you want to fill out the scene, window or windows is also a good term to set a mood. other than those most anything you want to prompt should work but may need weighting to come through.

example:
flwindoors, room, furniture, couch, chair, coffeetable, window, nachos and salsa, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up,source_anime, plants

描述:

训练词语: flwindoors,room