This model is trained using about 150 images of cyberpunk vehicles.

It is one of my first LoRAs trained, so I still have a lot to learn, including any shortcomings this model may have. However, it does generate some nice looking vehicles alone in landscape orientation. Sometimes, using an even shorter image (768x400 as opposed to 768x512) can produce better results if vehicles seem too tall.

Some possible keywords to try:

car
van
motorcycle
av (for flying vehicles. Doesn't work terribly well, but there are a few images in the dataset)

sports car
police style
luxury sedan
military style