This LoRa was trained with 600 pics of very odd looking cars from the late 1940s to the mid 1970s. The training pics were not of high resolution, so they result isn't either.

If it's made by me, things are always very experimental. This lora should work as well with realistic checkpoints, as with non-realistic ones.

I used the

https://civarchive.com/models/82098/add-more-details-detail-enhancer-tweaker-lora

and it improved the quality a little bit.

Trigger word is omg_cars

I recommend the LoRa's strength between 0.5 and 1.0,

but it's yours to be more experimental with generation.