Shrunk (PDXL/SDXL) Reviews

If I could, I would give this model 3.5 stars rather than 4. It's definitely a great start and I really appreciate all the work the creator put into creating this admittedly niche LoRA. However, it currently presents a couple challenges.

One is that it seems to lack flexibility. Results appear to be inconsistent across checkpoints: for some it doesn't seem to have much of an effect, and for others it seems to dramatically degrade the quality of the output. (It works reasonably well in NightVisionXL, for example.)

The other issue is the use of the word "miniature" as a trigger, which causes problems because many models understand miniature to mean a particular style of art/diorama that includes dolls or other tiny representations of people and objects. So using this trigger with some models (e.g. Dynavision) just gives everything a toy look.

I understand that the creator probably worked really hard on this and that it was probably challenging to find good training images, so I want to cut them slack. Hence why I'm rounding it up to 4 stars and hoping that it improves with some refinements--because it's a great start.