Only about 5% of generations were close to usable. Seem under-trained. Very prone to creating body horror and seems to have been trained on not-so-diverse men—it has basically no concept of Asian guys.

I feel like this may have fallen into the trap of being overly descriptive in the captions when it comes to positioning. Others' experiments have shown that when doing LoRAs for poses, less is more when it comes to captioning.