BRKN MOSS INNY-PUSSY

This LoRA was trained on 50 images over 4,000 steps. It performs well across most checkpoints, samplers, and schedulers.

Recommended strength: 0.3–0.8

• 0.3: Very subtle influence.

• 0.5–0.7: Good balance between accuracy and flexibility.

• 0.8: Stronger influence while generally maintaining consistency.

• Above 0.8: Increased risk of hallucinations, distortions, and unwanted artifacts.

There is no specific trigger word required, but prompt wording has a significant impact on the results.

The training dataset covers a wide variety of camera angles, lighting conditions, and compositions, including close-ups, extreme close-ups, low angles camera looking up, high angles, upward-looking perspectives, and multiple lighting setups.

If you experience issues with certain angles, poses, or compositions, try lowering the LoRA strength slightly. This usually resolves the problem. Some poses and perspectives are naturally more strongly represented than others.

I'm already planning a few corrections and improvements for Version 2, which should be coming soon.

Other than that, it's a pretty straightforward LoRA to work with.

Have fun experimenting with it!