FastNegative_sdxl

All credit for this one goes to Lykon, This is simply their FastNegative V2 SD1.5 embedding converted to SDXL using FoodDesert's embedding conversion tool.

Of course, this is for use in the negative prompt.

This embedding Improves clarity and removes extraneous details that tend to make generated images noisy, it also helps improve anatomy and adds interest to posing. The effect is subtle, but noticeable. I find it really improves outputs on 2D-focused checkpoints more than 3D-focused ones.