This is the base SDXL checkpoint, with the following changes:
DPO Lora merged in at 0.25 model / clip strength
SPO Lora merged in at 0.75 model / clip strength
Baked VAE replaced with SharpSpectrumVAE XL
CLIP-L replaced with ZeroInt's ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors
This is basically an opinionated set of changes I found after a bunch of testing to provide an overall better basis to work off of.
Description
Initial Version.
FAQ
Comments (4)
i got randomly blurred results like hyper was mixed in. Some was sharp, most was quite dark.
There's no Hyper, just exactly what I said there was. Also this isn't really intended to be great for direct inference, it's supposed to be a better building block than the original SDXL Base, as the name suggests. Can't really say much beyond though without even knowing what the prompts were lol.
@diffusionfanatic1173 Then I wonder why i'm getting the blurry spots, I wonder if those models have the hyper in it.
@frankmike I honestly don't really get what you mean TBH.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.