stabilityai/stable-diffusion · REIMAGINE XL vs. SDv2.1-unCLIP

Hi, I was intrigued by REIMAGINE XL Demo and I read it is based on the open source Stability Stable unCLIP model. However, when I tried the UNCLIP model, results were quite different, and the quality was poor.

This is the reference real image I gave as input:

This is what I obtain with REIMAGINE XL on ClipDrop Demo https://clipdrop.co/stable-diffusion-reimagine:

These are the images I am getting from the stableunclip.py streamlit script on Stability-AI GitHub (https://github.com/Stability-AI/stablediffusion/blob/main/doc/UNCLIP.MD):

Has anyone tried to generate images similar to a reference without the online demo, and found similar issues?