Closed Adenialzz closed 4 days ago
Thanks for your interest in out work!
Different prompts present varying levels of difficulty for the model, leading to different editing outcomes. To address this issue, we suggest slightly adjusting the --inject and --timestep, selecting the best results (which is also mentioned in the paper).
OK. I will try more and read the paper in detail. Thanks, nice work.
Hi, thanks for your open-sourcing.
When I tried different prompts (brown dog -> white cat), the original image changed a little. I think it should be kept unchanged, why this happend?