Type | |
Stats | 5,919 |
Reviews | (422) |
Published | Aug 6, 2023 |
Base Model | |
Hash | AutoV2 60C1E43E4C |
WIP MODEL
Aims to remedy the issue in some detail sliders where the "detail" increase ends up effecting the overall composition of the image. Results are similar to Add More Details as the model uses it as a base, which I highly recommend using.
Versions:
Main: The base one
Alt: Slightly better results IMO but 300mb because I hate having room on my harddrive and 256 dim sounded fun
Alt2: Less lighting weight, probably better but 🤷♀️
Current issues:
Still noticeable impact on characters, namely their ethnicity/skin color.
Neither have clip/text encoding training, this isn't really an issue though IMO
I'm working on a better version that should produce better results overall, but for now this is good enough. I might also release a few that target specific areas instead of producing more defined/sharp results, like one for realism