You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This repository will contain extra models that users can download and use in Upscayl. Upscayl v2.5 comes with a Custom Models option where users can select a models folder and use their own imported models.
RealESRGANv3 - Lightweight and faster versions of the default model, with (very slightly) worse quality.
RealESRGAN_General_WDN_x4_v3 - wide and deep network model
RealESRGAN_General_x4_v3
unknown-2.0.1 - "Unknown" - We still don't know what this model is, but we accidently included it in place of Ultrasharp in v2.0.1 and @royal-rigolo liked it, so here it is!
uniscale_restore by Kim2091.
4xLSDIR by Phhofm.
4xLSDIRplusC by Phhofm.
4xLSDIRCompactC3 as a SRVGGNET (Compact) model, meaning inference times should be faster, by Phhofm.
Upscayl will use the model file name to detect the scale that it supports. If the model file name has 'x2' or '2x' in its name, Upscayl-NCNN will set the target resolution for upscaling accordingly. This doesn't affect the post-processing scaling option, this only affects the actual upscaling via ESRGAN.