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.
models
folder in the earlier extracted folder.The Upscayl wiki has a guide on how to convert pre-existing ESRGAN models!
4x_NMKD-Siax_200k
- NMKD Siax - Universal upscaler for clean and slightly compressed images (JPEG quality 75 or better), based on CX loss + PatchGAN.4x_NMKD-Superscale-SP_178000_G
- NMKD Superscale - Perfect upscaling of clean (artifact-free) real-world images.RealESRGAN_General_WDN_x4_v3
- wide and deep network modelRealESRGAN_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.4xNomos8kSC
by Phhofm.4xHFA2k
by Phhofm.animevideo models for scales x2, x3 and x4.
realesr-animevideov3-x2
realesr-animevideov3-x3
realesr-animevideov3-x4
[!NOTE] 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.