Z Image 4X UPSCALER by Helaman {this is not LoRA.}
4x photo upscaler which handles jpg compression. This model will preserve noise.
General upscaler: Realistic - Restoration
Training details: AdamW optimizer with U-Net SN discriminator and BFloat16. Degraded with otf jpg compression down to 40, re-compression down to 40, together with resizes and the blur kernels.
Losses: PixelLoss using CHC (Clipped Huber with Cosine Similarity Loss), PerceptualLoss using Huber, GANLoss, LDL using Huber, YCbCr Color Loss (bt601) and Luma Loss (CIE XYZ) on neosr.
ComfyUI/
└── models/
├── upscale_models
└── 4xZImageTurbo_upscaler_255.safetensors
CREDIT: https://github.com/Phhofm/models/releases/4xNomos8k_atd_jpg
Description
License: CC BY 4.0
Dataset: nomos8k
Number of train images: 8492
OTF Training: Yes
Pretrained_Model_G: 003_ATD_SRx4_finetune
FAQ
Comments (6)
How does this get used in a workflow, exactly?
I have shared my workflow, you can use it. there is a node you have to install (I believe it's from rgthree-comfy.)
Link "Load Upscale Model" node your Ksmapler. since Ksampler doesn't have the upscale_model thing, we use the one from rgthree.
Sorry, have not seen the link for the workflow.
Awesome upscaler
Amazing upscaler..thanks for sharing the great work.
extreme retention of every detail without any pixelation..
it's too slow