Downloads · 30 days
0
IzumiKeita/KQube-Illustrious-v1.0
KQube-Illustrious-v1.0 is a text-to-image model from IzumiKeita. Use it when you need an image from a text prompt. It is set up for kqube. The card lists the license as cc-by-nc-4.0.
KQube (.kqube) is a lightweight cubic projection format that improves image quality in diffusion models without modifying their original weights. It is applied at inference time as an additional layer over the UNet.
Downloads · 30 days
0
Access
Public
Updated Jun 29, 2026
Repo size
173 MB
Likes
0
Public
Click a slice to open those files.
.kqube173 MB · 100%
From the Hugging Face model README
KQube (.kqube) is a lightweight cubic projection format that improves image quality in diffusion models without modifying their original weights. It is applied at inference time as an additional layer over the UNet.
This is the first public release, trained on waiIllustriousSDXL_v170 using the ComfyUI-KQube custom node.
.kqube fileComfyUI/models/kqube/[Load Checkpoint] → [KQubeLoader] → [VAE Decode] → [Save Image]
↑
KQube-Illustrious-v1.0.kqube
Recommended settings:
strength: 0.8 — 1.2 (1.0 = full weight)strict_dna: familia (default)| Format | KQube v3.0 |
| Architecture | SDXL (compatible with Illustrious, Pony, NoobAI, etc.) |
| r_cubic | 8 (cubic projection rank) |
| Size | ~165 MB |
| Base model | waiIllustriousSDXL_v170 |
This .kqube is compatible with any SDXL-family model (architecture_family: sdxl):
CC BY-NC 4.0 — Non-commercial use with attribution.
Created by IzumiKeita. Trained with the KQube framework. Civitai Keita_Izumi