NexBlend Illustrious
NexBlend Illustrious is a semi-realistic checkpoint based on the Illustrious model. It focuses on high-density semi-realism, acknowledging the inherent stylistic traits of the base.
• Visual Style: Designed for a semi-realistic aesthetic rather than pure photorealism.
• Ethnicity: Trained on Asian (specifically Korean) datasets, resulting in a distinct Korean-mixed/hybrid facial appearance.
• Technical Focus: Enhanced lighting response and texture density for portrait rendering.
• Enhanced lighting responsiveness and portrait details through a 7:3 merge of a private trained model and an external model.
• Warning: This model has very strong NSFW triggers. Use with caution.
Illustrious 베이스의 특성을 반영한 반실사(Semi-realism) 체크포인트입니다. 완전한 실사보다는 Illustrious 특유의 화풍이 섞인 고밀도 반실사 표현에 집중했습니다.
•외형: 한국인 데이터를 중심으로 학습되었으나, 전형적인 한국인 얼굴보다는 한국인 혼혈 느낌의 마스크가 주로 생성됩니다.
• 스타일: 실사와 그래픽의 중간 단계인 반실사 질감에 최적화되어 있습니다.
• 렌더링: 조명 반응도를 높여 피부 질감과 포트레이트의 입체감을 보완했습니다.
• 자체 모델과 타 모델의 7:3 병합을 통해 조명 반응성과 포트레이트 디테일을 보완했습니다
• NSFW 강도: 성인용 콘텐츠(NSFW) 생성 기능이 강력하게 작동합니다. 프롬프트 입력 시 의도치 않은 결과가 나올 수 있으니 사용에 주의하시기 바랍니다.
Description
First public release
FAQ
Comments (5)
The authors, can you tell me? With any generation parameters, with any samplers, the final image turns out to be faded and with small highlights... Although everything is fine in the preview window... I am using the latest stable version of Comfortable, and there is no such thing with other models...
Thanks for reporting this.
This usually happens when the VAE is missing or incorrect.
Since this model is based on SDXL / Illustrious, please make sure you are using a proper SDXL VAE (such as the standard SDXL VAE or Illustrious recommended VAE).
In ComfyUI, please double-check:
- Your VAE Loader node is set correctly
- The VAE is actually connected to the VAE Decode node
Preview may look fine because it uses latent preview, but the final image is decoded through VAE — so issues often appear only in the final output.
Recommended VAE:
getting an error getting the file though that might just be the site
awesome great checkpoint, too many checkpoints unfortunately produce the same face, great to see something new hit the scene. Checkpoint doesn't seem to follow the prompt. I get very different results despite cfg 7
Thanks for reporting this.
This usually happens when the VAE is missing or incorrect.
Since this model is based on SDXL / Illustrious, please make sure you are using a proper SDXL VAE (such as the standard SDXL VAE or Illustrious recommended VAE).
In ComfyUI, please double-check:
- Your VAE Loader node is set correctly
- The VAE is actually connected to the VAE Decode node
Preview may look fine because it uses latent preview, but the final image is decoded through VAE — so issues often appear only in the final output.
Recommended VAE: