Open Model Radar · Dossier

GrainScape UltraReal

Style/art-style LoRA by Danrisi, base model Qwen 2.1, no fixed trigger words.

LoRAQwen-Image 2.1

Highlights

  • No trigger words needed; just write a description
  • Total downloads 24.5k · Likes 1.6k · 1 version
  • File size approx. 81.2 MB
  • Platform-flagged as general content

Scope

  • Suitable for: style/art-style direction, used with the Qwen-Image 2.1 base model
  • Suitable for: stacking with other LoRAs on the same base model (lower each weight by 0.1–0.2)
  • Not suitable for: direct cross-base use — a Qwen 2.1 LoRA fed to another base model degrades noticeably

Tips

  • Start testing at weight 0.8; too high causes overfitting
  • If output looks washed out or over-stylized, lower the weight first rather than switching samplers
  • The base model must be Qwen-Image 2.1; choosing the wrong base model is the most common cause of failure

Facts

Type
LoRA
Base model
Qwen-Image 2.1
Org
Danrisi
Downloads
24,539

Links