[Lumina2] Asian Mix Realistic Lora

推荐设置：

Lora 权重：0.3 ~ 1.0
步数：40
CFG：4.0
Shift：4 ~ 6
采样器：gradient_estimation / res_multistep
调度器：sgm_uniform / simple

系统提示：

You are an advanced image generation assistant designed to generate high-quality realistic images, specialized in creating highly detailed, high-resolution photography that precisely matches user prompts, including tag-based prompts. <Prompt Start> ...your prompt here...

简单工作流：

https://raw.githubusercontent.com/hinablue/comfyUI-workflows/refs/heads/main/lumina/lumina2_workflow.json

---

v0.1-rank16

这是非常早期的实验性测试，使用与 [SD3.5] Asian Mix Medium / Large Lora 相同的数据集进行训练，旨在增强整体细节表现，但可能产生干扰性输出，请自行斟酌使用权重。

已知问题：

• 手、脚等肢体表现不佳。

• 无法生成 NSFW 内容。

This is a pilot test of the whole early stage, trained with the same dataset as [SD3.5] Asian Mix Medium / Large Lora. It is used to enhance the overall fineness of the performance and produce effective results. Please use it at your own discretion.

Known issues:

• Poor body expression of hands, legs, and feet.

• Lawless manifestation NSFW contents.

これは初期段階全体のパイロットテストであり、 [SD3.5] Asian Mix Medium / Large Lora と同じデータセットでトレーニングされています。パフォーマンスの全体的な細かさを高め、効果的な結果を出すために使用されます。ご自身の判断でご利用ください。

既知の問題:

• 手、脚、足の体の表現が不十分です。

• 無法な表現、NSFW コンテンツ。

이것은 [SD3.5] Asian Mix Medium / Large Lora 와 동일한 데이터 세트로 훈련된 전체 초기 단계의 파일럿 테스트입니다. 성능의 전반적인 정교함을 향상시키고 효과적인 결과를 생성하는 데 사용됩니다. 귀하의 재량에 따라 사용하세요.

알려진 문제:

• 손, 다리, 발의 신체 표현이 좋지 않음.

• 무법적 표현 NSFW 콘텐츠.