BS
Details
Download Files
About this version
Model description
Model Type: LoRA (Low-Rank Adaptation)
Base Model Trained On: FLUX.1
Overview & Focus:
This BS-V7 EP2 LoRA is a refined version, trained specifically on the FLUX.1 base model. It is designed to produce sharp, high-quality portraits with a focus on natural skin detail and cinematic/realistic lighting.
This LoRA excels at generating soft, expressive portraits of Asian women with a clean, aesthetically pleasing look.
Best For: Close-up Portraits, Photorealism, Aesthetic Imagery.
Epoch: 2 (Focused training for quality over quantity)
Steps: 20 (Total training steps)
How to Use:
1. Essential Prompting:
- To activate the style, you must include the LoRA file name in your positive prompt. (Insert your LoRA filename here, e.g.,
<BS-V7-EP2:1.0>).
2. Recommended Positive Prompts:
Look & Feel: realistic photo, detailed skin, soft volumetric lighting, natural blush, subtle smile, looking at viewer, cinematic lighting, ultra detailed, photorealistic, 8k, best quality
Subject: young woman, korean style, k-pop idol, model ( Adjust if your subject is different )
3. Recommended Negative Prompts:
- nsfw, ugly, deformed, blurry, low resolution, worst quality, out of frame, extra limbs, bad anatomy, jpeg artifacts, noise, grain, lowres
4. Optimal Settings:
LoRA Weight: Start between 0.7 and 1.0.
Sampler: DPM++ 2M Karras or Euler A is recommended.
CFG Scale: 5 to 7
Resolution: 512x768 or 768x512.
Hires Fix: Highly recommended for superior detail and sharpness.
License and Permissions (CRITICAL!):
Based on the permissions shown in your TensorArt image:
Credit: MUST provide credit to the creator (You).
Merging: Allowed (Share merges of this model).
Commercial Use: STRICTLY PROHIBITED.
- WARNING: This model CANNOT be used for commercial purposes, selling generated content, or running on generation services. Please adhere to the non-commercial license terms.