Kontext-Five5k-Revert LoRA

Model Description / 模型描述

Kontext-Five5k-Revert is a specialized LoRA trained on the MIT-Adobe FiveK dataset, designed for inverse image processing. While most AI tools aim for "perfect" aesthetics, this model reverses the professional retouching pipeline.

Kontext-Five5k-Revert 是一个基于 MIT-Adobe FiveK 数据集训练的逆向图像处理 LoRA 模型。与大多数追求“完美画质”的模型不同，本模型专注于逆向还原。它能够将经过专业后期修饰、高动态范围的精修图像，还原为具有“原始感”的标准画质，并能模拟常见的摄影瑕疵。

Key Capabilities / 核心能力

• Quality Reversion (画质逆向): Strips away high-end post-processing to restore a standard, unedited aesthetic. / 去除专业的后期处理痕迹，还原图像的标准化未编辑外观。
• Exposure Simulation (曝光模拟): Accurately mimics digital artifacts associated with underexposure and overexposure. / 精确模拟欠曝（Underexposure）和过曝（Overexposure）带来的视觉特征。
• Hardware Simulation (设备模拟): Replicates the visual profile, color curves, and limited dynamic range of mobile phone photography. / 复制手机摄影的色彩曲线、噪点特征及有限的动态范围。

---

🧩 Trigger Prompts / 触发指令

Task / 任务	Prompt / 提示词
Simulate Underexposure / 模拟欠曝	Downsample refined high-quality images to standard quality and simulate underexposure
Mobile Photography / 手机感	Downsample refined high-quality images to standard quality and simulate mobile phone photography
Simulate Overexposure / 模拟过曝	Downsample refined high-quality images to standard quality and simulate overexposure

---

⚙️ Recommended Settings / 建议参数

• LoRA Weight: 0.6 - 1.0
• Base Model: FLUX.1-Kontext-dev
• Sampling: Match standard FLUX settings