Qwen Image Edit 2509 Light_restoration

Introduction

Qwen Image Edit 2509 Light_restoration is a LoRA designed to remove existing lighting from images and re-illuminate them with soft, even light. It's particularly useful for creating high-quality datasets for training other lighting LoRAs, by allowing you to "strip" lighting from real photos.

How to Use

• Trigger word: None needed.
• Installation: Download the LoRA file and place it in your models/loras folder.
• Prompt: Use the following prompt: "移除光影,使用柔和光线对图片进行重新照明" (This translates to: "Remove light and shadow, re-illuminate the picture with soft light").
• Companion LoRA: For best results, use this LoRA together with Qwen-Image-Lightning.

FAQ

Q: What's the trigger word?

A: No trigger word required.

Q: Why would I want to remove lighting? Doesn't that make images worse?

A: This LoRA addresses a common challenge in training lighting models. Traditionally, creating a dataset for a lighting LoRA meant finding images without lighting and then trying to add artificial lighting, which often looked unnatural and required extensive trial and error. This LoRA flips that process: you can now take any real photo with the lighting you want to emulate, use Qwen Image Edit 2509 Light_restoration to remove its existing light and shadow, and then use that "clean" image to create a much higher-quality dataset for training new lighting effects. The resulting datasets are significantly better than those created with the old method.

Q: What prompt should I use?

A: The recommended prompt is "移除光影,使用柔和光线对图片进行重新照明". This prompt is fixed and doesn't need modification.

Q: Do I need any other models to use this LoRA?

A: Yes, it's recommended to use this LoRA alongside Qwen-Image-Lightning.

Technical Details

• Base model: Qwen/Qwen-Image-Edit-2509
• Training info: Trained code-free on ModelScope.
• Source: Hugging Face