This model is a Lora trained based on the Qwen Edit model, mainly used for character lighting and image composition.

The core function is to effortlessly match the subject with perfectly matched lighting effects to the background environment while maintaining character consistency.

All you need to do is upload a roughly composed image of the character and background, then click run to generate a high-quality image where the character and the scene environment naturally blend.

This model is widely applicable. For images where real people/CG characters/animals/sculptures/IP images, etc. do not blend well with the environment, it performs excellently. Additionally, it has low requirements for prompts; you only need to upload an image to generate results, making it very suitable for designers to quickly explore creative ideas and preview effects.

Fixed prompt: Adapt and adjust the lighting of the character, optimize the character's shadow texture and highlight position based on the lighting logic of the background environment. Throughout the process, no modifications are made to the background image, while strictly preserving the original appearance, hairstyle, clothing details, and facial expressions of the character, achieving precise matching of the character's lighting with the background environment. (Simply input directly)

Recommended weight: 1

Recommended sampler: euler  Recommended scheduler: simple