Qwen-Image-Edit Semi-Realistic Detailer (NSFW-Aware)

Recommended sampler & scheduler

• dpmpp_2m_sde + karras
• Lightning LoRA 8 steps (8 - 10)

Qwen-Image-Edit-semi-realistic-detailer (NSFW-Aware)

This LoRA does the opposite of Qwen-Image-Edit-smoothify and transforms images in two ways:

• For anime or cartoonish inputs: If the style of the input image is included in the dataset, it converts them to a semi-realistic style, blending cel-shading into more lifelike textures while adding photorealistic details and 3D-like depth. If the style is not included in the dataset, it converts them to a semi-realistic style, preserving exaggerated features like large eyes while adding photorealistic details and 3D-like depth, blending cel-shading into more lifelike textures while keeping some artistic charm.
• For realistic inputs: Enhances skin details by adding subtle imperfections (pores, textures) and increasing sharpness for a more natural, lifelike look.

It handles various anime styles, but results vary: Flatter, less detailed anime works best with lower strength (0.2–0.5) to prevent over-processing. More defined or 3D-smooth anime benefits from higher strength (0.5–1.0) for a fuller transformation.

Usage Recommendations

• Strength Adjustment:
  • Flat/less detailed anime: 0.2–0.5
  • Defined/3D-smooth anime: 0.5–1.0
  • Realistic images (for detail enhancement): 0.6–0.8

Workflow is embedded in demo outputs

NSFW Warning

This model was trained on datasets including NSFW content to improve body and skin accuracy in revealing or tight clothing. As a result, it may generate suggestive or explicit outputs, especially with certain inputs. Use responsibly and avoid sensitive or inappropriate themes. Not suitable for all audiences.

Trigger words

You should use transform the image to semi-realistic image to trigger the image generation.

Download model

Download them in the Files & versions tab.