🎨 LoRA Fine-tuned Kontext Model for Anime Character Colorization

Repository: https://huggingface.co/svjack/Kontext_Gray_Anime_Character_Colorize_lora

This repository provides a LoRA (Low-Rank Adaptation) fine-tuned variant of the Kontext image consistency model, specialized in colorizing grayscale anime character art while preserving facial features, skin tones, hair details. The model uses color reference images to achieve high-fidelity, context-aware colorization.

✨ Key Features

Precise Feature Retention
- Preserves facial contours, hair structure, and skin tones during color transfer
- Maintains original shading and lighting consistency
LoRA Efficiency
- Faster fine-tuning with Low-Rank Adaptation, reducing GPU memory usage
Prompt
- Use color image to colorize black and white image

🖼️ Colorization Examples

Character: Monkey D. Luffy (One Piece)

Prompt: Use color image to colorize black and white image

Grayscale Input	Reference Image	Colorized Output

Characteristics Preserved: Straw hat texture, scar under eye, muscle shading

Character: Monkey D. Luffy (One Piece)

Prompt: Use color image to colorize black and white image, with black cloth

Grayscale Input	Reference Image	Colorized Output

Characteristics Preserved: Straw hat texture, scar under eye, muscle shading

Character: Klee (Genshin Impact)

Grayscale Input	Reference Image	Colorized Output

Characteristics Preserved: Dodoco eyes, backpack details, hair ribbons

Character: Oloron (Genshin Impact)

Grayscale Input	Reference Image	Colorized Output

Characteristics Preserved: Feather accessories, armor gradients

Character: Robin Bird (Star Rail)

Grayscale Input	Reference Image	Colorized Output

Characteristics Preserved: Wing feather patterns, beak gradient

Explore more examples:

svjack/OnePiece_Luffy_Kontext_Colorize
note: you can change your cloth color by prompt as your need

Model Details

Base Architecture: Kontext (12B-parameter rectified flow transformer)
Comfy workflow: https://huggingface.co/svjack/Kontext_Gray_Anime_Character_Colorize_lora/resolve/main/flux_kontext_dev_basic_luffy_colorize.json
Model File: my_first_flux_kontext_lora_v1_000008500.safetensors