AI Style Transfer Tool

October, 2025

Overview

This Python-based AI tool was initially developed during a 2-week SCADpro x Electronic Arts Design Challenge to solve a critical production challenge: maintaining consistent art style across multiple concept pieces. As part of the programming team, I created custom style transfer tools that allow artists to apply visual characteristics from reference images to new content, ensuring cohesive art direction throughout the production pipeline. I have continued to develop and expand the tool's capabilities following the challenge's conclusion.

AI Models and Tools

• VGG 19: A Pre-trained AI model used to extract semantic features from images at multiple levels of abstraction including edges and color, textures and patterns
• AdaIN (Adaptive Instance Normalization): A neural style transfer technique that enables real-time transfer of artistic styles between images. This technique is widely used to maintain consistent visual aesthetics across multiple pieces of content, making it ideal for concept art pipelines where style consistency is critical.
• Stable Diffusion XL (SDXL): Large-scale text-to-image diffusion model for high-quality 1024x1024 generation
• IP-Adapter: It transfers style and color from reference paintings by encoding them through CLIP vision and injecting visual characteristics into the diffusion process. This captures artistic qualities like brushstrokes, color palettes, and mood
• Softedge ControlNet: Preserves sketch composition by detecting edges with Gaussian blur, creating soft boundaries that produce natural, painterly results without harsh artifacts.

Key Features

• Style-Transfer Tool: Transform images by applying artistic styles from reference photos, with detailed control over the placement and intensity of effects. Using luminance masks, users can independently adjust how strongly the style appears in bright areas, midtones, and shadows, and also assign different reference images to each tonal range for complex, layered artistic results.
• Sketch-to-Render Tool: Converts hand-drawn sketches into stylized images using ControlNets and Stable Diffusion XL. It intelligently identifies and preserves objects from the sketches through shape-based computer vision and renders them with appropriate colors and artistic style based on a reference image.

HP AI Studio | Python | Hugging Face