
Local Content-Style Control for Diffusion-based Image Stylization
ConferenceACM SIGGRAPH Asia Technical Communications, 2026
Computer Graphics Engineer and Researcher
Bringing artistic control to expressive image stylization — from GPU-based filtering and neural style transfer to on-device generative AI.
I lead research and development at Digital Masterpieces, driving research and product development for expressive image stylization apps such as BeCasso and ProFuse. With over 15 years in non-photorealistic rendering, my work has been recognized in both research and industry — including best paper and demo awards at SIGGRAPH Asia and Expressive, and App of the Day features in the App Store.
My contributions span GPU-based mobile image filtering, neural style transfer and on-device diffusion. I received my doctorate from the Hasso Plattner Institute, University of Potsdam, in 2016 and was a post-doctoral researcher in its Computer Graphics Systems group until 2023, working on real-time rendering, computational aesthetics and cartography-oriented visualization.
Currently
With ProFuse, I am shifting my focus to bring local artistic control to generative AI — drawing artists deeper into decisions the models would otherwise make alone.

ConferenceACM SIGGRAPH Asia Technical Communications, 2026

JournalThe Visual Computer, 2024

ConferenceProceedings International Conference on Cyberworlds, 2023
On-device generative AI art app for iPhone, iPad and Mac. Turns photos and text prompts into artworks with local control over content and style — running privately on the Apple Neural Engine.
Photo-to-painting app for iPhone, iPad and Mac. Interactive oil paint, watercolor and cartoon filtering on the GPU, combined with on-device neural style transfer, Custom Styles and Magic Eraser.
Interactive photo-to-drawing stylization on mobile devices — pencil, pastel and ink renditions with fine-grained local control.
Giving users fine-grained artistic control over filtering and stylization — local and multi-level parameterization, art-directable strokes, editable texture and shape, temporal consistency in video, and local content and style control for diffusion models.
Interactive image stylization with an oil paint look — feature-aware color quantization with dominant color palettes and real-time paint-texture synthesis, with per-pixel control through interactive painting on desktop and mobile GPUs.
Non-photorealistic rendering of virtual 3D city models, landscapes and water surfaces following cartographic design principles — the topic of my doctoral thesis.