AI Work
Over the past three years I have explored a wide range of AI based workflows. It is difficult to present all of them here, so I selected a curated body of work that feels most relevant and compelling.
A central focus of my practice is digital avatars. I began researching this during my thesis in art school and continue to develop it. I work extensively with LoRAs to train and refine specific faces, identities and stylistic traits. This allows me to build consistent and recognizable characters that function across different scenes and contexts while maintaining coherence. LoRAs also offer control over subtle features such as age, expression, surface quality and photographic realism, which makes them effective for character driven work and AI influencer concepts.
Another key area of interest is the automation of abstract brainstorming and image generation using ComfyUI with minimal input. I often provide only a few prompts that are expanded by a large language model such as Gemini. These systems can process reference images to guide composition, materiality and mood. I focus on shifting more authorship into system design by building node based pipelines that generate variations, reinterpret references and produce unexpected visual directions.
Alongside 2D image generation I explore AI supported 3D workflows. This includes using generated imagery as a base for modeling and look development in Blender, training LoRAs on specific products or faces and integrating them into 3D scenes. I also combine procedural geometry with AI generated textures. I am interested in the relationship between generative systems and structured 3D environments where controlled lighting and physically based materials meet probabilistic image synthesis.
At times I deliberately push models toward abstract outputs to test the limits of representation and visual coherence. The balance between control and unpredictability defines much of my current research.
A central focus of my practice is digital avatars. I began researching this during my thesis in art school and continue to develop it. I work extensively with LoRAs to train and refine specific faces, identities and stylistic traits. This allows me to build consistent and recognizable characters that function across different scenes and contexts while maintaining coherence. LoRAs also offer control over subtle features such as age, expression, surface quality and photographic realism, which makes them effective for character driven work and AI influencer concepts.
Another key area of interest is the automation of abstract brainstorming and image generation using ComfyUI with minimal input. I often provide only a few prompts that are expanded by a large language model such as Gemini. These systems can process reference images to guide composition, materiality and mood. I focus on shifting more authorship into system design by building node based pipelines that generate variations, reinterpret references and produce unexpected visual directions.
Alongside 2D image generation I explore AI supported 3D workflows. This includes using generated imagery as a base for modeling and look development in Blender, training LoRAs on specific products or faces and integrating them into 3D scenes. I also combine procedural geometry with AI generated textures. I am interested in the relationship between generative systems and structured 3D environments where controlled lighting and physically based materials meet probabilistic image synthesis.
At times I deliberately push models toward abstract outputs to test the limits of representation and visual coherence. The balance between control and unpredictability defines much of my current research.