Virtual Influencer Content Creation Pipeline

2025 
To transform Chef Jade from a traditional 3D character into a scalable digital personality, I engineered an AI-powered content creation pipeline for Jade & Zelda’s virtual influencer startup. Using custom FLUX LoRA model training, the workflow generated thousands of consistent, high-quality images while preserving Chef Jade’s signature style across every variation. This pipeline accelerated content production, fueling explosive growth that expanded their Instagram following from under 1,000 to over 9,000 in just two months.
Creative direction by Kiki Wu, with production support from Jieyi Lee.

Category

AI Burger Commercial Test Case

2025
As the technical artist for RiTE Media and Undercurrent XR’s burger commercial test case, I designed custom LoRA training processes and multi-platform evaluation systems. Handling dataset curation, model training, and performance optimization, I worked across AI video generation tools, including Runway ML, Kling AI, and Google Veo 3. This systematic approach delivered hundreds of commercial-quality assets with 30–40% approval rates while establishing reproducible workflows for future campaigns.
Creative direction by Kiki Wu, with production support from Jieyi Lee

Category


Semantic Seeker —
A Dialogue between Randomness and Patterns


2024
“Semantic Seeker,” an interactive installation, explored Taiwanese residential architecture through generative design. For this project, I trained specialized models using LoRA fine-tuning with Kohya on Stable Diffusion 1.5 and developed an image generation pipeline that produced culturally accurate interior design variations. By curating datasets of authentic Taiwanese homes, the system maintained architectural integrity while generating stylistically precise outputs throughout the process..

Category
Gen AI, Dev
              


CONTEXTU(AI) — AI x 3D Pipeline Development


2024
Leading the development of CONTEXTU(AI), I built an advanced AI stack that transforms social media content into architectural design iterations and visualizations through custom workflow integration. The pipeline combined ComfyUI for image generation with 3D production workflows across Unreal Engine 5, Blender, Houdini, and Maya, leveraging fine-tuned large generative models and 3D Gaussian Splats. The result was a seamless bridge between visual storytelling and architectural visualization.

Category
Creative Workflow Integration