Core Contributor
LingBot-World-Infinity
A real-time world model for open-ended interaction through camera control, actions, and text.
I build interactive generative systems for images, video, and world models.
Selected projects
First author and lead researcher
MagicQuill Series
Interactive image editing from free-form strokes, color, edge, and layered visual cues.
MagicQuill V1 became one of Hugging Face's most visited editing projects and earned broad open-source adoption. V2 extends the same interaction language with layered visual cues.
Selected publications
First- and co-first-author papers.
Dynamic Typography
Bringing text to life through video diffusion priors.
MagicQuill
Interactive image editing from free-form strokes, color, and edge cues.
3.7K stars · Nearly 8M demo visitsMagicQuillV2
Precise interactive image editing with layered visual cues.
The second generation of the MagicQuill research lineCausalCine
Real-time autoregressive generation for multi-shot video narratives.
Causal video generation · Multi-shot storytellingAbout me
I am a Ph.D. student at HKUST, advised by Prof. Qifeng Chen, and a research intern at Ant Group. My work focuses on controllable generation, interactive systems, and real-time visual models.
I enjoy working across the full stack of a research project, from model training and post-training to interfaces and deployment. Recent projects include MagicQuill, CausalCine, and LingBot-World-Infinity.
- Controllable generation
- Interactive systems
- Image and video editing
- World models
Experience
Ant Group
Research Intern
Interactive image editing, world models, and real-time generative systems.
Jun 2024 – Present
Hangzhou, China
Apple
Machine Learning Engineer Intern
Machine learning engineering.
Jun – Aug 2023
Beijing, China
Education
The Hong Kong University of Science and Technology
Ph.D. in Computer Science and Engineering
Visual Intelligence Lab · Advised by Prof. Qifeng Chen
2023 – 2027 expected
The Hong Kong University of Science and Technology
B.Sc. in Data Science and Technology
Graduated in the top 5%.
2019 – 2023
News
- Released LingBot-World-Infinity, an open real-time world model for infinite and versatile interaction.
- Released CausalCine, a real-time autoregressive system for multi-shot video narratives.
- MagicQuillV2 was accepted to CVPR 2026.
- Dynamic Typography was selected as an ICCV 2025 Best Paper Award Finalist.
- Dynamic Typography was accepted to ICCV 2025 as an Oral presentation.