Controllable Video Generation Research Intern
Core
Research and design multimodal controllable video generation frameworks for game characters and environments, supporting inputs like keyboard, controller, trajectory, and voice.
Role type
Research Intern (Controllable Video Generation)
Builds
Controllable video generation capabilities for multi-player collaborative control with low-latency response.
Domain
Game AI / Computer Vision / Deep Learning
Deliverable
production ML models
Required skills
conditional generation models, diffusion models, trajectory control, camera control, motion control, posture control, Python, PyTorch
Preferred skills
Game AI, game development, real-time/interactive video generation
Technologies
PyTorch, diffusion models
Responsibilities
Design multimodal controllable generation frameworks; Research high-fidelity generation under sparse control signals; Achieve decoupling and combination of trajectory, camera, motion, and posture controls; Develop video generation capabilities for multi-player collaborative control; Collaborate with Engine/Agent teams to define standardized interfaces; Drive joint optimization of controllable generation and Reinforcement Learning post-training.
