光子 AI-可控视频生成研究员
Core
Researching AI-controlled video generation frameworks for interactive scenarios, enabling high-fidelity generation from sparse control signals like keyboard, controller, trajectory, and voice inputs.
Role type
Senior Research Scientist (Multimodal AI & Video Generation)
Builds
Interactive video generation systems with decoupled motion, camera, and pose controls for gaming and simulation.
Domain
Artificial Intelligence, Computer Vision, Interactive Media, Gaming
Deliverable
production ML models
Required skills
Conditional generative models, Diffusion models, Multimodal control, Trajectory/Camera/Pose control, Python, PyTorch, Reinforcement Learning
Preferred skills
Game engine integration, Real-time video generation, Multi-agent collaboration
Technologies
PyTorch, Python, Diffusion Models, RL
Responsibilities
Design multimodal control frameworks supporting diverse inputs; Research high-fidelity generation under sparse control signals; Implement decoupled and combined control for trajectory, camera, motion, and pose; Develop specialized control modules for different game types; Enable low-latency multi-player collaborative generation; Define standardized interfaces for engine/agent teams; Drive joint optimization of generation and RL fine-tuning.
