Show Lab, National University of Singapore
Show Lab is a National University of Singapore research lab led by Mike Zheng Shou that studies multimodal intelligence: the synergy and conversion among image, video, language, audio and tactile modalities. Its work spans video understanding, multimodal models, video generation, GUI/computer-use agents and embodied/robotic agents.
At a Glance
- Universities and academic AI/robotics researchers
- AI model and developer communities
- Technology companies working on video, multimodal AI, GUI agents and robotics
- Media, journalism, filmmaking and online-content creators
- +2 more
AI Tools by Show Lab, National University of Singapore
(1)
Show-Harness
Open Source VLM Robot Control
Discussions
No discussions yet
Be the first to start a discussion about Show Lab, National University of Singapore
Latest News
NUS announced that Mike Shou was named a 2026 Asian Young Scientist Fellow; the fellowship supports his Video-Language-Action and Video World Model research.
Show-Harness public release: embodied harness, GUMI collectors, plugins, training pipeline, six LoRA adapters and demonstration data.
Show Lab released ShowUI-Aloha for human-taught computer use on Windows and macOS and ShowUI-π for GUI dragging.
Show-o2 released 1.5B and 7B models with video understanding, unified text/image/video generation and training/inference code.
Products & Services
An Apache-2.0 embodied harness that lets vision-language models control robots through a compact semantic action interface. It supports zero-shot frontier VLM control, fine-tuned open models, embodiment-specific interpreters, GUMI browser-based demonstration collection, real robots and simulation; associated models and data are released through Hugging Face.
An open-source, lightweight, end-to-end vision-language-action model for GUI agents and computer use. It offers a 2B model, Hugging Face demo/model, iterative grounding refinement, quantization, local desktop execution, navigation-task support, API calling and integrations with Qwen2VL/Qwen2.5VL.
A human-taught computer-use agent for real Windows and macOS desktops, supporting a human demonstration/recording workflow.
A 450M flow-based vision-language-action model for unified GUI dragging and dexterous GUI control.
Market Position
Show Lab positions itself as an academic, open-research group at the intersection of multimodal video intelligence, generation and action. Its differentiators are unified understanding-and-generation models (Show-o), open video-diffusion work (Show-1/Tune-A-Video), lightweight GUI vision-language-action models (ShowUI), and a shared semantic interface that lets general VLMs control robots (Show-Harness). Relevant comparison points include proprietary frontier VLMs and agent platforms from Google, OpenAI and Anthropic, open VLA/robot-policy projects such as π0.5 and GR00T, and GUI-agent projects such as OSWorld, Mind2Web and computer-use implementations.
Leadership
Founders
Mike Zheng Shou
Assistant Professor with joint appointments in NUS Computer Science and Electrical & Computer Engineering; Presidential Young Professor and National Research Foundation Singapore Fellow. He earned his PhD from Columbia University, interned at a Bay Area startup and Microsoft Research, and was a Research Scientist at Facebook AI developing video AI platforms and models before joining NUS in May 2021.
Executive Team
Mike Zheng Shou
Founder and Lab Director; Assistant Professor, NUS; Presidential Young Professor; NRF Fellow
Former Facebook AI Research Scientist; PhD from Columbia University; prior Microsoft Research and Bay Area startup internships. His research focuses on multimodal video intelligence, video generation, GUI/computer-use agents and video-language-action/robotic agents.
Founding Story
Show Lab was established when Mike Shou joined NUS in 2021 with an NRF Fellowship, which supported him in quickly building an independent research group. The initial vision was to develop video AI that uses the multimodal nature of video for practical assistance, including video understanding, creation, robotics, smart-city monitoring and other real-world applications.
Business Model
Revenue Model
Show Lab is an academic research lab rather than a commercial company. Its outputs are primarily research papers, open-source repositories, open model checkpoints, datasets and demos; the public sources do not describe subscriptions, API billing or commercial licensing for the lab as a whole.
Target Markets
- Universities and academic AI/robotics researchers
- AI model and developer communities
- Technology companies working on video, multimodal AI, GUI agents and robotics
- Media, journalism, filmmaking and online-content creators
- Autonomous-vehicle, AR/smart-glasses and smart-city developers
- Healthcare and scientific-discovery robotics
- Research and development in multimodal AI, computer vision and deep learning
- Video understanding, retrieval, captioning, summarization and recommendation
- Text-to-video generation, video editing and creative-content production
- GUI automation, computer-use assistants and desktop navigation
- Robotic manipulation, embodied agents, robot learning and healthcare/scientific robots
- Augmented-reality/smart-glasses perception and egocentric assistants
- Facebook/Meta, Google and Tencent have used or benefited from Mike Shou's video-AI models according to NUS's Forbes 30 Under 30 announcement; the sources do not identify these as commercial Show Lab customers.