The next frontier in generative video isn’t sharper frames or longer prompts. It’s models that can sustain a character across minutes of streaming generation, without drift, and in real time. Cantina is building toward this from its new Singapore research lab, while NUS Show Lab, led by Prof Mike Shou, has been at the forefront of long-form video generation, identity preservation, and unified multimodal modeling.

Prof Mohan Kankanhalli, Director of the NUS AI Institute, will open with NUS’s broader perspective on partnering with emerging AI startups such as Cantina. Timo Mertens, CTO of Cantina, will share Cantina’s research roadmap across video, speech, and multimodal generation. Timo and Prof Mike Shou will then discuss the live collaboration between their teams, beginning with real-time streaming generation and extending into research areas such as mixture-of-experts architectures, omni models, and autoregressive long-form video generation.

Attendees will leave with a candid view of the technical problems Cantina is hiring and building against in Singapore, as well as a clearer understanding of where collaboration with NUS is expected. Expect a highly technical discussion, followed by Q&A and networking.