SLIDE 001
≈1 minuteLECTURE 01 · FALL 2026/27
How machines learn
to make moving worlds
The narrated lecture video is no longer hosted on this site.
Use the slide player below with the transcript and captions; the video can be shared on request.
Use the slide player below with the transcript and captions; the video can be shared on request.
LECTURE 01 · PRIMARY SOURCES
Read the systems, not the hype.
Three 2026 sources anchor the lecture’s current model landscape. Read each for what it discloses, what it measures, and what remains outside the evidence.
READ · APR 202601
Evolution of Video Generative Foundations
A technical survey of architectures, training objectives, data, evaluation, and the shift toward multimodal long narratives.
arXiv ↗TECHNICAL REPORT · APR 202602
Seedance 2.0
Evaluation across text-to-video, image-to-video, reference-to-video, audio, and multimodal task following.
ByteDance / arXiv ↗OPEN-SOURCE REPORT · 202603
MiniMax H3
A disclosed audio-video system with Context-IR, a 33B dense omni-modal transformer, latent visual and audio paths, and 2K regeneration.
MiniMax ↗TEACH OR STUDY
