When
AMCC 5160 · Fall 2026/27
AI-Driven Animation
& Video Generation
An art-and-technology course for creating, directing, understanding, and critically rethinking generative media—from images to interactive worlds.
Course at a glance
Where
MIL13002
Credits
3
Class size
50
Course description
This course treats generative AI as both a technical system and an artistic medium. Students learn how contemporary image, animation, video, audio, and world models work; use them to direct purposeful creative experiments; situate the work in visual culture; and evaluate authorship, provenance, consent, bias, and audience experience.
Understand systems
Build mental models that outlast individual products.
Direct experience
Connect technical choices to intention and audience.
Evaluate evidence
Make limitations and responsible use part of the work.
One medium · two inseparable lenses
See the system.
Direct the experience.
Lecture + discussion + lab
Art lens
Intention becomes form.
- Visual language, composition, color, time, sound, and rhythm
- Art direction, performance, narrative, interaction, and audience
- Reference, context, authorship, critique, and exhibition
Technology lens
Capability becomes control.
- Data, representations, transformers, diffusion, and flow
- Conditioning, editing, adaptation, temporal modeling, and agents
- Evaluation, reproducibility, provenance, safety, and limits
By the end, students can
explain a pipeline · direct a coherent work · build and document an experiment · critique with evidence · account for AI use · present a refined project
Assessment
A clear path from sharing to showcase.
Letter grades · 100% total
Weeks 2–11
Weekly presentation
10%20% each
Assignments 1–3
60%Nov 17 or 24
Final presentation
30%Assignment deadlines Details will be announced in class and on Canvas.
Weekly presentation10%
Share current
or past work.
Around 10 minutes. Bring one artistic, design, research, or technical project and connect it to the course.
Live capacity by weekChoose a slot →- Explain its context, creative intention, and technical approach.
- Share one lesson, limitation, or question.
- Sign up for one of five slots in Weeks 2–11.
Final presentation30%
Refine one
assignment.
Choose one of Assignments 1–3 and present it as your strongest refined showcase. Ten minutes plus critique and Q&A.
Live capacity for both datesChoose a slot →- State the problem, audience, or artistic intention.
- Show evidence, one representative failure, and your own contribution.
- Explain AI-tool use, iteration, and what should be tried next.
Tentative weekly schedule
Thirteen weeks.
One evolving medium.
Open a week to see its focus.
01Sep 1 · This weekFrom generative media to multimodal simulationSeeing + systems＋
Focus
How images become time, systems, and worlds. Technical mental models meet visual language, authorship, critique, provenance, and a first controlled experiment.
02Sep 8Representation, visual grammar, and unified multimodal modelsForm + representation＋
Focus
Pixels, latents, and visual tokens alongside composition, framing, color, and reference. How representation shapes what artists can edit, preserve, and vary.
03Sep 15Transformers, diffusion, flow—and the aesthetics of variationModels + chance＋
Focus
Autoregressive prediction, denoising, diffusion transformers, flow matching, guidance, seeds, and sampling—paired with rule-making, chance, seriality, and selection.
04Sep 22Image generation, editing, and art directionImage + direction＋
Focus
Subject and style control, reference grounding, masks, iterative editing, visual prompting, and agent-assisted workflows. Build a coherent visual system—not a bag of adjectives.
05Sep 29Adaptation, creative workflows, and authorshipCraft + workflow＋
Focus
LoRA and adapters, distillation, quantization, open versus hosted models, node and code pipelines, dataset curation, reproducibility, appropriation, and artistic voice.
06Oct 6Video foundations: time, motion, and cinematographyTime + motion＋
Focus
Video latents, temporal attention, motion representation, camera control, shot scale, blocking, rhythm, montage, and current open and hosted model families.
07Oct 13Controllable motion, choreography, and video-as-promptControl + choreography＋
Focus
Identity, pose, trajectory, depth, optical flow, and reference-video semantics—applied to performance, choreography, camera movement, continuity, and failure diagnosis.
08Oct 20Native audio-video, performance, and real-time charactersSound + performance＋
Focus
Synchronized sound, dialogue, lip-sync, expressive avatars, low-latency generation, voice, soundscape, performance direction, and cross-modal evaluation.
09Oct 27Critique, evaluation, provenance, and responsible releaseCritique + responsibility＋
Focus
Human preference and technical metrics meet studio critique. Prompt alignment, consistency, physics, synchronization, model cards, consent, copyright, and disclosure.
10Nov 3Interactive world models and explorable artWorlds + interaction＋
Focus
Action-conditioned prediction, explorable environments, persistent scenes, real-time simulation, installation, games, and physical-AI data. Case studies include Genie 3 and Cosmos 3.
11Nov 10Video reasoning, creative agents, and speculative practiceReasoning + speculation＋
Focus
Generation as a reasoning tool, multimodal agents, limits of treating video generators as world models, speculative futures, exhibition strategy, and presentation coaching.
12Nov 17Final presentations IShowcase＋
Focus
Students present one selected assignment as a refined showcase, followed by peer critique and discussion.
13Nov 24Final presentations IIShowcase＋
Focus
The second showcase session, peer critique, course synthesis, and future directions.
Current frontier · Fall 2026
The syllabus follows capabilities—not product hype.
Unified multimodality
Understanding, generation, and editing increasingly share one context and one workflow.
Native audio-video
Systems such as Veo 3.1 generate synchronized dialogue, ambience, effects, and moving images together.
Video as reasoning
Recent work tests whether generative video models can also segment, edit, infer affordances, and simulate tool use.
Interactive worlds
Genie 3 and Cosmos 3 push from passive clips toward action-conditioned, explorable, and physical-AI environments.
References & further reading Open list ＋
Presentation sign-up
Choose a slot.
Bring work worth discussing.
Use the access code shared in class to see live remaining capacity and reserve directly on the course website—no account required.
Choose a presentation slot →Weeks 2–11
See the live number of slots left for every week, then enter your family name, student ID, and topic.
Nov 17 or Nov 24
Compare live capacity across both final dates and reserve your preferred slot.
Reserved slots show only family names and the last four ID digits. Full student IDs are not stored.
Teaching team
People, not prompts.
Instructor
Harry Yang
Assistant Professor, AMCyangharry@ust.hk ↗hyang.org ↗Teaching assistant