MOONWALK: The AI That Stops Animation Feedback From Getting Lost

When an animation or visual-effects studio makes a movie, the early stage is a long chain of notes. A director says 'make it feel warmer,' a supervisor passes it to a junior artist, and three days later someone realizes nobody agreed on what 'warmer' meant. The paper calls this 'criteria drift' — the goal quietly changes each time the work is passed along, and the reason behind a single request usually doesn't survive the handoff. Everyone re-explains, everyone redoes work, and studios burn time and money.

MOONWALK is a software system designed to stop that. It keeps one shared record of what the team actually wants, ties every judgment to a concrete piece of evidence (a drawing, a reference clip, a written spec), and converts approved decisions into clear revision tasks. The key detail: the AI does not create anything. It acts like a very organized production assistant — flagging when a note is missing context and tidying up scattered comments — while artists keep full creative direction. That line matters, because the study found people refused to hand over taste.

The researchers tested MOONWALK with professional animators and supervisors using real production materials, comparing it against a plain chatbot interface. The structured tool won on three counts: people stayed truer to the original intent, they could trace why a decision was made, and the resulting to-do list was actually executable. The chatbot, by contrast, was less reliable at holding onto context.

The catch is honest and built into the design: aesthetic authority and final prioritization still rest with the humans. AI can organize your notes, but it can't tell you which shot looks better. Anyone who manages teams — designers, editors, marketers, even software groups — will recognize the underlying problem, and the fix may not require a chatbot at all.