Mumu Li n3xta
v2.0.0-20261017
← Writing

Movement Study

My central gesture is a pick–inspect–place loop. I repeat the same task so that changes in my pace, posture, attention, and fatigue become visible.

The Unreal robot will mirror my motion. If I show a reinforcement-learning dashboard, I will label it as a parallel training process, unless I build a direct connection between my captured motion and that system.

ID Movement What I want to show Variations Digital house
M01 Clock in: enter, take my position, lower my gaze. My body becoming available for work. Eager, hesitant, exhausted. The robot assumes a service pose.
M02 Pick: reach, grasp, and lift an object. Anticipation of weight and adjustment after contact. Near or far; either hand. The robot lifts a luxury object.
M03 Inspect: rotate the object and decide where it belongs. Judgment as labor, even when motion data barely records it. Quick glance or prolonged uncertainty. Inspection looks like admiration.
M04 Sort: place the object in one of two bins and return. Repetition, asymmetry, and practiced efficiency. Left or right; correct or mistaken. The robot arranges a display.
M05 Assemble: align and press two pieces together. Coordination and material resistance. Smooth fit or repeated attempts. The robot appears to work effortlessly.
M06 Reset: move the trays and clear my station. The hidden work that enables the next cycle. Routine, rushed, or paused. The room returns to a pristine state.
M07 Correct: retrieve an item and redo an action. Error and repair as part of production. Self-correction or external cue. I show the motion with and without the correction.
M08 Pause: freeze mid-reach, look toward a cue, resume. Who controls my pace, and the effort of holding still. Short or long hold; resume or restart. The robot freezes in an elegant pose.
M09 Accumulate: repeat M02–M06 for several minutes. How a real body changes through repetition. Comfortable, imposed fast, then slower. Early and late takes appear side by side.
M10 Refuse / exit: stop mid-cycle and leave. What happens when I stop supplying motion. Set the object down or walk away holding it. The robot stops, loops, or remains posed.

Capture units: I will record M01–M08 and M10 as short takes, starting and ending in a neutral stance. M09 will be one continuous take. I will log the props, bin positions, pace, repetitions, and prompts, and keep imperfect takes as part of the database.

Rehearsal and shot sequence

Sequence Physical view Digital view Purpose
1. Establish Wide shot of my body and station. Wide shot of the house. Establish the mismatch between spaces.
2. Learn the loop Side view of the full task. Robot performs the corresponding motion. Make the live connection clear.
3. Show what capture misses Close video of my fingers, grip, and corrections. Skeleton or robot beside it. Compare lived action with captured motion.
4. Repeat Fixed, uncut wide shot. Early and current takes, or an archive of loops. Let small bodily changes become visible.
5. Stop Empty station after I leave. Robot’s behavior after I stop. End on the question of whether the output continues without me.

I will shoot the close views on a regular video camera, separate from OptiTrack. They can preserve touch, facial expression, and hesitation that may never reach the avatar.

Capture and preproduction

I will draw a floor plan, mark my standing point and reach limits, and rehearse with large, lightweight, matte props. I need an open station that keeps my markers visible, especially when I cross my arms or lean over the bins. I will test my worker uniform over the marker suit and avoid reflective fabric or shoes. OptiTrack camera placement · Skeleton tracking

I will prioritize full-body actions: reaching, lifting, turning, transferring, resetting, holding, and walking away. If precise finger movement matters, I will need separate hand tracking; otherwise, I will show it in video or animate the robot’s grip deliberately. Glove setup

My test day will include calibration, a slow loop, a normal loop, a long take, and an occlusion test. I will ask someone to watch me and the avatar together and note where they diverge. I may also seek feedback from a movement practitioner and someone familiar with repetitive work.

Preservation and erasure

I can preserve timing, posture, repetition, and correction. The avatar may erase touch, resistance, uncertainty, and the conditions of work. I want that gap to remain visible: a difficult assembly on video might sit beside a robot that completes it smoothly.

I will keep the raw and failed takes alongside the cleaned animation, and label motion as live, recorded, edited, or looped. If I use Xiaomi’s public RL display, I will identify it as a separate process. The question I want to leave open is: what survives when my labor becomes reusable motion, and what disappears?