A signing table at a convention. An artist draws your portrait, hands it over, and the thing you take home is partly the drawing and partly the fact that you watched it being made by a particular person at a particular time.
Present Drawing takes that apart to see which bits are load-bearing.
The paper was posted to arXiv on 23 October 2026 (arXiv:2610.32166) by Kojiro Tanaka, Takuya Iwamoto and Yuki Okafuji. The Comments field reads “Accepted (ACM UIST 2026)” — and UIST 2026 runs 2–5 November 2026 in Detroit, so this is accepted work whose venue is a week away. There is no DOI yet; the arXiv record is what to cite.
The arrangement
The proof-of-concept is a nice piece of stagecraft:
- An observer wears a head-mounted display and sees an avatar.
- The avatar’s motion is synchronised with a customised pen plotter.
- The plotter’s mechanical parts are hidden from the observer.
So you watch a figure draw, and afterwards there is a drawing. The marks are real ink on real paper, made by a machine you were never shown.
The stated aim is to “reconfigure conventional drawing events by relaxing their physical constraints” — and the paper is explicit that this is an organizer-oriented design space, which is an unusual and useful framing. It is written for the person staging the event, not for the person having the experience.
The design space
Four elements, and they are worth walking through because each one names a constraint that a normal drawing event cannot escape:
Time. A conventional drawing event happens now. Decoupling the drawing from the watching means the performance can be recorded and replayed — and the paper’s evaluated configuration is exactly this.
Site. One artist, one table, one queue. Relax it and the same drawing event can occur in several places.
Control Source. Who or what drives the avatar: a live human, a recording, or a system.
Avatar Embodiment. What the observer sees doing the drawing — a person’s representation, a mascot, something else.
The five application examples map onto those: live, playback, multi-site, mascot, and system-controlled.
The mascot case is the one that reveals what the design is really about. A costumed character cannot draw in character — the gloves are wrong, the hands are wrong, and the illusion that the character itself made the mark is unavailable. Here it is available, because the hands you see and the hands that move the pen are different hands.
What was actually tested, and what wasn’t
This is where the paper is commendably narrow. The evaluation is described as:
an initial evaluation of one representative playback autograph configuration, showing the basic plausibility of the core Present Drawing experience as an avatar-mediated drawing event and highlighting factors associated with perceiving the avatar as the drawer
Read that carefully. One configuration of five. Playback, not live. “Basic plausibility”, not effectiveness. And the outcome of interest is “factors associated with perceiving the avatar as the drawer” — whether the attribution lands at all, rather than how much anyone enjoyed it.
No participant count, no effect sizes, and no comparison condition are given in the abstract. Four of the five configurations are described but unevaluated. This is a concept and a prototype with a first look at whether the central illusion holds, which is the appropriate amount of claim for a UIST paper of this kind.
Why installation artists should read it
The transferable idea is not the avatar. It is the deliberate concealment of the mechanism combined with a physical residue.
Most mixed-reality installation work produces an experience and nothing else: you take off the headset and the piece is over, existing only as a memory and a photograph someone else took. The pen plotter is doing something specific here — it means the encounter ends with an object that persists, was genuinely manufactured during the encounter, and cannot be regenerated.
Three practical notes if you want to build in this direction:
Hiding a plotter is harder than hiding a screen. Plotters are audible. A gantry moving a pen makes a distinctive stepper whine and a paper-contact tick, and an observer in an HMD has their eyes occupied but not their ears. Any serious version of this needs the sound designed — either masked, or deliberately matched to the avatar’s motion so the noise becomes evidence rather than a giveaway.
Synchronisation is the illusion. If the avatar’s hand and the pen diverge by enough to notice, the attribution the paper is measuring collapses. The hard engineering is not the rendering; it is keeping a mechanical system and an animation in agreement, including through plotter acceleration limits that a drawn hand does not have.
Playback is the configuration that scales, and the one with the ethics question. A recorded performance replayed to many observers in many places is operationally attractive and quietly changes what the observer is being told. “This was drawn for you” and “this was drawn once and is being drawn again for you” are different claims, and only one of them is true in the playback case. The paper does not raise this, and anyone staging it commercially will have to.
If you want to build the physical half first, a plotter is the cheapest way in — our pen plotter walkthrough covers the AxiDraw and vpype toolchain, and the hiding is a cabinetmaking problem after that.
Sources
- Present Drawing: Exploring Avatar-Mediated Drawing Events with Physical Traces — arXiv:2610.32166
- arXiv PDF
- ACM UIST 2026
- arXiv listing for cs.HC