sounding games

Hangos Játékok · six strands, six playable examples

A family of games played on buttons and answered entirely in sound. No screen, no text, no picture. Six faculties held deliberately apart — inference, articulation, timing, imitation, surprise, and the involuntary body — so that each can be played with, measured and composed for on its own terms. The suite began as a 414-word note in Hungarian written in October 2017; what follows is one worked example from each of its six strands, built to be played here in the browser.

The rule of the whole work is a single subtraction: the output is sound and only sound. Not sound-first, not an accessibility mode bolted onto a visual game — there is nothing to look at. A player with their eyes shut, and a player who has never seen anything, are playing the identical game. Every panel on these six pages will run with the monitor off. The numbers each game keeps are folded away under what the instrument recorded, and nothing in them is needed to play.

why the subtraction is productive

Reward has to be composed rather than decorated. On a screen, reward is trivially cheap — a number goes up, a bar fills, confetti. The sound is then free to be arbitrary because it is only confirming what the eye already knows. Remove the picture and reward must be made: audible as such, immediately, to someone who was told nothing. That is why success in these games is a change in timbre, clarity or room rather than a bleep. A lock does not congratulate you. It opens, and you hear that it is open.

Failure stops being punishment. If success is clarity, failure is simply less of it: the phrase comes back a little duller, the room a little more absorbent. Nothing is taken away and nothing scolds. That matters for the people this work is most useful to — children who have learned to fear being wrong, adults in cognitive decline, anyone for whom a red cross is a door closing.

The body comes back into the loop. Screens hold the eyes and, through them, the head, the neck, the posture. An audio game leaves the body free: it can be played walking, lying down, in the dark, by someone whose vision is exhausted, by someone ill in a room where a screen is one more demand.

And latency becomes honest. Vision forgives; hearing does not. Forty milliseconds between a button and a picture is invisible — between a button and a sound it is the difference between an instrument and a remote control. The target for the built work is under 10 ms from action to sound with under 1 ms of jitter, after Wessel and Wright (2002), and every hardware decision descends from that one line.

The span of the work A logarithmic time axis from one audio sample at 44.1 kHz (0.023 ms) to a heart-rate-variability window (30 s), with the six strands drawn as bars over the neighbourhoods they occupy. Articulation answers between 1 and 10 ms, rhythm between 10 and 120 ms, imitation between 0.4 and 4 s, surprise between 0.25 and 3 s, biofeedback between 1 and 60 s. Inference is unbounded. The 10 ms action-to-sound threshold is marked. 0.023 ms 1 sample 0.36 ms 1 block 1 ms jitter budget 10 ms fused 100 ms syllable 600 ms a pulse 3 s skin answers 30 s HRV window action → sound threshold I inference unbounded — wherever the player puts it II articulation button → sound III rhythm rhythmic judgement IV imitation turn-taking V surprise recognition VI biofeedback autonomic response
The span of the work. A logarithmic scale from one audio sample at 44.1 kHz to a heart-rate-variability window. Six strands occupy five different neighbourhoods of it, and that fact — not a theme, not a genre — is what actually separates them. A button and a sweat gland cannot be composed for at the same tempo.

the six games

Each page carries the strand’s paragraph from the 2017 source in Hungarian and in English, one game built and playable, and the other designs of that strand described. The suite catalogues nineteen games in all; six of them are built here, and the pages say which thirteen are not and why.

what a browser can and cannot do

A browser cannot honour this project’s timing budget. Web Audio scheduling plus pointer event delivery costs roughly 15–40 ms with several milliseconds of jitter — one to two orders of magnitude worse than the target. Every page reports what your machine admits to, above the pad. The rhythm game in particular measures you through that error, and says so rather than quietly folding it into the result. This is not a flaw in the demonstration; it is the demonstration, and it is exactly why the installed work runs on dedicated hardware — Bela for the research builds, a bare-metal Daisy for the hand-held object.

One consequence is worth stating plainly before you start: Bluetooth headphones break four of the six games. Standard pairing adds 150–250 ms; even a low-latency codec adds 40–60. Use wired headphones or speakers.

Every session runs off a seeded generator and the seed is in the address bar, so a session can be handed to somebody else exactly as it was played. Each game has a link button that copies it. Nothing is uploaded, nothing is stored, and no account exists to store it in; the one game that opens the microphone analyses it frame by frame in the page and keeps nothing.

the apparatus these stand in for

The nineteen games run on one object: a hand-held device roughly 180 × 90 × 30 mm, six mechanical buttons in a 2 × 3 grid with one key differentiated by profile, force-sensing under each, one continuous rotary with a centre detent. Wired audio out. One haptic actuator, used only where a strand needs a non-auditory reference. No lights except a dim power indicator, no screen at all, titles engraved in raised type and Braille. A single sensor port takes the biofeedback modules — skin conductance, respiration, pulse, EEG, gaze — as separate objects, so nobody carries electrodes they are not using. It runs standalone and is never required to be online to function.

A button is the smallest honest gesture available. It has no dynamics beyond force and no expression beyond timing, and it cannot be faked — which is exactly why it is the right controller for a project trying to measure something. On these pages the six keys become on-screen keys with keyboard equivalents; on a phone they are large enough to find without looking.

the 2017 source

The note that started this named the six strands and nothing else about them was written down for nine years. Its six paragraphs are reproduced, one per page, in the original Hungarian with a translation, so that anything built here can be checked against what was actually claimed. Its central methodological move is separation: six faculties held apart on purpose, in the source’s words, “so that we can examine and develop players’ skills in finer detail”. Ordinary games fuse them — a rhythm game is also a reaction game is also a pattern game is also a reward schedule — and the fusion is exactly what makes commercial games entertaining and useless as instruments. Isolating them costs entertainment and buys legibility.

where the work comes from

Rules that structure a game without prescribing its content: Xenakis’s Duel and Stratégie, Cardew’s Scratch Orchestra, Zorn’s game pieces. Play as pedagogy: Kodály, and Kokas Klára’s movement-and-listening work. Listening without an image: Schafer’s ear-cleaning exercises, Oliveros’s Deep Listening, radio drama’s whole craft. Games played with the ears: Baer and Morrison’s Simon (1978), and the small tradition of audio-only games that followed it. The instrument as toy: Toshio Iwai’s Electroplankton, the Skoog, Soundbeam. The body as the source: Lucier’s Music for Solo Performer (1965) and Kurenniemi’s DIMI-S (1971). And the access tradition this work belongs to whether or not it is invited: Oliveros’s AUMI, the EyeHarp, the OHMI Trust’s competition for one-handed instruments — where the shortage has never been of ideas but of instruments that are actually built and actually good.

start with I — the lock → ← installations