Your voice is the controller. Sound is light, words lift stone, and a shout drives back what hunts you.
A micro-horror puzzle game set inside an undiscovered pyramid that works as one vast acoustic machine. Your voice is the controller. This is a prototype, and it changes often.
Tuning. In the first chamber the tomb learns three notes from you: Muu (low), Aah (middle) and Kree (high). Any three notes you can sing comfortably will do. You can retune at any time from the pause menu.
Relics. The heart-jar answers to Muu, the ankh to Aah, the feather to Kree. Hold a relic's word and it locks to you. Carry it, then say the same word again to let it go.
The distant clunk. A relic set down on its own plinth settles quietly. Dropped anywhere else, it falls, and the sound carries.
Echoes. They cannot be seen unless your voice lights them. They wait behind columns, move only in the dark, and listen for a wavering voice and every clunk. A whisper and a heartbeat give them away.
The ward. A sharp shout throws them back.
The Tomb (Beginners). The Throat, The First Weighing, The Reach and The Hall of the Scale, then the Apex.
The Deep (Advanced). Opens once you have finished the Tomb. Three chambers, made new on every descent, drawn from The Forest, The Breathing Stones, The High Places, The Second Scale and The Hunter. It gets harder the deeper you go.
Listening. Headphones, speakers, or a keyboard voice if you cannot or would rather not sing.
Echo of the hall. Rich or Light. Light is easier on a phone.
Microphone. Use the default, or avoid a Bluetooth headset's microphone.
View. First person, or Spirit view: a fixed camera from above that never turns, playable with one thumb.
Comfort. Removes head bob, screen shake, grain and pulsing, and widens the view, for anyone prone to motion sickness.
Distant Clunk is one self-contained web page with no libraries, images, models or recordings. It was designed by Ollie Francis and written in conversation with Claude (Fable 5.2): each version was played on a Mac and a phone, and what felt wrong was described and reworked.
The picture. There is no 3D model of the tomb. One fragment shader raymarches a signed distance field, so every wall, plinth, relic and colossus is a formula. The shader's source is written by the game from a list of rooms, which is how The Deep can build a new tomb for each descent.
The stone. Giant blocks with hairline joints, weathering, grain, pitting, fine cracks and a worn path down each floor, all layered noise. Two inlaid gold lines ring with your voice.
Two cameras. First person and Spirit view share the same renderer. Spirit view cuts the roof away and draws you as a wisp of light.
Speed. The tomb is split into four segments and only those in view are worked out. Small objects sit inside bounding volumes, unlit stone skips its texture, and fine detail fades with distance. The picture's resolution settles to suit your device in the first few seconds.
Listening. The microphone is filtered to the range of a voice, then a pitch detector (the McLeod method) finds the note. A note must be clear and steady for a tenth of a second to count, so rumble, hiss and the game's own sound are ignored. The tomb quietly re-learns your low and high notes if they drift.
Sound. Everything you hear is synthesized as you play with the Web Audio API: filtered noise, a few oscillators and an echo. There are no audio files.
Echoes. Each one picks a column to hide behind, moves only while unlit, and weighs how loud and how unsteady your voice is, and how hard each relic lands.
Size. The source is minified into a single file. The budget is 100 KB.
Your voice never leaves your device.
0.23 These release notes.
0.22 Work on sound crackle on phones. New Echo of the hall setting.
0.21 The top bar fills as you hold a word. Sound and light reward a lock and a correct placement.
0.20 Faster engine, steadier picture, optional speed readout. The right thumb hurries the spirit in Spirit view.
0.19 More tolerant listening, a choice of microphone, and a message when the tomb cannot hear you.
0.18 Menus sized for phones, a shorter introduction, first fixes for phone sound.
0.17 Simpler title screen and a Settings panel.
0.16 Lock and release replaces holding a note to carry. Comfort mode and Spirit view. Echoes learn to hide.
0.10 to 0.15 The Deep. Giant stone blocks, great scale and the colossi in the Hall. The big coloured words. A more forgiving carry. Privacy and information.
0.5 to 0.9 Touch controls and fullscreen on phones. The pitch meter moves to the top. The game stops hearing itself. Version and file size on the title screen.
0.1 to 0.4 First prototype. The three-note tuning ceremony, the ward, recurring hints, the pause menu, lower and moodier sound.
The single-file build, the adaptive resolution and the approach to synthesized sound draw on Abyssal Garden by Maksymilian Dendura (MIT licence).
© 2026 Distant Clunk. All rights reserved.
Your voice never leaves your device. The game listens through your microphone to follow the pitch and loudness of your voice, and that is all. Nothing is recorded, stored or sent anywhere.
Your progress is saved in this browser only, so the game can remember that you have unlocked The Deep.
© 2026 Distant Clunk. All rights reserved.
The tomb waits.