A real-world escape room with a living AI game-master — a voice-cloned persona who watches the players, guards the final secret, and delivers a cinematic, AI-generated reveal. The whole experience runs on a single offline computer in the room. No cloud. No internet. No per-token bills.
Try the finale Build one for your groupThis is the actual intro film for The Lost Fountain — the briefing every team of explorers watches before the clock starts. A talking-head British scholar, an animated legend of the Fountain of Youth, and a sixty-minute timer.
Every Escape Room AI gets its own character — a persona with a face, a voice, and a purpose. For The Lost Fountain, that character is The Keeper: a grave, theatrical British game-master who guides the explorers to the end of their quest.
The Keeper is an Elyan-class agent — a local language model wrapped in a written character, a cloned voice, and a set of tools he can actually use. He speaks in a voice cloned from the production's own narrator, so the AI on screen is continuous with the story the players have been living.
He stays in character under pressure, replays clues on command, and — at the right moment — breaks the final seal. He never reveals the ending early, and he never breaks the spell.
When the players bring the final code, The Keeper plays a film made for this room — AI-generated "rivers of living water," narrated in his own voice. This is the actual finale from The Lost Fountain.
This is exactly how the room ends. In The Lost Fountain, the last code the players recover is the phrase "It is finished." Type it below and break the seal — just as a player would.
Hint: it's the same phrase a real player would carry out of the codex lock.
Drag to rotate. Scroll to zoom. The relic at the heart of the quest, rendered live in your browser.
Live WebGL — running on your device, right now.
Four AI systems, composed into one in-real-life game — and all of it runs on hardware that fits in a backpack.
The persona speaks in a voice cloned from your production's narrator, so the AI is continuous with the story. Loud, clear, and in character.
A bespoke film for the finale — cinematic, themed to your story, narrated in the persona's voice. Made for your room, not stock footage.
A character-driven AI runs on a local machine — no cloud, no API bills, no data leaving the room. It can chat, give hints, and replay clues.
The critical reveal is on a deterministic gate, not the model's whim. The right answer fires the finale every time — in front of a room of kids, it cannot miss.
The entire experience runs on one computer in the room. Bring it anywhere — a church hall, a classroom, a fair booth — and it just works.
Any story, any character, any lesson. Mystery, history, scripture, science — the persona, puzzles, and finale are written to your theme.
The Keeper never breaks character because he is guarded by Harmonia — Elyan Labs' published, config-driven LLM guardrail framework. The same weighted-scoring engine that keeps a clinical AI on medical topics keeps the Keeper on his quest, turning away off-mission questions and jailbreak attempts in microseconds, before he ever speaks.
He is an Elyan-class agent: a persona with a written character, a cloned voice, and his own tools, running entirely on local hardware — no cloud, no per-token bills, your data never leaves the room.
Built for Joshua Migues — real estate professional and Sunday-school teacher — to run for his class. The players solve physical lockbox puzzles, recover a scroll, and crack a codex lock, then bring the final answer to The Keeper. He breaks the seal, plays the living-water film, and delivers the closing lesson in his own voice (John 4) — the Lost Fountain was never a place; it is the living water of Christ.
Church, classroom, youth group, team event, or attraction — we write the story, the persona, the puzzles, and the finale, and hand you a box that runs it. The first conversation is free.
Start the conversation scott@elyanlabs.ai