Hack #6: Zed

I built a Pokemon-style startup sim where you choose a founder, recruit a company, survive pitch battles, and race to Demo Day. ElevenLabs powers the game’s entire personality layer: rival founders taunt you with custom voices, mentors brief you in-character, music shifts with each startup stage, and a voice-command mode lets you make key choices out loud. Built entirely in Zed without hand-editing code, it turns startup culture into a fast, funny, replayable RPG. I added some fun easter eggs (for example PG - paul graham, father of startups, is there, and so is Garry Tan as a cameo - unofficial).

It's 2 AM. Joe wakes to a scream from the apartment next door. He goes up to check on Mrs. Hollis. He won't be coming back. Five cassette tapes. Four locks. One key. One way out — and it's not really a way out at all. Apartment 4B is a 15–20 minute browser horror game built in 36 hours for ElevenHacks. The thesis is simple: audio IS the game. Every meaningful beat of fear, bonding, and betrayal lands through ElevenLabs-generated voice — the cold-open narration, Mrs. Hollis's six tape recordings, Joe's reactions, the screams, the heartbeat under the climax, even the music bed. Three custom voices carry the entire emotional arc: Mora as Mrs. Hollis, Zelda as Joe, Jerry B. as the narrator. Every audio file is pre-baked at build time and committed as MP3 — the deployed site makes zero runtime API calls. The visual side is intentionally bare-bones. Every mesh is a Three.js primitive (box, cylinder, plane). Every texture — wallpaper, polaroids, the wall calendar, the spice rack — is drawn programmatically into a CanvasTexture. No GLTF imports, no Sketchfab. The PS1-era lo-fi aesthetic is on-trend for indie horror and lets sound design carry the weight. A 10-minute lighter battery counts down whenever you're playing. When it runs out, the game ends. Solve five gated tape puzzles to find the front-door key, turn it, and trigger the time-loop ending — pixel-identical to the cold open. You wake on the couch at 2:47 AM. There is no escape. Built with: Vite · React · TypeScript · React Three Fiber · Three.js · Drei · Howler.js · Zustand · @react-three/postprocessing · ElevenLabs (TTS + Sound Generation + Music API) · Zed (with its AI agent for pair-programming).

🕵️‍♂️ Spioniro-Golubiro 🌍🇩🇪🇯🇵🇮🇹 Can you prove you’re not the spy? A voice-driven interrogation game where you must survive by sounding like a local. Each level, an officer speaks a phrase in their language — you repeat it. Your pronunciation is analyzed in real time using AI. If you’re convincing enough, you pass. If not, suspicion rises… and failure means capture. 🚨💀 🎮 Core 🎙️ Listen → Speak → Survive 🌍 Multiple countries & languages 📊 Accuracy-based progression (60% / 80% / 90%) 🚨 Suspicion system builds tension 🔊 Tech ⚡ Built with Zed + Zed Agent 🔊 Powered by ElevenLabs for voices & audio 🤖 Speech similarity model for real-time feedback

Meet the monkey who can't stop dancing. He's been grooving since before you got here, arms up, maracas shaking, locked to a rhythm he'll never break. Here's the catch: you can move him. You can't control what he does next. Rhythm Kingdom is a "split-control rhythm arcade game". You control where the monkey goes — but not what he does. Actions are locked to a beat sequencer: eight slots that repeat. You fill them with actions in advance. When the beat hits, they execute. No take-backs. Jump. Roll. Throw coconuts at enemies. Miss the timing and you die. All audio — from chill to intense beats, animal sounds, voices, and tribal chants — is generated using the ElevenLabs Music and Sound Effects API, with ElevenLabs TTS voicing the game, layered to build the atmosphere. The screen shakes to the beat. Lights pulse. Every action triggers its own audio stem. All locked to tempo — everything perfectly synced, nothing out of time, in a single flow. When everything lines up, it feels like a live performance. You lose the rhythm. You die. You try again. Built on the Phaser 3 game engine — runs entirely in the browser, no installs, no backend. Built with Zed. The dev flow stayed fluid and clutter-free. I used Zed's Parallel Agents feature to work across different parts of the game simultaneously. With MCPs like Context7 keeping everything in sync, nothing drifted. Zed let me move fast without losing control. "Rhythm Kingdom : Place the runes. Hit the beat. Don't fall behind."

Growing up, I loved memory games, flip two cards, find the pair. Simple but addictive. So I rebuilt it with a twist. Not pictures. Sound. EchoJong is a Mahjong-inspired memory game where every move is guided by what you hear. Flip a tile, maybe it is a church bell, maybe an ice cream truck. Listen closely, hold onto it, and find its match. Every sound you hear is generated using the ElevenLabs SFX API and Music API, every audio clip in the game is AI generated. Built entirely using Zed, leveraging its AI editing, fast navigation, and integrated workflow to design and develop seamlessly in one place. This is not just a game, it is memory reimagined through sound. Can you remember what you hear?

An incredible game where you carry the last flame through a dying world. It features immersive audio, atmospheric world design, and a character that feels truly present. ElevenLabs for high quality sound and insanely emotional music that pulls you into the world. Developed using zed which is insanely fast to work in.

If you've had an old dog, you know the math... You've already done it. You did it last night. But we don't lay down and we don't give up. We're too strong to die and too stupid to quit. This is a folk punk beat-em-up about defiantly swinging a baseball bat at the inevitable. Three nights. Five hit points. The dawn is the win condition. One more day. The dog is asleep on the porch the whole time, the way old dogs sleep. Like a log. You stand watch in the swamp watching wisps and shades creep toward your cabin while the music gets louder. Shadows come for him. You decide they don't get him... Not tonight. Then you do it again the next night. And the night after. Until you can't. The vet said it's just a matter of time. The vet is right. But the night isn't tonight, and tonight you have a bat. **For Thelma, who made it to sixteen. And for Bea, who's turning ten soon.** ____________________________________________________________________________________________________________ Built end-to-end in Zed with the Anthropic Claude and ElevenLabs integrations, for ElevenHacks. All three combat tracks (drums, bass, guitar stems per wave) generated with the ElevenLabs Music API, then conducted at runtime in Tone.js so the mix layers up with your combo and slumps to a record-player crawl when you die. Title music and outro music generated the same way (but without stems), same emotional family, different tempo, so the cold open and the gut-punch close mirror each other. All combat SFX (bat swing, hit, death stinger, the dog whine on defeat) generated with the ElevenLabs Sound Effects API and pooled through HTMLAudioElement for sample-accurate retrigger. Asset generation scripts in /scripts are reproducible, the full pipeline committed. Because, why not? Pixi.js renders at native 480x270 pixel-art canvas, integer 4x scale.

Horror is hard to design. Anyone can make a player jump. Few can make them dread. Most horror games chase scope — sprawling maps, hours of mechanics, scares on a timer. The dread fades. The player adapts. Tension dies in playtime. The Hollow House goes small on purpose. It is a fifteen-minute first-person horror experience. Three friends take a shortcut through the woods and stumble into a house that does not want them to leave. They solve riddles posed by the house itself, examine documents, listen to whispered hints from each other, and survive whatever is hunting them — before they become the next set of scratched-out faces on the family photographs. Atmosphere comes first. The house breathes through audio. Every character has a voice. Aarav, the nervous calm one. Priya, who hears it before anyone else. Sam, who is too quiet. The narrator is dry. The presence is none of those things. Lines are generated dynamically from a single dialogue script, then routed by character. The entity passes through a Presence bus — layered pitch-shifted playback through lowpass, distortion, and reverb — so it feels diegetically wrong, not just loud. Footsteps duck under speech. Heartbeat and whispers ramp as you descend. A CRT shader sits over every frame — scanlines, vignette, faint flicker — like found footage you should not be watching. The entity only spawns in the basement. That one rule turns every other room into anticipation. Under the hood: Godot 4 on the gl_compatibility renderer for portability. Every voice line is generated and cached via the ElevenLabs Text-to-Speech API across five distinct voices — including a custom-tuned narrator and a layered demonic mix for the presence. The closing credit is pre-generated in parallel with the ending so the audio lands instantly. Built entirely in Zed. Every GDScript, every scene tweak, every gameplay system was iterated through Zed's AI agent — spec, draft, refine, ship — at the pace of conversation. The editor was the workflow, not just the surface. The Hollow House does not chase scares. It earns them.

Signal Lost is a 3D interactive sci-fi experience where players restore a failing space station by activating terminals and rebuilding signal strength. It solves the problem of passive AI demos by turning audio into a core gameplay mechanic. The game uses ElevenLabs to deliver immersive, dynamic voice interactions and was built rapidly using Zed for an AI-native development workflow.

NOX is a brutalist horror maze game where the creature hunts entirely by sound. You have 4 seconds to memorise the maze before the lights go out. Then you navigate blind — collect fuses, find the exit, survive. The creature's voice lines are generated live using the ElevenLabs TTS API. "I'm here." when it spawns. "I hear you." when the hunt begins. "Got you." when it finds you. Powered by Dante — ElevenLabs' growly monster voice — and cached locally so every new player gets fresh AI-generated audio on their first run. Built and coded in Zed. Single HTML file. No frameworks. No music. No ambient noise. Sound becomes the only language — which makes the AI voices land harder.

Soundle is a mobile-first word-guessing game where every clue is audio — no images, no text hints, just sound. Players listen to 2–3 audio clues and build the hidden word by tapping letters from a scrambled tile bank. Clues range from raw sound effects ("a wall clock ticking") to AI-narrated descriptions ("a flat panel of buttons you press to type") — combined, they reveal the answer: CLOCK, KEYBOARD, RAINBOW, and 17 more across 4 themed packs. The problem it solves: Word games are visually dominated. Soundle makes audio the primary interface — training your ears instead of your eyes, and creating a genuinely novel puzzle format. How it uses ElevenLabs: Every audio clue is generated by ElevenLabs using two APIs: - Sound Generation (`/v1/sound-generation`) — produces natural sound effects and music background from text prompts like *"friendly dog barking twice, clean studio audio"*, with a custom prompt wrapper that explicitly strips out any accidental speech. - Text-to-Speech (`/v1/text-to-speech`) — narrates description and word-part clues using the `eleven_multilingual_v2`. How it uses Zed: The entire project was built in Zed. Zed's AI agent was used throughout — most critically to design and wire up both ElevenLabs integrations inside a single Next.js API route, generate the batch audio script, and iterate on the puzzle data schema. Zed's speed and inline AI made it possible to go from zero to a working audio-driven game in a single hackathon sprint.

SUSPECT is an AI-powered interrogation game where you sit across from Detective Harlow and answer his questions with your actual voice. There are three cases — a murder, a missing person, and a suspicious accident. Each one starts with a cinematic briefing that puts you in the situation. Then the interrogation begins. You speak your answers out loud. The game transcribes them in real time. The detective reads exactly what you said and responds — not from a script, but generated fresh each time based on your specific words. If you contradict yourself, he catches it. If you stay silent, he notes it. A 45-second timer ticks audibly throughout your answer window. As pressure escalates across 6 questions, the room visually closes in — the vignette darkens, the light turns red, the detective's voice gets colder. At pressure level 5 the atmosphere is genuinely suffocating. The voice cloning feature is the centerpiece: after you've spoken enough, the game silently clones your voice using ElevenLabs' API. At a critical moment in the interrogation, it plays your own words back at you — in your own voice — as evidence. The disconnect between what you said then and what you're saying now is what breaks most players. At the end: CONFESSION, RELEASED, or NO COMMENT. The AI judge reads your full transcript and decides. Built entirely in Zed — the AI-native code editor. Zed's built-in AI assistant was used throughout development to scaffold components, write the interrogation logic, and debug the real-time audio pipeline. The speed of Zed's AI panel made it possible to ship this in 48 hours. ElevenLabs APIs used: — Text-to-speech (Detective Harlow's voice, generated per question) — Voice cloning (suspect's voice captured and replicated mid-game) — Sound effects generation (ticking clock, ambient drone, slam SFX) — Speech-to-text (transcribing the player's spoken answers) The goal was to make a game that makes you feel something. Most players confess.

OUTRUN is a synthwave endless runner with one twist that changes everything: the AI radio DJ chants whatever you type into the menu. Type ADITYA, see those six letters spawn on the runway, hear them chanted back at you in the same voice that greets you by name and reads your final stats aloud. Three sectors, escalating speed, a cyber shop between levels, gold shield power-ups that wrap Rocky the pentapod in a shimmering aura, and an AI-narrated recap that reads your run back to you when it's over. Every voice line, every sound effect, every music track — generated by ElevenLabs. Built entirely in Zed.

Letra Letra is a 3D letter-learning adventure for pre-K kids ages 3–6. Players walk through a googly-eyed alphabet world, bumping into letter characters to hear them, spelling missing pets, and matching phonemes to letters. Three modes ship: Spell the Word, Find the Alphabet, and Match the Sound, plus a trophy shelf for repeat play. Before each game, the kid picks an avatar, a biome, and the letter case: Avatars: Kid, Car, Rocket Biomes: Park, Moon Letter case: UPPERCASE, lowercase, Mixed Plays in any browser, installs as a PWA, and works offline. Try it: https://www.playletra.com The Problem Most pre-K alphabet apps are flat 2D tapping games with stock TTS, and most ask non-readers to read on-screen instructions just to navigate. Letra is the opposite: a tactile 3D world where every prompt is voiced, so a child who cannot read yet can still drive every screen alone. Big buttons, generous proximity collection, voice-on-hover menus, and a hint timer that gently re-orients instead of scolding, all designed around one core assumption: the user cannot read yet. ElevenLabs ElevenLabs voices every line in the game: every letter name, every phoneme, every prompt, and every celebration, all in one warm, consistent custom narrator voice, Marissa. The full corpus of roughly 90 short MP3s is pre-baked at build time, so a child’s 40th replay costs zero tokens and starts instantly from cache. Token economics flip when the audience replays, and a 3-year-old will replay the same letter 40 times in a session. Without ElevenLabs, the game falls back to the browser’s Web Speech. It works, but it does not feel like a friend. Zed Zed’s agentic editor wrote the majority of the React shell and the three.js engine, including geometry, lighting, and the per-frame loop. It also handled the input layer, including keyboard, gamepad, and on-screen joystick support, plus the trophy state machine and the biome registry. I drove the design and kid-UX judgment calls, (with help from my four-year-old assistant): proximity collection radius, hint phrasing, joystick ergonomics, and voice-on-hover. ⸻ Source: https://github.com/billums123/letra

Earwitness is a detective game where the gameplay is purely auditory. There is no combat, no exploration, no inventory — just four AI-generated suspect voices and the suspicion that one of them is lying. The premise. Each case opens with a forensic case file: victim, location, time of death, and four suspects. Players listen to each suspect’s recorded testimony, watch the transcript light up word-by-word as they speak, and flag the lines that sound suspicious. Once all four are heard, the player accuses the killer and identifies which line gave them away. A correct accusation triggers a full ElevenLabs-generated confession; a wrong one ends with an innocent person executed. The core mechanic. Lying lines are generated with lower voice stability and higher style settings in ElevenLabs Multilingual v2 — producing a subtle vocal tell that real human listeners can detect. The player isn’t guessing; they’re catching micro-cadence shifts the same way a detective in an interrogation room would. The textual layer. Every claim-dense line carries an evidence tag — ALIBI, TIMESTAMP, WITNESS, MOTIVE, ADMISSION — color-coded inline in the transcript. When a player flags a line, the notebook surfaces any cross-suspect contradictions: another suspect’s testimony, displayed in italic serif, with a short forensic note and a one-click jump to that exact moment in the other recording. Detective work earns the cross-reference, not handouts. What’s there. Three full cases: a 1960s-coded estate murder (Vance), a 1.2M-follower influencer drowned at her own brand-launch party (Wipeout), and a Hinge first date ending in an elevator (Double Tap). 112 individually-generated MP3 clips, 16 distinct ElevenLabs voices, 8 cross-suspect contradictions, full per-line stability tuning on the killers’ lies. The build. Built in Zed in 24 hours. Static HTML + React UMD + Babel — no build pipeline. Deployed on Vercel as a single static site. Audio generation pipeline is a 90-line Node script that reads case data, hits the ElevenLabs API per line, and writes per-suspect manifests the audio engine consumes at load time. Try it: earwitness.vercel.app

Voiceprint is a psychological horror game where the monster hunts you using your own voice cloned in real-time by ElevenLabs. WHAT IT DOES You explore an abandoned radio station that mysteriously kept broadcasting after closing eleven years ago. Halfway through, the game asks you to speak into your microphone for fifteen seconds. ElevenLabs clones your voice instantly. Within moments, the monster starts whispering at you from the dark in your own voice. "I'm right behind you." "Don't turn around." Players hear themselves say things they never said. HOW WE USED ELEVENLABS (5 APIs) - Instant Voice Cloning clones the player's voice live, mid-game, then uses the cloned voice ID to generate monster whispers - Text-to-Speech (Cedric voice, Eleven v3) cinematic narrator throughout - Sound Effects API ambient drone, footsteps, monster sounds, distant whispers, transition impacts - Music API original 60-second horror score for the trailer - All routed through Vercel serverless functions to keep API keys server-side HOW WE USED ZED Built end-to-end in Zed using its AI agent panel for rapid iteration on the Three.js scene, voice-cloning audio routing, monster AI, and shader pipeline. Zed and ElevenLabs are both built into the world as in-game sponsor posters pristine on one wall, defaced with red spray paint on another. Worldbuilding, not branding. WHAT MAKES IT SPECIAL This game cannot exist without ElevenLabs. Voice cloning isn't a feature added on top it's the entire mechanic. Last year's Supernormal made you talk to ghosts. This year's Voiceprint makes the ghost talk back, in your voice. Built in 4 days. Runs in any browser. Microphone required. Live demo: voiceprint-alpha.vercel.app

Guitar Hero is gone and rhythm game fans on mobile are stuck with tap-to-the-beat clones that have no soul — no crowd energy, no feedback, nothing that makes you feel like a rockstar. FretHero is a work-in-progress mobile rhythm game built in Unity that uses ElevenLabs to generate dynamic audio reactions — crowd cheers, voice callouts, and sound effects that respond to your performance in real time — coded entirely in Zed, and designed to eventually ship on the App Store.

Martian is a 3D driving game set on the surface of Mars. You pilot a rover across red, rocky terrain — navigating craters, Martian rock formations, and glowing lava fields under a dusty, reddish sky.

I built AI Tom, an upgraded version of the classic Talking Tom experience, enhanced with real-time conversational AI. The project was developed during ElevenHacks with the goal of combining nostalgia with modern AI capabilities. Traditional Talking Tom applications were limited to voice mimicry and predefined animations, offering no real intelligence or meaningful interaction. AI Tom reimagines this concept as a fully interactive AI companion where users can engage in natural conversations and receive intelligent, real-time responses while also interacting with dynamic emotional reactions such as happiness and mood changes. The classic slap feature has been retained but enhanced with a behavior-based system in which repeated interactions activate a “Roast Mode,” allowing the AI to respond humorously and more aggressively. Additionally, the system can deliver personalized interactions, including light-hearted roasts based on a user’s social media presence, making the experience more engaging and relatable. AI Tom leverages ElevenLabs for high-quality, natural-sounding voice synthesis, enabling expressive and realistic interactions, while the ElevenLabs conversational agent powers real-time dialogue and intelligent responses. The application was developed using Zed IDE, which enabled rapid prototyping and efficient implementation of the system. Behavioral logic is used to manage different interaction states, including normal conversation, emotional responses, and roast mode, allowing the AI to dynamically adapt its tone and personality based on user actions and creating a more lifelike experience. This project addresses the limitation of many existing AI applications that are functional but lack engagement and personality. By combining conversational intelligence with interactive gameplay elements, AI Tom transforms AI into a more entertaining and socially engaging experience. It introduces a shareable and viral aspect through humor and dynamic responses while reviving a familiar concept with modern AI capabilities. The project stands out by blending nostalgia with advanced AI, introducing behavior-driven personality switching, and delivering a simple yet highly engaging and scalable concept. AI Tom is not just a chatbot or a game, but an interactive AI personality that responds intelligently, adapts dynamically, and creates a unique user experience.

Hushfall is a small, polished stealth game built for the Zed + ElevenLabs Hackathon. You play as a Finder moving between a frozen living world and a fractured broken world to recover trapped memories while avoiding Echoes that react to sound. We built the project in Zed and used ElevenLabs to shape the game’s entire sonic identity, not as a layer on top of the gameplay, but as the system that drives it. ElevenLabs was used in three core ways. First, we created the game’s musical atmosphere: separate tracks and cues support the emotional contrast between the stillness of the living world and the tension of the broken world, so crossing between realities feels like a tonal shift, not just a visual one. Second, we used ElevenLabs-generated sound effects to build the actual stealth language of the game. Different surfaces produce different footstep sounds and threat profiles, object collisions create distinct noise events, and moments like mirror crossing, memory return, Echo alerts, and release beats all have their own sonic signatures. Third, we used ElevenLabs voice generation for the released souls, so when a trapped memory is restored, the player hears a fragile, personal response instead of a generic reward cue. That matters because in Hushfall, sound is the foundation of play. Footsteps on wood, rug, stone, grass, or water do not just add atmosphere, they communicate risk. Bumping furniture or breaking an object does not just make the scene feel alive, it can expose the player’s position and trigger pursuit. The emotional payoff also depends on audio: the voices of freed souls turn progression into something intimate and human. ElevenLabs let us make sound carry mechanics, mood, and narrative at the same time. Zed helped us move quickly and stay focused on a quality-over-scope prototype where every audio decision mattered.

ElevenSlice is a reimagined take on the classic Fruit Ninja experience. This hand-gesture fruit-slicing arcade game lets players slash flying fruits using only their webcam, no controller required. It combines real-time hand tracking via MediaPipe, fast-paced gameplay, and immersive audio powered by ElevenLabs’ SFX and Music API. Developed with Zed and vibe-coded for rapid iteration, it delivers a fun, touchless experience that blends motion control and arcade nostalgia.

Atlas Cipher is a voice-first detective game where players receive missions from a handler, review case briefings, travel across a 3D globe, interrogate witnesses, collect clues, compare suspect dossiers, and ultimately decide who to arrest. Each case is structured as a focused but complete investigation loop: players manage limited time and travel resources, revisit their notebook, track evidence across cities, and piece together route clues, suspect clues, and corroborating details before issuing a final warrant. We built Atlas Cipher in Zed and used ElevenLabs as a core part of the experience, not just as an audio layer. ElevenLabs powers the live conversational interactions with the handler and witnesses, character voice design, narration, ambient sound, mission music, and supporting sound effects that shape the emotional pacing of the investigation. Our goal was to create a smaller, polished game experience where audio drives immersion, tension, and personality, making every conversation, city visit, and final accusation feel more immediate and memorable.

After Midnight is a late-night radio game built around a simple idea: every song is generated live from a real conversation. It recreates the feeling of old 3 AM radio — when you call in because you can’t sleep and someone actually listens — but turns it into an interactive experience. There are two game modes: In Caller Mode, you dial into a fictional station (WMID 88.7). An AI host (Kai, Luna, or Hank) picks up, talks with you, and after a short conversation writes and performs a 60-second original song using your words. The lyrics come directly from what you shared, so each song is completely unique and never repeated. In DJ Mode, the roles are reversed. AI callers (like a truck driver or a kid who just lost his goldfish) call into your show. You talk to them, understand their situation, and choose a track from three suggested vibes. Your voice is transformed in real time so you sound like a real radio host on air. Every generated song is saved into a shared library, organized by mood (e.g. Dusty Country, Synthwave Heartbreak, Cozy Lo-Fi). The game library of songs grows with every interaction, creating a persistent, communal radio experience. The core idea is that the conversation itself becomes the gameplay. Each interaction is a unique run, where players shape the outcome through what they say and how they respond. Two players can ask for the same kind of song, but end up with completely different results because the system adapts to the emotional signals and choices within each playthrough. We built this using ElevenLabs across the entire loop: * Conversational AI powers both the hosts and the callers in real time * The Music API generates full songs with vocals mid-conversation * Voice Changer lets the player become the DJ * Text-to-Speech was used for narration The app is built with Next.js and Zed IDE for managing the two gameplay flows, with all songs persisted into a shared library.

Chronos Grid is a browser game that turns navigation into an exercise in senses and doubt. You move through a 2D maze in near darkness with only a small circle of light. Walls do not stay still: they shift on timers, so the map you are building in your head is always provisional. ElevenLabs is not polish on top of the game; it carries the narrative and the deception. Synthetic voices sketch roles like dispatch and guidance, while generated sound cues can confirm reality or mislead on purpose. The design treats audio as evidence you must weigh, not wallpaper. Built with Zed as the hacking context and ElevenLabs for TTS and sound generation (plus dev middleware for music and effects), it lines up directly with sponsor tech: expressive speech and sound design that change how the player interprets each moment. Cognitively, it pushes spatial working memory, time-based prediction, selective attention in darkness, and calibrated trust: you rehearse separating signal from noise, updating beliefs when the world reshapes itself, and holding two timelines in mind when major and minor shifts run on different clocks. It is perception and reasoning under uncertainty in game form.

I built a murder mystery game. It’s fully interactive and actually immersive. You interrogate, ask questions you like, the suspects respond with real voices. Your goal is simple: find the killer. I built everything in Zed, using its ai-assisted workflow to move fast. ElevenLabs brought it to life: • Text-to-Speech + Streaming for real-time dialogue • Speech-to-Text for player voice input • ElevenLabs Voices for characters and narration • ElevenLabs Music for in-game atmosphere.

Ghost in the Machine is a real-time multiplayer social deduction game for coders - Among Us inside a code editor. 2 - 4 players join a room, one is secretly the Ghost, and the rest are Engineers racing to fix a broken codebase before time runs out. The Ghost sabotages from the shadows using abilities like code injection, screen blackouts, and phantom cursors - all while whispering through a demon voice. ElevenLabs powers 7 distinct audio layers: Speech-to-Speech transforms the Ghost's real voice into a demon broadcast heard by all players, TTS generates dynamic narrator lines for key game events, whisper-mode TTS creates eerie paranoia whispers that intensify as tension rises, and Sound Generation produces all SFX and 3-layer adaptive ambient music that crossfades based on the paranoia meter. The Ghost can also send custom demon taunts via TTS. Built with Zed — using the Agent Panel for rapid prototyping, Inline Assistant for code transformations, .rules file for project-wide AI conventions, and MCP servers for extended tooling.

Zombow — Survive the Horde Zombow is a fast-paced, 3rd-person survival archery game built entirely in the browser using raw JavaScript, Three.js, and WebGPU. Set in a lush, procedurally generated tropical environment, players must survive an endless, dynamically scaling horde of zombies. The game features high-performance 3D spatial audio, realistic projectile physics with continuous collision detection, and over-the-shoulder camera mechanics for precision aiming. Zombow pushes the boundaries of what is possible for immersive, lag-free 3D gaming directly in a web browser.

Sirens of the Shore is classic "tower defense" type game with some interesting twists found only on the high seas! The ancient Sirens protect their homeland from being plundered by the greedy pirates, and they use their enchanting voices (via ElevenLabs AI) to change the physics of their world, and drive the pirates ships into the dangerous rocks! I thought it would be interesting to use the ElevenLabs api to inflluence the physics of the game and wasn't sure it was possible, but it came out quite nice. I used the power of ZED ai development to create a unique weapon system, gameplay, and make it fast and efficient for every computer as a browser-based game. What should have taken me a few months I was able to complete in days. I have never made any serious game before, certainly not to this caliber. The catchy siren music and pirate music is ElevenLabs music. The Pirates try to find their way ashore, to plunder the land and take as they wish. Who will win in this classic battle? The sounds of the Sirens or the brute force of the Pirates?! It's up to you to fight as the Sirens and use your voices to win the battle, and save your land! Video: https://www.youtube.com/watch?v=BMzXTuc-EeY

PaperCraft Theatre 2.5D paper theater choose your own adventure game. A single shadowy puppeteer voices every character by lifting masks. As the play unfolds, the masks slip and their own voices leaks through.

CODE RUNNER is a cyberpunk endless-runner game where you play as NEXUS, a rogue AI escaping through an infinite code editor while a hostile compiler hunts you down. I built it to solve a common issue in browser games: AI audio is often treated like a gimmick and quickly becomes repetitive. In CODE RUNNER, audio is part of the gameplay loop, so each run feels more alive and reactive. The game is built with a single-file HTML5 canvas frontend and a lightweight Node/Vercel backend for secure API proxying. ElevenLabs is central to the experience: real-time text-to-speech narration reacts to gameplay events, procedural sound effects are generated for actions like jumps and collisions, and adaptive music shifts with game intensity to create momentum and tension as difficulty increases. This week’s sponsor, Zed, shaped the project’s identity and presentation. The entire game world is styled like a living terminal with a code-native interface, scanline visuals, and a compiler-chase narrative inspired by editor workflows. The result is a fast, stylish web game where AI voice, generated sound, and dynamic music combine into a cohesive playable story.

Rhythm Grid — Transforming the minimalist black-and-white aesthetic of the 2010s into a portal leading to infinite soundscapes. Do you remember that sense of pure tension? The screen scrolling endlessly downward, your fingertips hovering between black and white—where a single touch could be a matter of life and death. We have preserved the soul of that minimalism: no flashy lanes, just the binary aesthetic of right and wrong; a tempo that accelerates with your heartbeat; and the unforgiving arcade brutality where a single mistake spells the end. Yet, we have transformed every successful tap into a creative brushstroke. When the final black tile vanishes beneath your fingertips, what plays is no longer a pre-programmed MIDI loop, but an original AI composition generated in real-time via the ElevenLabs Music API (music_v1). Your combo count, your speed tier, and your chosen genre tags are assembled in real-time into a prompt injected directly into a neural network. "Easy" mode might conjure flowing Lo-fi piano melodies, while the dizzying descent speed of "Expert" mode unlocks a completely reconstructed Drum & Bass track or a symphonic variation. The same level, three distinct universes—and absolutely zero canned music. Built on a tech stack of React, TypeScript, and Vite (developed using Zed), the game features three built-in starter tracks. No API key is required—allowing you to dive back into that "purely for the high score" golden era in under 10 seconds. But this time, every S-rank playthrough expands a personal music library that never repeats itself. The black-and-white aesthetic serves as the nostalgic shell, while AI generation provides the futuristic core.

Earshot is a horror game where your microphone is the controller. You wake up trapped in an abandoned office building. The lights are dying, the doors are locked, and something in the dark is listening. To get out, you need a keycard, the power back on, and a way through the front door before the building takes you. Every sound you make matters. Whisper and your light lasts longer. Speak and something turns its head. Run and you wake the whole place. The monsters don't care if they can see you. They care if they can hear you. Three of them share the building. The Listener hunts by sound. The Whisperer won't let you past the trapdoor until you whisper a phrase she gives you, into your actual microphone. The Jumpers wait in the vents. You have a few things going for you. A breaker puzzle solved by ear. Broken tapes you reconstruct on a workbench. Lockers to hide in. A charm that lures monsters away once. Crouch to move silent and slow. Run to move fast and loud. Speed or sound, you don't get both. You make it to the door. You step onto the porch. You turn around. They're already coming.

Ori Says is a polished 3D multiplayer party game inspired by Simon Says, where players must obey only Ori’s voice while fake AI commands try to trick them into moving at the wrong time. Players jump, freeze, race to colored zones, and survive quick reaction rounds in online rooms or offline bot matches. I built it in Zed with real-time multiplayer, a ready-up lobby, offline play, and a neon arena designed for short, fun gameplay sessions. ElevenLabs powers the core experience with generated voices, fake command lines, sound effects, music, and AI-created video/audio assets for the demo.

Puaro - psychological noir detective game where the player acts as an invisible interrogator, using a radio link and biofeedback to break suspects through dialogue and physical evidence. Here you can feel like a real detective, communicating directly with suspects, listening their voice and examining their personal belongings. The core gameplay is powered by ElevenLabs Text-to-Speech, using high-stability professional voices processed through bitcrush and radio-static filters to create a haunting, lo-fi atmosphere. The whole complex system was build on Zed, which found excellent solutions for any process in Puaro.

"Deadly Sentencing" is a voice-interactive, theatrical courtroom simulator inspired by the Domain Expansion from Jujutsu Kaisen anime. Players step into the shoes of a defendant trapped in a digital trial facing two AI entities: an impartial Judge and a ruthless Prosecutor. Over three intense rounds, players use their actual microphone to verbally defend themselves against absurd charges, while the Prosecutor dynamically generates hidden evidence to catch them in a lie. Use your wit, defend yourself and win. What problem does it solve? Most LLM-based games rely on text inputs, which breaks immersion and slows down pacing. "Deadly Sentencing" solves this by turning "prompt engineering" into a high-stakes, real-time verbal sparring match. It gamifies AI red-teaming and negotiation, proving that voice-first AI interactions can create genuinely stressful, immersive, and entertaining gameplay loops using minimal visual UI. How does it use ElevenLabs and this week's sponsor? The entire atmosphere of the game relies on audio. We used the ElevenLabs Text-to-Speech API to bring our AI antagonists to life with theatrical presence. The Judge utilizes a booming, authoritative voice to pass verdicts, while the Prosecutor uses a sharp, cunning voice to dynamically interrogate the player. The ultra-low latency and emotional delivery of ElevenLabs are what make the "Domain Expansion" feel truly deadly. The game was built with agentic coding powered by Zed.

Seven Minutes is an emotional choice-driven narrative game about grief, memory, and the impossible things we wish we could say. You play as Kira, who has spent six years saving for seven minutes with the father she lost. He does not know who she is. She cannot warn him. She cannot change what happens next. All she can do is sit across from him in a diner and choose what to say before time runs out. We used ElevenLabs to create the voice performances, music, ambience, and sound effects that carry the emotional weight of the story. We built the game in Zed, using it to rapidly write, debug, and shape the interactive narrative experience.

Built an Interactive horror game set in eerie backrooms where each level increases the number of ghost entities you must defeat. Move using your phone as a controller and fire by shaking it - powered by iphone motion sensors like accelerometer and gyroscope for immersive gameplay. Features SFX from @elevenlabsio and built using @zeddotdev Agent. #ElevenHacks

I built a space survival game where AI is not a gimmick — It's the reason you make it home. Last Signal: You're stranded on station Aurora-7, day 847. Your only way out is ARIA, an AI assistant who guides you through an emergency repair using her robotic arm. You talk to her. She listens. She responds. And somewhere along the way, she starts to care. Every voice is generated in real time with ElevenLabs.

FREQUENCY is a real-time hostage negotiation game where the bomber on the other end of the phone is a live AI. What we built: A browser game where every playthrough is unique. Claude API generates the bomber's responses in real time he has memory, emotional state, and will reach a natural ending based on how the player handles him. ElevenLabs Turbo v2 voices every single response live with dynamic voice parameters that shift based on his emotional state. Players can respond by voice using ElevenLabs Speech to Text, making it a true two-way voice conversation with an AI character. How ElevenLabs is core to the experience: ElevenLabs isn't background music it's the game mechanic. The bomber's voice changes with his emotional state. When he's calm, stability is 0.85. When he's desperate, it drops to 0.30. The trap state "resolved" outputs stability 0.90 but style 0.05, creating flat affect that sounds like calm but signals he's made his final decision. Players who only read the transcript fail. Players who actually listen to the voice win. How Zed is core to the build: The entire game was designed, architected, and built using Zed's AI features. From the system design documents loaded into context, to the milestone-by-milestone build driven by Zed AI agent, to the real-time iteration on every component. Zed wasn't the editor it was the co-developer. Stack: Bun + Hono backend, vanilla JS frontend, Claude Sonnet for the game master, ElevenLabs Turbo v2 TTS + STT for full voice loop, isometric pixel art scenes generated with GPT Image, animated with Seedance 2.0. The ending where he goes silent for 4 seconds before you find out if you saved him that silence is the whole game.

"Echoes of the Maze" is a 2D Maze survival game where the player is in a dark maze where he can only see so far. He has to reach the exit and proceed to the next level. However, there are enemies of various kinds, who would like to stop him from doing that. He has to trust a guiding voice as it tries to help him navigate the maze safely while the enemy taunts and confuses him in his guttural voice. The sound effects generated by elevenlabs, add to the feeling of dread and tension. This project has been built in Zed's AI Developement environment.

Voice Shooter is a cyberpunk action shooter that redefines player interaction through real-time voice control. Built entirely using Zed as an AI-powered development environment, and enhanced by ElevenLabs APIs, the game delivers a fully immersive, voice-driven combat experience.

Suvo's Pawnshop is a 3D pawn shop simulator game where you haggle with AI-powered NPCs, authenticate rare historical artifacts, and try not to get arrested. Sellers walk through the door with items some authentic, some fake, some stolen. You negotiate prices using text or your real voice, call experts on your in-game iPhone for authentication advice, and build your inventory across 5 levels. The game features NPCs with randomized appearances and unique personalities, a full economy system with rent, ads, and expert fees, and an authenticity mechanic where one bad deal lands you in jail. ElevenLabs powered Text-to-speech for NPC responses and speech-to-text for player voice input. The shop radio is powered by Eleven labs music. Sound effects were also generated using ElevenLabs. The entire project was built on Zed and Zed Agent.

I built a custom quest generator game that allows users to create fully interactive adventures tailored to their imagination. From fantasy worlds to entirely original storylines, users can design unique characters, branching narratives, and multiple endings. They can enrich their games with custom sound effects, character voices, and immersive elements, then publish them online or share directly with friends. The platform also enables users to explore and play quests created by others, creating a collaborative storytelling ecosystem—similar to a digital, shared experience inspired by games like Dungeons & Dragons. At the core of this project is ElevenLabs, which plays a central role in delivering a deeply immersive experience. Through its APIs, the game dynamically generates high-quality voiceovers for characters and custom sound effects that respond to the narrative. Beyond that, ElevenLabs Creative Studio was used extensively to generate all visual and multimedia assets—including images, videos, and animations—making the entire creative pipeline AI-driven and cohesive. This heavy integration of ElevenLabs transforms the game from a simple storytelling tool into a rich, multi-sensory interactive experience. The prototype was built using Zed, which proved invaluable throughout development. Its ability to handle parallel, independent tasks significantly sped up prototyping, allowing multiple components of the system to be developed simultaneously. Combined with its exceptional speed and fluid file navigation, Zed enabled rapid iteration and experimentation, making it a powerful environment for building and refining the project efficiently.

Mad Lib Music turns pure chaos into fully produced songs. You answer a few ridiculous questions out loud — "dancing raccoon," "spaceship," "pickle" — and thirty seconds later you have a sea shanty about your roommate's ex, or a Broadway number about taxes, or a trap song about pickles. I built it during ElevenHacks to answer a simple question: what happens when you take the randomness of Mad Libs and force it through a state-of-the-art music generation pipeline? The result is a complete, genre-specific track generated end to end from whatever absurdity you fed it, with the classic Mad Libs mechanic intact except the output is a real song you actually want to send to the group chat. Three ElevenLabs features do the heavy lifting. A voice agent collects the inputs conversationally, so it feels like a game show (users can use a form if preferred). Those answers get reshaped into a structured lyrical prompt that emphasizes the spirit and mechanics of the Mad Libs game, then the music API generates the track and returns timestamped lyrics, which drive a lightweight web frontend for perfectly synced playback and shareable outputs. The whole pipeline — conversational input, structured prompt, genre-conditioned generation, synchronized output — was built in Zed. Most AI music tools start with intention. Mad Lib Music starts with chaos, and that constraint is exactly what makes every output surprising, personal, and coherent enough to be genuinely funny.

Voices in the Dark is a fantasy-inspired psychological decision game where players survive by interpreting mysterious AI-generated voices in real time. Inside a dark minimal world, players are guided by three entities: a calm voice that rewards trust, a panic voice that pressures urgency, and a dark voice that manipulates obedience. Each voice carries consequences. Correct decisions increase your score, while listening to deceptive commands can instantly end the run. The result is a tense emotional loop built around trust, fear, and reaction under pressure. The project was built in Python using Pygame inside Zed, with ElevenLabs APIs powering the dynamic voice generation system. Instead of using static prerecorded audio, the game integrates emotionally distinct AI voices to create a living atmosphere where every command feels personal and unpredictable. What makes the project unique is the emotional gameplay layer. The player is not fighting enemies or solving puzzles traditionally. They are navigating uncertainty itself. The experience feels less like a conventional game and more like entering a haunted conversation where every voice claims to know the way out.

Code Noir is a browser-based noir detective game where the crime scene is a developer's codebase. Players solve a murder by investigating real syntax-highlighted code files, running terminal commands, tracing blockchain transactions, and unlocking the victim's phone — every clue is a real security vulnerability developers face daily, from hardcoded secrets to malicious npm scripts. ElevenLabs voices every character — a cinematic narrator reacts to each file opened, suspects have distinct voices during interrogation, and an anonymous caller interrupts mid-investigation with a distorted warning. Built entirely in Zed — the game recreates its editor aesthetic as the full game world. File tree, syntax highlighting, working terminal. Players never leave the developer environment, which makes the moment they realize a murder happened inside their own tools genuinely unsettling.

Walking Simulator: An ASMR Walking Experience Walking Simulator is a calm browser experience built to encourage movement and short breaks. You guide a small character through a quiet landscape that slowly shifts from sunrise to night. Birds in the morning, soft ambient sounds in the evening, and gentle lo-fi music throughout. You simply hold forward and walk. At intervals, the experience pauses and invites you to stretch. The character stretches with you, turning it into a small, shared moment. All audio including music, footsteps, and ambient soundscapes was generated using ElevenLabs Music Compose API and Sound Effects API. Built with Zed Agent. From idea to deployment in one session. Pure HTML canvas with no runtime dependencies. A simple space to walk, listen, and reset.

MathQuick: audio-first adaptive mental math trainer most math drill apps are broken in the same way. fixed question banks, no memory, same difficulty for everyone. ace multiplication today, come back tomorrow, it starts from scratch. MathQuick is different. questions are read aloud via text-to-speech. you solve them in your head. no peeking at the problem, no reading, just pure mental computation under a countdown. what makes it technically interesting we combined two things that haven't been put together this way before: FSRS (free spaced repetition scheduler, the algorithm now powering Anki) and LLM-based question generation via Venice AI. FSRS decides which skill to drill and when, based on a memory model tracking difficulty, stability, and retrievability per skill. Venice AI writes the actual question, fresh every time, calibrated to the learner's current level. the learner model tracks 6 skill dimensions (addition, subtraction, multiplication, division, percentages, chained operations) across 5 difficulty levels each, unlocking progressively as the learner improves. question generation uses a hybrid approach: a pre-built static bank of ~900 questions ships with the app for zero-latency first sessions, with Venice replenishing buckets in the background. audio is cached in IndexedDB so every question is only synthesized once. the adaptive algorithm - FSRS schedules which skill is most overdue based on predicted retrievability - zone of proximal development targeting: below 60% accuracy drops a level, above 85% for 3 consecutive promotes - promotion thresholds scale with level: level 1 promotes on 3 correct in a row, level 3+ requires 4 of 5 - 20% of questions come from already-mastered skills to prevent fluency decay - full research documentation at mathquick.vercel.app/research stack React, Vite, ElevenLabs TTS, Venice AI (llama-3.3-70b), ts-fsrs, localStorage + IndexedDB for offline-capable persistence. no backend, no account required. the full learner model fits in under 2kb of localStorage. try it mathquick.vercel.app

A game where the final boss is just asking a local where the best street food is. Learn practical, conversational languages through interactive, real-life travel scenarios. We coded the entire game from the ground up using standard web tech (JavaScript & HTML Canvas) so it runs instantly in any browser without downloads. To keep our development lightning-fast, we built it entirely inside the Zed editor.

Padmavyuh: One Door is a voice-driven retelling of the Mahabharata, set in a parallel universe where Abhimanyu, the sixteen-year-old who in the original epic died trapped in the Padmavyuh formation, gets one chance to be saved. You play Arjuna. Three voices call from the maze at every gate. Only one is Krishna's; the others are deceivers wearing his cadence. The mechanic is audio: you walk Arjuna toward the diya whose voice you trust. Get it wrong twice, and the maze closes. Subhadra waits at the gate; she has already grieved this boy once, and cannot grieve him twice. The problem it explores: most "voice AI" demos use voice as a layer, which is a narration over a game that would still work without it. Padmavyuh asks what happens when audio is the gameplay. You can't see your son. You can only hear three voices, one of which is the divine guide and two of which are the maze pretending to be him. Listening becomes the act of the game. ElevenLabs use: I used four ElevenLabs APIs in concert. Voice Design generated six custom Indian-English personas like Krishna (steady, philosophical), Narrator (formal, grave, Mahabharata-scroll cadence), Abhimanyu (younger, wounded), Subhadra (warm, breaking), and two Deceivers (one urgent and panicked, one seductively reasonable). Text-to-Speech rendered every line with per-character voice_settings (stability, style, similarity_boost). Krishna's settings stay rock-stable across all three gates because he is divine and constant; the deceivers' settings destabilise progressively gate-by-gate, where gate 3 deceivers have the lowest stability and highest style; so the maze itself sonically unravels as Arjuna gets closer. Music API pre-generated three tension beds (Dawn - Indian classical raga; Battlefield - distant war drums; Chamber - sustained tanpura drone) that crossfade between scenes. Sound Effects API generated the conch-shell horn at gate openings, a single bowstring pluck on wrong choices, layered arrow ambience during voice playback, and Subhadra's exhale on the closing "Beta." Each diya plays with stereo panning via Web Audio's createStereoPanner, so direction matters as much as content. Zed use: every code change, such as game logic, audio integration, character animation, spawn-opposite logic, the Subhadra arrival timing fix, the lives system, and visual feedback effects, was prompted to Claude inside Zed's Agent panel.

FrolicZone A premium game hub with real-time multiplayer — built with Zed and ElevenLabs. Play solo. Race the clock. Compete with anyone, anywhere. Every action is heard, every win is celebrated — out loud. Overview FrolicZone is a multi-game puzzle platform playable on mobile and web at https://frolic-zone-game.vercel.app It includes three complete games — Word Rush, Sort Splash, and Emoji Combos — each featuring: Structured progression systems Time-based challenge modes Real-time multiplayer tournaments using shared puzzles The defining feature is its audio-first design. Sound is not an afterthought — it is a core gameplay layer. All audio — sound effects, music, and voice — is generated using ElevenLabs and integrated directly into gameplay feedback loops. Core Experience FrolicZone is built around three pillars: 1. Immediate Interaction Each game is simple to start and intuitive to play: Swipe to form words Tap to sort Decode emoji clues No tutorials required — interaction is learned through play. 2. Competitive Structure Every game supports: Classic Mode — structured progression Timer Mode — score-based time challenges Tournament Mode — real-time multiplayer with identical puzzles Tournament Mode uses Firebase to synchronize gameplay and update leaderboards live during matches. 3. Audio-Driven Feedback Every action produces a response: Correct actions trigger tonal rewards Power-ups trigger distinct audio signatures Completion triggers layered sound + voice feedback A set of 20 dynamically selected voice lines (generated via ElevenLabs TTS) plays on key successes, creating a reactive and expressive experience. Games ⚡ Word Rush Swipe across a circular letter wheel to form words that fill a solution grid. 10000+ puzzles across multiple difficulty levels Bonus word system for additional rewards Animated letter connections and particle transitions Modes: Classic progression Timer-based scoring challenges Real-time tournament play 🫧 Sort Splash Reorder colored balls across tubes until each tube contains a single color. Deterministic level generation with seeded randomness Increasing difficulty from simple to complex configurations Designed to be solvable at every level Modes: Infinite progression Timed efficiency challenges Shared-puzzle tournaments 🧩 Emoji Combos Interpret four emoji clues to identify a hidden word. ~1000 curated puzzles Multiple valid answers per puzzle Hint system with in-game currency Correct solutions trigger visual and audio feedback, reinforcing discovery moments. Audio System (ElevenLabs) Audio is central to FrolicZone’s design. Sound Effects 11 distinct effects generated via ElevenLabs Sound Generation, each designed with a specific gameplay purpose: reward signals interaction feedback error cues progression milestones Each sound is defined by emotional intent, not just function. Voice Feedback 20 encouragement phrases generated with ElevenLabs Text-to-Speech. Randomized playback prevents repetition Tuned for expressive variation Triggered on successful actions Voice lines act as real-time reinforcement, replacing traditional UI feedback with human-like responses. Background Music 5 looping ambient tracks generated via ElevenLabs: Calm Upbeat Mystical Lo-fi Epic Music is dynamically reduced in volume when voice lines play to maintain clarity. Audio Pipeline All assets are generated through a single script: npm run generate-audio-assets This pipeline: Generates SFX, music, and voice clips Tracks assets via manifest Ensures all required audio is present before runtime Multiplayer System Tournament Mode enables real-time competition: Players receive the same puzzle instance Progress is tracked and synced live Rankings update dynamically during gameplay

Prism lets you upload any EPUB and instantly turn it into a clean, distraction free reading experience in your browser. It preserves formatting, handles large books smoothly, and gives you full control over how you read font, layout, and navigation without needing any external apps or downloads. Beyond reading, Prism transforms books into natural sounding audiobooks using ElevenLabs, while an AI companion helps you understand what you read. You can ask questions, get summaries, and explore ideas in real time, making reading more interactive, accessible, and deeply engaging. Built with precision on Zed.

Tatakae Lodge is an anime-inspired strategic deck-building combat game where poker hands become cinematic battle techniques. Instead of simply playing cards, players build hands like pairs, flushes, full houses, and royal flushes to trigger attacks, combos, limit breaks, and ultimate arts. A weak hand might become a basic strike, while a powerful hand can become a dramatic finishing move like Water Slice, Hinokami Kagura, or Sun Halo. The game solves the problem of making card combat feel more emotional, readable, and exciting. Traditional deck-builders can feel abstract, so Tatakae Lodge turns every move into a visual and audio spectacle: the player understands the strategy of the cards, but feels the impact like an anime fight. I built the project using Zed as my fast AI-assisted development environment, helping me prototype UI, combat flow, and game systems quickly. ElevenLabs is used to bring the game to life with dynamic battle narration, character voice reactions, attack callouts, sound effects, and cinematic audio moments. Every major action can be voiced or enhanced with adaptive sound, making each combo feel like a true fighting-game moment. At its core, Tatakae Lodge asks: What if poker hands were not just scores but superpowers?

Shout Runner is a voice-controlled endless runner — yell "JUMP," "SLIDE," "FASTER" to play. ElevenLabs powers a live AI crowd that reacts to your run (cheering, gasping, falling silent on death) and generates every in-game SFX. Built so audio isn't decoration; it's the game.

SPELLSPEAK is a voice-driven magic battle game where your vocabulary is your weapon. There are no buttons and no predefined spell list — you describe a spell out loud, and the game generates everything in real time. It uses 7 ElevenLabs APIs deeply integrated across the entire battle loop: • Scribe v2 for speech-to-text transcription of player spells • SFX V2 for generating unique spell sound effects from text • Eleven v3 + Audio Tags for emotional enemy wizard voice responses • Voice Design for procedurally generating a unique villain voice each session • Eleven Music for archetype-matched battle soundtracks • Music Inpainting for injecting orchestral critical-hit stings into the live track The AI scores every spell on creativity, specificity, coherence, and drama — a boring "fireball" loses to "a tornado of frozen glass shards that screams at the frequency of breaking bones." The enemy wizard analyzes your spell, taunts you with emotional voice acting, and counters with its own generated magic. Built entirely in Zed with Gemini Flash for spell scoring and image generation. No two battles ever sound or look the same.

In 2005, military strategists classified memes as cognitive weapons. Today, we built the simulator. A 4X Roguelike of memetic warfare. You are the Chief Meme Commander (MCO), tasked with overthrowing the "Collaborative Corruption Government" by hacking the mindset of the population across a hexagonal grid using memes and memeology. At some point in the game during a random epoch, an event occurs and you must speak (elevenlabs) and what you say alters your influence on the population.

Duke — heads-up no-limit Hold'em against an AI you can actually talk to. A weathered Vegas pro plays real poker (deterministic strategy + opponent profiling) while a separate ElevenLabs Conversational AI agent voices him in character — free to bluff, lie, and trash-talk you in real time. Built with ElevenLabs Conversational AI and Sound Generation.

Echo Impostor is a procedural social stealth game where AI-generated voices are the core gameplay mechanic. Each run creates a new spaceship with crew suspects — one is an impostor. Players interrogate characters using ElevenLabs Text-to-Speech, listen for contradictions, and accuse the right suspect. Sound effects and music are also generated via ElevenLabs APIs, making every run sound unique. Built with Zed.

ECHOES is a browser-based audio memory game where you listen to a sentence and try to reconstruct it as it slowly fades away. Each sentence is spoken once using an AI-generated voice, and across four rounds, that voice progressively degrades through pitch shifts, distortion, noise, and filtering. What starts as a clear signal becomes increasingly fragmented until it’s almost unrecognizable. Your task is to recall and fill in the missing words before the memory disappears completely. The game focuses on a small, polished experience combining memory, attention, and audio perception. Using AI-powered voice and sound processing, ECHOES aims to create a sense of tension and emotion, making players feel the fragility of memory in real time. Built using Zed for rapid development and ElevenLabs APIs for dynamic voice generation and audio effects.

I built Whisperwoods, an ai game that listen to everything you type and talks back. You're lost in a dark forest. The game narrates what you see and hear using ElevenLabs TTS, with ambient sounds (branches snapping, wind, whispers) from the Sound Effects API. You type choices ("go left", "hide", "run") or choose what the ai suggests . Tension builds through audio alone. Short, atmospheric, and memorable.

The Chaos Kitchen is a web-based kitchen chaos management game where players act as Head Chef commanding a team of hilariously dysfunctional AI staff. The core problem it solves is showcasing how ElevenLabs' full API suite can make game characters feel genuinely alive — not through pre-recorded audio, but through fully dynamic, context-aware AI voices generated in real time. How it uses ElevenLabs: Every single sound in the game comes from ElevenLabs — there are zero pre-recorded audio files in the codebase. -Text-to-Speech: Three staff characters (Marco, Kevin, Isabelle) each have unique voices and 60+ contextual dialogue lines across 5 stress levels (Relaxed → Meltdown). As kitchen chaos rises, TTS parameters shift in real time — lower `stability`, higher `style`, faster `speed` — so voices literally sound more panicked and unhinged the worse things get. At 90%+ chaos, staff start calling each other by name and arguing. -Sound Effects Generation: All SFX (fire alarms, plate crashes, success bells, background music, victory/defeat jingles) are generated on demand via text prompts — no MP3 files, just prompts like *"Large ceramic plate smashing on a tiled floor"*. -Speech-to-Text (Scribe v1): Players can speak a custom dish name into their microphone — the game transcribes it via ElevenLabs STT and creates a live order on the board, fully playable on mobile browsers. -Instant Voice Cloning: Players can upload a short voice recording of a friend. ElevenLabs IVC clones it into a new staff member — so your friend's voice gets blamed for burning the food.

Fivetran is a real-time voice compliance monitoring application built for financial service companies using Elevenlabs Scribe V2 for speaker ID, violation detection, and sensitive data flagging

What we built** **Future Selves** is a voice-first interactive fiction game disguised as a daily ritual. Each day, the player gives the game one word about how they feel, receives a short “transmission” from a possible future version of themselves, and makes one small choice that nudges the story forward. Over time, new future voices unlock, and the cast expands from encouraging voices like **Future Self** and **Future Mentor** to more unsettling possibilities like **The Shadow** and other unchosen futures. **What problem it solves** Most self-improvement tools feel clinical, generic, or guilt-driven. They ask users to track, optimize, and correct themselves. **Future Selves** solves a different problem: it makes reflection emotionally engaging. Instead of asking, “What’s wrong with you?”, it asks, “What would the person you’re becoming want to tell you right now?” That makes self-reflection feel intimate, memorable, and motivating rather than dutiful. **The core experience** The hook is emotional contrast. Every future character has a distinct tone, perspective, and voice. **Future Self** sounds warm and grounded. **Future Partner** sounds intimate and vulnerable. **The Flatlined** feels muted and absent. **The Exhausted Winner** sounds polished but hollow. Players feel the difference before they even process the words, which makes the audio itself part of the gameplay. **How it uses ElevenLabs** ElevenLabs is central to the game, not just decorative. Every transmission is turned into character-specific speech using the ElevenLabs text-to-speech API. Each cast member maps to a different voice profile with tuned parameters like stability, similarity, and style, so the voices carry distinct emotional identities. The generated audio is then stored and played back in the app, turning written narrative into a performance the player can hear and feel. **How it uses Zed** We built the project in **Zed** and used it as our primary environment for rapid iteration across the Expo + Convex codebase. Zed’s speed and AI-assisted workflow helped us move quickly between narrative design, backend generation logic, and frontend polish. It was especially valuable for iterating on prompts, refactoring TypeScript across the monorepo, and tightening the game loop fast enough for hackathon pace.

I have built a geometric rhythm game which is enhanced by elevenlabs music generation. Defeat your geometric foes by placing attack, move, and shield elements in a rhythmic sequence to catchy music loops.

Voice Thief is a noir heist puzzle game where the player has no voice — but everyone else does. The core mechanic: record NPCs in the wild, steal their voices, and use them to manipulate voice-locked doors, intercoms, and phone calls to break into a bank vault before the clock runs out. It solves a creative design problem — how do you make voice AI the gameplay itself, not just a UI layer? Every locked door in the game is a voice puzzle. The vault accepts only the manager's calm voiceprint. The hallway intercom belongs to the secretary. Use a panicked recording and you trip the alarm. The player must social-engineer the right people into saying the right things, in the right emotional state, at the right time. ElevenLabs powers the entire voice layer: all 14 NPC dialogue clips are rendered with ElevenLabs TTS to give each character a distinct, human voice. The phone call system uses ElevenLabs to synthesize real-time impersonation audio when the player makes a call in a stolen voice. The architecture is IVC-ready for live Instant Voice Cloning flows on auth surfaces. Zed was used as the primary editor throughout the build.

Voice Thief is a noir heist puzzle game where the player has no voice, but everyone else does. The core mechanic: record NPCs in the wild, steal their voices, and use them to manipulate voice-locked doors, intercoms, and phone calls to break into a bank vault before the clock runs out. It solves a creative design problem — how do you make voice AI the gameplay itself, not just a UI layer? Every locked door in the game is a voice puzzle. The vault accepts only the manager's calm voiceprint. The hallway intercom belongs to the secretary. Use a panicked recording and you trip the alarm. The player must social-engineer the right people into saying the right things, in the right emotional state, at the right time. ElevenLabs powers the entire voice layer: all 14 NPC dialogue clips are rendered with ElevenLabs TTS to give each character a distinct, human voice. The phone call system uses ElevenLabs to synthesize real-time impersonation audio when the player makes a call in a stolen voice. The architecture is IVC-ready for live Instant Voice Cloning flows on auth surfaces. Zed was used as the primary editor throughout the build.

IPL Auction Game is a browser-based cricket auction simulator that lets you experience the thrill of the Indian Premier League auction firsthand. Featuring real 2024 IPL player data, you manage a team budget, bid against AI franchises, and strategically assemble your squad before the purse runs dry. Built for cricket fans and devs alike — fast, fun, and surprisingly addictive. Features: 🏏 Real 2024 IPL player roster with stats and base prices 💰 Team budget management across all franchises ⚡ Live bidding simulation with competitive AI 📊 Squad overview and remaining purse tracker 🎯 Strategic player targeting — all-rounders, pace attacks, finishers

DORA: The Compliance Roguelike is a browser-based card game about surviving an EU regulatory audit. You defend five compliance indicators by playing regulatory control cards while an ElevenLabs-voiced Regulator reads increasingly painful DORA findings. The game uses ElevenLabs-generated regulator dialogue, boss voices, music, and sound effects. Audio is preloaded before gameplay so every finding, card response, boss transition, and fine notice lands instantly after the mobile tap-to-start unlock. What makes it different: the horror is real. The card mechanics are built around DORA, NIS2, and GDPR controls, including a Dutch regulator mode, GDPR boss variant, hard mode for systemic institutions, and a Coffee Break card that does absolutely nothing except draw two cards. Demo video : https://youtu.be/iolfzKFnRJI

Say It is a pronunciation game that doesn't let you move forward until you actually say the word right. A word appears on screen — French, Japanese, Arabic, German, or brutally hard English — and the gate stays shut until your mouth gets it right. Stuck? Hit Hear It and ElevenLabs speaks it perfectly. You copy it. You pass. Next word flies in. No waiting, no skipping, just you vs your mouth vs every language on earth. Kids sound ridiculous. Adults sound worse. Everyone wants one more round.

Sound Garden is a musical painting game inspired by the Jennifer Lawrence synesthesia meme. what if you could literally hear colors and see sounds? Paint near glowing seeds to make them bloom into musical plants that create an evolving symphony based on where you paint (pitch), what color you choose (instrument), and how you move (rhythm). ElevenLabs powers the entire audio experience: AI-generated garden sound effects (blooms, wind, water), adaptive background music that grows richer as your garden flourishes, and a poetic AI narrator that responds to your progress. Built entirely in Zed editor (which is incredibly lightweight on Windows compared to vscode), it transforms simple painting into an emotionally resonant creative experience where anyone can compose beautiful music without musical training. TikTok is banned in India, but I used a VPN to create a new account and post the demo anyway - deserves at least 1 extra point! 😄

(Mood Wave) A Text-to-Voice-to-Music Blend AI-generated voice with music.It can turns text into a spoken voice line, then blends that voice with adaptive background music based on the text’s tone, emotion, and pacing.

The Snap & Speed Date is a fast-paced, insane dating game that recreates the speed dating process in the age of Artificial Intelligence. Instead of the tedious swipe-swipe-swipe, you need to take one quick selfie and enter into a 60-second date with an individual AI personality. Every character has its unique character – some might flirt with you, some might roast you, but some can leave you utterly astonished! The Snap & Speed Date game app is all about reacting quickly and playing along to keep up with the dynamic nature of the game. It’s a combination of a dating game and survival of the fittest, as the ticking timer increases the intensity of the game while each AI reacts differently!