Why this project?
From thriller to conversational hypothesis.
Why extend a film through conversation — rather than an app, a game or a series?
01·The Project·Founding Prototypes
The concrete origins of GamiWays
GamiWays was not born from an abstract idea. It is the generalization of two functional prototypes, tested with real users, which each revealed the same fundamental challenges: latency, experience personalization, immersion and voice quality.
Storygami·Où est Ava ?
Edugami·Le Dilemme Plastique
00From Prototypes to Platform
Two experiences, one shared engine.
Où est Ava ? and Le Dilemme Plastique are two radically different experiences on the surface — one narrative and cinematic, the other pedagogical and scientific. Yet they share exactly the same technical challenges: STT→LLM→TTS pipeline, conversational agent orchestration, session memory management, perceived latency and French voice quality.
This convergence gave birth to GamiWays: a generic platform capable of supporting both the Storygami experience (interactive cinema) and the Edugami experience (guided learning), sharing the same orchestration engine.
Latency & fluidity
STT→LLM→TTS pipeline <2s
French voice quality
Naturalness, prosody, immersion
Session memory
Continuity without token explosion
Expressive avatar
Lip-sync + body language sync
01Où est Ava ? — Storygami
Storygami·Interactive cinema
Où est Ava ?
Prototype 1 — v0.26.0
The experience
Dive into a dystopian thriller with "Où est Ava ?" — an immersive conversational audiovisual experience accessible on desktop and mobile. You are catapulted into a mountain chalet where the family of Emma, Max, Où est Ava ? (9 years old) and Léo (15 years old) tries to survive a virus that transforms women into men (the "protogyny").
"Où est Ava ?" is the interactive extension of the film "Où est Ava ?", a dystopian thriller in production (public presentation planned for 2026). Explore the film's initial context, discover the family's stakes and secrets, and prepare to live a personalised conversational story, full of twists and surprises.
Through conversational AI and hyperrealistic synthetic videos, engage in unique dialogues with the characters of this world on the brink. Ask questions, express your hypotheses and explore complex themes in a world in peril. Each conversation reveals a fragment of a possible reality, with all its contradictions and complexity, to confront you with your own values.
The objective of this digital project: to make you think. Explore your reactions to the themes addressed and experiment with your relationship to your own certainties and moral principles through this original interactive experience.
Key metrics
v0.55.1
Current version
1 562
Commits (since Mar 2026)
5+
Active TTS providers
memoways/ava-proto1 · v0.55.1 · last commit 2026-08-07 · Streaming Avatar (HeyGen/Tavus) · Max→Emma handoff · RAG Voyage 4 · GM Orchestration · 1,562 commits
Explore Où est Ava ?
Hover a card to see the question that page explores.
Read the answer
The experience
Booth, audience journey, what the film leaves open.
What does someone actually experience in the booth — minute by minute, after the screening?
The prototype
Max and Emma, live metrics and safeguards before any wider release.
What already works — and what evidence is still needed before a release?
Back-Office
The narrative control room — without redeploying every tweak.
How do you edit voice, RAG and Game Master without touching code on every take?
Technology
17-node stack, sovereignty, real-time constraints.
Which building blocks sustain five to ten minutes of continuous dialogue?
Learnings
800 field commits — what we know, what we refuse to lock in.
What did we learn by building — and which hypotheses stay open?
Roadmap
Shipped, in progress, end-of-summer 2026 target.
What's stabilized — and what are we aiming for before the first festival showings?
02Le Dilemme Plastique — Edugami
Edugami·Voice-first education
Le Dilemme Plastique
Public teaser prototype — v2.8.0 (Light)
The project
Dilemme Plastique is an innovative project combining documentary expertise, artificial intelligence and active pedagogy to raise awareness of environmental issues related to plastic pollution. It is designed to evolve into the generalisable educational tool Edugami.
The project was born from the meeting between Peter Charaf, a documentary filmmaker who spent a decade (since 2015) travelling the world twice to document the plastic problem in partnership with the Race for Water Foundation, and the Memoways team. This collaboration resulted in a unique proprietary database: field-verified images, videos and testimonials covering environmental pollution, health impacts, recycling challenges and emerging solutions.
After a first deterministic prototype presented at the Semaine des Médias in French-speaking Switzerland in February 2024, the team made a major technological pivot: abandoning the linear no-code approach in favour of a system 100% powered by generative AI. The student can communicate with Peter Charaf's avatar via voice or text, in an environment designed to maximise pedagogical engagement.
Long-term vision: Edugami is an educational tool generalisable to other themes. "We replace plastic with something else, another theme, history, another social issue." — Peter Charaf
The Light teaser
Peter guides 12–18 year-old students through analysing an image of Geneva's Place des Nations to discover 6 hidden clues about plastic pollution. Challenge: find all six clues within 8 exchanges; discussion stays open until the 15th exchange. Peter receives a server-side tracking block each turn — he never re-asks for a clue already found.
Voice-first pipeline: Deepgram nova-2 live during recording (visual feedback), Whisper as ground truth at end → OpenAI Assistants API → per-sentence Gradium TTS (rolling early dispatch). Welcome message pre-generated per session with the student's first name; contextual resume on return to /tutorial via POST /api/sessions/:id/resume. Hosted on Coolify (proto-dilemme2.edugami.app).
Key metrics
v3.0.0
Current version
201
Sessions tested
6 / 8→15
Clues · exchanges
1 209
DB messages
Ulrfis/Dilemme-Plastiqueproto-Light · Deployed on proto-dilemme2.edugami.app (Coolify) · PostgreSQL 16 · 20 feedback responses
A fuller Flowise prototype (28 nodes, 20–30 min sessions) exists in internal R&D — diagrams below are kept as technical archive.
Tools used
Light / Tutorial Prototype
Replit Agent (initial vibe coding + hosting), Claude Code, OpenAI GPT-4o-mini Assistants API + Whisper, Gradium TTS. Hosted on Coolify (proto-dilemme2.edugami.app) since July 2026.
Flowise / Full prototype
Replit AgentVibe coding + hosting
FlowisePeter orchestration — 28 nodes (Memoways self-hosted)
OpenAIGPT-4o via Flowise — Peter conversations
GradiumPer-sentence TTS (replaces ElevenLabs since v3.0.0)
ElevenLabs ScribeSTT (WER 2.11% French) — Flowise archive
DeepgramNon-final live transcription (visual feedback)
PostHog + RectifyProduct analytics + session recording
Technical challenges → documented solutions
STT / Speech-to-Text
ElevenLabs Scribe vs Deepgram vs Whisper — WER, latency, French accuracy. Dilemme uses Scribe (WER 2.11%).
TTS & Voice Synthesis
ElevenLabs eleven_multilingual_v2 — 1507ms avg vs 2373ms OpenAI. Cache LRU strategy for classroom use.
Voice-to-Voice Pipeline
Interactive V2V diagram — visualize the full pipeline latency budget as used in both Dilemme prototypes.
Opportunities & Gaps
Flowise orchestration limits, LLM latency, STT accuracy on children's voices — gaps that GamiWays addresses.
03Convergence → GamiWays
The same challenges, one answer.
Two very different projects — an interactive film and a science class — ran into the exact same walls. Those walls became GamiWays' research agenda. Click each challenge to understand what's blocking and where we stand.
01
The slowness problem
When the avatar takes 8 seconds to respond, the conversation dies. The user disengages, the illusion collapses. This is GamiWays' number one problem.
Où est Ava ?
Où est Ava ? reduced latency by launching several operations in parallel rather than in sequence.
Le Dilemme Plastique
Le Dilemme cached the most frequent responses to avoid regenerating them each time.
GamiWays response
GamiWays targets a strict goal: respond in under 2 seconds. This requires rethinking the architecture from the ground up.
See latency & UX analysis02
The avatar that forgets everything
At each new session, the avatar starts from scratch. It doesn't remember your name, your preferences, or what you said the day before. The experience stays shallow.
Où est Ava ?
Où est Ava ? summarized the conversation every 4 exchanges to keep track of important facts.
Le Dilemme Plastique
Le Dilemme stored key student information in a permanent instruction file.
GamiWays response
GamiWays is building a three-layer memory system: what just happened, what we know about you, and how you prefer to interact.
See the memory system03
The voice that sounds off
A poorly tuned synthetic voice or one that mispronounces French words immediately breaks immersion. Voice quality is a critical issue, especially for education.
Où est Ava ?
Où est Ava ? tested multiple ElevenLabs settings to find the right balance between naturalness and clarity.
Le Dilemme Plastique
Le Dilemme precisely measured transcription error rates to validate voice recognition quality.
GamiWays response
GamiWays compared 25 voice synthesis engines and 15 recognition engines to identify the best French options.
See voice benchmark04
The conductor who loses the thread
Coordinating multiple AI systems simultaneously — one to listen, one to respond, one to manage the scenario — is very complex. When it goes wrong, the experience becomes incoherent.
Où est Ava ?
Où est Ava ? designed an autonomous 'Game Master' that makes narrative decisions independently from the rest of the system.
Le Dilemme Plastique
Le Dilemme used a visual tool (Flowise) to assemble 28 AI blocks, but the complexity became a bottleneck.
GamiWays response
GamiWays is building a 7-dimension context engine that clearly separates the role of each component.
See multi-agent orchestration05
The avatar that doesn't react
An avatar that speaks without changing expression, without moving naturally, without reacting to the interlocutor's emotion — it's unsettling. The technology to fix this exists, but it's expensive and slow.
Où est Ava ?
Où est Ava ? tested pre-recorded videos with dynamic triggers to simulate reactions.
Le Dilemme Plastique
Le Dilemme opted for a photo thumbnail rather than a video avatar, lacking a satisfactory solution.
GamiWays response
GamiWays is studying new real-time avatar platforms (Simli, HeyGen, D-ID) to find the right quality/cost/latency tradeoff.
See avatar benchmark06
Data leaving the country
Most AI services are hosted in the United States. For an application used in classrooms or with minors, this is a major legal and ethical issue in Europe.
Où est Ava ?
Où est Ava ? secured API keys server-side to avoid any client-side data exposure.
Le Dilemme Plastique
Le Dilemme identified the GDPR issue as a barrier to adoption in educational settings.
GamiWays response
GamiWays targets an architecture where processing can be hosted in Switzerland or Europe, with open-source models if needed.
See sovereignty roadmapFrom these two prototypes, GamiWays was born.
Discover the product vision, target architecture and Core Engine build status.