Starting from Où est Ava ?, Edugami and GamiWays Core, this page turns research into product, partnership and business-model questions for the next settings.
What this page does not repeat: model comparisons, performance benchmarks and research protocols. They live in the state of the art. Here the question is: which experience should we dare, with whom, under which conditions, and how will we know it deserves to continue?
Start from the role conversation plays in a cultural or learning journey; defer model, voice and avatar choices to the existing comparisons.
Sell framing, scenario, facilitation and evolution; make inference and streaming costs visible instead of hiding them in a vague SaaS promise.
The Core connects replaceable bricks. Content, distribution, infrastructure and training alliances matter as much as the next multimodal model.
Transparency, rights, data, oversight and human handover are components of the experience and institutional adoption criteria.
Established findings, sources and related protocols remain in the state of the art. The scenarios below start from them without duplicating their tables.
Open state of the artNone of these scenarios is a product promise. They are directions of work combining a setting, offer shape, partners, an explicit uncertainty and a test able to advance or stop the hypothesis.
01 · Guided experience
Perspective question
How can a conversation with a character become a mediation experience visitors choose, understand and leave with a useful trace?
Où est Ava ? can become a format co-produced with venues, festivals and rights holders: editorialised, bounded in narrative scope and supported before and after the conversation. The future to test is not an avatar everywhere, but a situated conversational presence at the right moment in a cultural journey.
Watch point
The experience must preserve the fiction/reality boundary, representation rights and an explicit way out of the conversation. Narrative quality matters more than generated minutes.
Assets to mobilise
Character System · asynchronous Game Master · scenario back office · streaming video avatar
Value and offer shape
Design and deployment engagement, complemented by event operation or time-bounded scenario licensing.
Ecosystem to compose
Producers, rights holders, mediators, cultural venues, voice/avatar partners and audiovisual production teams.
First decisive test
Compare classic mediation and a conversational session around the same work: comprehension, desire to continue, perceived relationship quality and facilitation cost per visitor.
Framing principle
Framing, scenario, facilitation, integration and maintenance should not be confused with API, streaming, compute or rights costs. Separating them from the pilot avoids confusing a technical success with a sustainable offer.
Framing & scenario
Design fee: intent, audience, character, mediation rules, journey and evaluation.
Integration & operation
Build, connectors, staging, team training, on-site presence or operational support.
Variable & reversible
Models, voice, avatar, hosting, streaming and rights remain traceable and replaceable.
Venues, schools, producers and trainers bring context, audiences and usefulness criteria. They should be involved in the scenario and test protocol, not only distribution.
Gamilab, voice/avatar/LLM providers, hosts and observability tools compose an evolving architecture. Core adapters preserve the ability to replace a brick without rebuilding the project.
Open protocols to connect agents and tools become an experimentation ground. They matter when a partner or client needs to connect its own data and services; they do not replace governance work.
Compliance is built with business teams, data owners and mediators. AI information, human roles, quality measures and reversibility become partnership deliverables.
Put the four scenarios in contact with real settings. Document what is reusable, what needs a partner, and what audiences reject or misunderstand.
Turn working formats into modular offers: framing, scenario, integration, support, operation and measures. Keep provider alternatives and fallbacks.
Open contracts or SDKs only after observing cross-project reuse. The platform is a consequence of successful scenarios, not a substitute for them.
Multimodal ambition does not justify collecting unnecessary data, opacity about AI use or an avatar simulating a relationship beyond its role. In educational settings, design choices must in particular exclude prohibited emotion-recognition uses and preserve meaningful human oversight.