Intelligence that learns the show
Speech, script and subtitles are the current core. Production memory, multimodal understanding and human-approved coordination form the research roadmap.
The CREW: rehearsal, anticipation and recognition
The Understudy represents rehearsal preparation: reading the script and learning its rhythm. The Prompter anticipates scripted lines. The Interpreter recognizes and translates improvisation. Speech and script are the existing input core; deeper learning and adaptation remain development work.
The production world model is an R&D goal
Ariel’s research direction combines production memory, live state and human intent. The aim is to understand what should happen, recognize what changed and suggest an appropriate response. Multimodal sensing, automated device control and transfer to unfamiliar rooms are future research goals.
Humans author. Intelligence assists.
Rehearsal is where people express intent and correct the system. The planned learning cycle uses authorized rehearsal and correction data, with evaluation before learned capabilities are deployed. Human approval and operator override guide the architecture.
Live inference stays local. Optional future pre-production or model-development workflows require separate data-use choices. The roadmap does not assume permission to reuse a production’s scripts, recordings or audience information.
The roadmap for live intelligence
Company roadmap, 9 September 2026. Future deliverables are targets, subject to development and validation.
- 2026
Supertitle delivery
Speech and script are the existing core. The focus is delivery in live performance, with Japan as the first market and a CEATEC keynote planned for 15 October 2026.
- 2027
Production memory
R&D target: package production memory and a model of show state. Planned toolkits and training would help venues operate the system themselves.
- 2028
Live performance assistant
R&D target: understand multimodal events and adapt to changes. A planned Taiwan–Japan co-production would provide a development setting.
- 2029
Venue coordination
R&D target: coordinate venue systems through human-approved actions. A first Broadway production is a commercial target.
- 2030+
Partner licensing
Long-term target: license the intelligence to partners and research how it transfers to unfamiliar environments. General-purpose intelligence has not been validated.
How does Ariel relate to AGI?
Ariel’s AGI-era strategy is to develop intelligence for live human environments, beginning with theatre. It is a domain-focused edge AI roadmap. Ariel does not claim to have built artificial general intelligence or validated general-purpose autonomy.
Discuss a project