¶ What is Tritium AI?
Tritium AI is the complete human-robot interaction stack that sits behind every conversation your robot has — bringing together the intelligence to listen, understand, and respond; the expressiveness to react naturally in the moment; and the social awareness to know who it's talking to, what language they speak, and what the situation calls for. It powers Agents, the interface where you define your robot's personality, knowledge, and behaviour.
Powered by AI tokens, everything runs through a single integrated platform. No managing separate AI subscriptions. No stitching together disparate services. Just a robot that shows up to every interaction ready to perform.
Tritium AI provides:
- Speech recognition (ASR) — converts spoken audio into text
- Large Language Models (LLMs) — generates intelligent, context-aware responses
- Text-to-speech (TTS) — produces natural-sounding speech output
- Computer vision — recognises faces and objects in camera feeds
- Expressions — drives facial expressions in real time during conversation
These services run in the cloud because they are computationally intensive. The robot needs an internet connection for conversational features.
¶ AI Tokens
Tritium AI is powered by AI Tokens — a single balance that covers everything your robot does. Speech recognition, language model processing, text-to-speech: every interaction draws from one place. No separate subscriptions. No juggling providers.
AI Tokens are not equivalent to tokens from other providers (fer example OpenAI/ChatGPT) as each service may calculate token consumption in a different way.
¶ Token Usage
The Organization Usage page shows how your organisation's tokens are being spent across robots and Tritium users, so you can track consumption and top up before you run out.
Note: Token usage is available to customers on token-based pricing only, and only users with admin rights can view it. Admin rights are managed in the Marketplace; changes made there are synced to Tritium.
¶ Finding the page
Click your profile icon in the top right corner of Tritium and select Usage.
¶ Subscription and Top-Up tokens
Usage is split by where the tokens come from:
- Subscription tokens are allocated to individual robots and renew on each robot's renewal date.
- Top-Up tokens are purchased separately and shared across your whole organisation. Tritium users draw only from Top-Up tokens.
¶ What the page shows
At the top of the page, a bar shows how many Top-Up tokens have been used and how many remain since your earliest active purchase. Hover over a section of the bar to see the size and date of that purchase. If your Top-Up tokens run out, the bar turns red and shows No Top-Up tokens remaining.
Below this, three tabs break down usage in more detail:
- Overview: the five robots and five users with the highest usage. Select View All Robots or View All Users to see the full lists.
- Robot Usage: every robot with its serial number, renewal date and tokens used, split by Subscription and Top-Up. You can search, choose a time period and sort the list.
- Tritium Usage: Top-Up tokens used by each person in your organisation. You can search, choose a time period and sort the list.
¶ Agents
Agents is where you configure how your robot listens, speaks, and behaves, built on top of Tritium AI. Create an agent once, then assign it to one robot or across your fleet.
Three platforms are available, each suited to a different level of control:
- Core — configured entirely in Tritium UI: prompt, voices, languages, knowledge base, and abilities. The fastest way to get a capable agent running.
- Pro — connects to the Engineered Arts Agent Builder for advanced control over models, data sources, and tools.
- ElevenLabs — connects your robot to a conversational agent built in ElevenLabs, for voice-first, multimodal agents.
See Agents Interface for full setup and configuration steps.
¶ Multilingual Conversation
Your robot can converse in multiple languages — switching naturally based on who it's speaking with, with accurate lip sync and expressions throughout. Select a primary language and add as many secondary languages as you need, each with its own voice.
¶ Expressions
Tritium AI drives your robot's facial expressions in real time during conversation. As your robot speaks, its face responds — lip-syncing accurately, shifting naturally between emotional states, and staying visibly present even when listening.
- In sync — speech and mouth movement are tightly synchronised in real time
- Emotionally fluid — the robot moves naturally between expressions without snapping between states
- Always present — subtle, continuous movement during listening means the robot never looks blank
¶ Group Conversation
Tritium AI's multiperson speaker detection identifies who is speaking at any given moment and directs the robot's attention accordingly. When a person speaks, the robot looks at them. When someone else starts, it switches cleanly.