The Platform

Our platform is where the magic happens.

Developed in-house over three years, the Captivation Platform brings together agentic AI, speaking avatars and rich media to create engaging, human-like digital experiences.

Contact
Captivation avatar
CAPTIVATION
RENDERING
ENGINE
Frontend
Widgets | WebApps
MAIN LLM
CAPTIVATION PLATFORM
Voice | Avatar | Gestures | Media
INTEGRATIONS
MCP, UCP,
API, OTHER
DATA

Super Fast Connections. Low-Latency Responses.

Captivation is engineered to make Al conversations feel fast, fluid and natural.

50-100ms: Initial WebSocket connection ~100ms - Filler response from user input

150-250ms: LLM response to avatar speech with stored voice

550-800ms: LLM response to avatar speech with newly generated voice

300-400ms: Average response time from LLM output* Fast enough to make a conversation feel like a conversation.

*Based on 60-80% of responses using stored voice files.

Our patent-pending proprietary technology combines multiple AI services and leading technologies with robust, proven engineering. This gives us the flexibility to use the right tools for each task while optimising compute, latency and cost without compromising the user experience.

Built for scale, flexibility and reliability, the platform integrates easily with third-party systems and services, supporting Captivation solutions across a wide range of industries and use cases.

Hosted in leading international data centres, the platform is designed to meet enterprise expectations for security, resilience and performance.

Platform FAQs

  • Instead of continuously generating and streaming a new video, Captivation uses pre-created visual assets that are delivered to the browser and brought to life using lightweight real-time instructions from the Captivation Platform. Voice, text, lip sync, gestures and expressions are synchronised as the conversation happens.

  • Browser rendering dramatically reduces the amount of server-side video generation, compute and bandwidth required. This helps Captivation deliver faster connections, lower operating costs and a lighter technical footprint than continuously streamed AI video.

  • Turning LLM text into voice is one of the more compute-intensive and expensive parts of a speaking Avatar service. Captivation can match around 80% of response phrases to audio already generated for a particular voice and reuse the stored file instead of making a new text-to-speech request. New speech is still generated dynamically when required.

  • The Media Window allows relevant products, images, video, graphics, maps, comparisons, forms and other visual content to appear alongside the Avatar and in sync with the conversation. Captivation's Media Trigger Service selects content according to the conversation and what the user is viewing.

  • The platform is designed to be LLM agnostic. Different AI models and services can therefore be used for different tasks, allowing Captivation to optimise for the required combination of capability, speed and cost.

  • Yes. Captivation is designed to work through common modern web browsers across smartphones, tablets and PCs without requiring the end user to download an application.

  • Captivation AI has filed patent applications covering several aspects of its technology and platform.