Latency
Latency is the delay between a person finishing what they say and the AI Avatar starting to answer. Low latency makes a conversation feel natural; long pauses make it feel broken. Vendors quote different kinds of latency, so the numbers aren't always comparable.
How latency adds up
A reply passes through speech recognition and turn detection, the language model, text-to-speech and the avatar render, then travels over the network. Each step adds time.
- Avatar render latency: the time the face layer takes to turn audio into video.
- End-to-end (reply) latency: from the person finishing a sentence to the Avatar starting to answer.
- Connection time: from pressing start to the first visible frame.
When comparing vendors, ask which of these each number measures.
How Beyond Presence does it
Genesis 2.0's Avatar render latency is about 100 ms. With Managed Agents, end-to-end reply latency is about 1.0–1.2 s, and global end-to-end latency is up to 1.2 s. The language model and voice take most of that time, so model choice matters as much as the face. Read the AI agent stack in 2026.
Frequently asked questions
How fast is Beyond Presence?
Genesis 2.0 renders the Avatar in about 100 ms. With Managed Agents, a full reply starts about 1.0–1.2 s after the person stops speaking, and global end-to-end latency stays at or under 1.2 s.
What takes the most time in a Beyond Presence reply?
The language model and the voice. The Avatar render is about 100 ms of the total.
How can I make Beyond Presence replies faster?
Choose fast LLM and voice models, keep answers short, and run close to your users, for example in the Beyond Presence EU cloud for European users.