Voice agent latency

Latency is the difference between a conversation and a wait. Here is the budget and the breakdown.

  • Simba 3.2 produces first audio in under 100ms
  • Full pipeline lands under 1 second end-to-end
  • Simba 3.2 is #1 on Artificial Analysis

Definition

Voice agent latency is the gap between a caller finishing their sentence and the agent beginning its reply. It is not one number but a sum: speech-to-text, the LLM’s reasoning, text-to-speech first audio, and the network hops between them. Cross about half a second and the caller feels the wait.

The budget

The lower the total, the better. Inside the per-turn budget, each stage takes a share: transcription, reasoning, and speech generation. The art is keeping the total under the threshold where a human perceives the reply as immediate rather than delayed.

The breakdown

Speech-to-text runs as the caller speaks, so much of it overlaps the utterance. The LLM’s reasoning is variable and depends on the model and the prompt. Text-to-speech first audio is the last stage and the most visible, because it is the moment the caller finally hears something back.

Why first audio dominates

First-audio time is the last thing between reasoning and sound. A model that starts speaking in under 100ms keeps the reply feeling immediate even when the earlier stages take their share. SpeechifyAI’s Simba 3.2 does exactly that while sitting at #1 on the Artificial Analysis TTS leaderboard, so the speed does not come at the cost of voice quality.

FAQ

Frequently asked questions

What is voice agent latency?
Voice agent latency is the delay from the moment a caller stops speaking to the moment the agent starts replying. It is the sum of speech-to-text, the LLM's reasoning, and text-to-speech first audio, plus network time. When it exceeds roughly half a second, the conversation starts to feel like a wait.
What latency should a voice agent aim for?
The lower the better: the goal is a reply that feels immediate rather than delayed. The single biggest lever is text-to-speech first-audio time. SpeechifyAI's Simba 3.2 produces first audio in under 100ms, which leaves headroom for the STT and LLM stages inside the per-turn budget.
Why does TTS first-byte latency matter so much?
Because it is the last stage before the caller hears anything. Even if transcription and reasoning are fast, a slow text-to-speech start makes the whole reply feel late. A streaming-native model that emits first audio in under 100ms keeps the perceived response immediate.

Start building

Latency is the difference between a conversation and a wait. Here is the budget and the breakdown.

Privacy preferences

Choose what we may store on this device. You can change this at any time from the footer.

Strictly necessary

Sign-in, security, load balancing, and remembering your privacy choices. These cannot be switched off.

Always on

Analytics

How the site is used in aggregate - which pages get read, where people get stuck - so we can improve it.

Marketing

Measures which campaigns bring people here, and lets us show relevant ads on other platforms.