# Sub-second voice agent pipeline

> Research and engineering to bring voice-to-voice response time low enough for natural phone conversations.

Source: https://aibyos.com/research/voice-latency

Research · Voice & speech

# Sub-second voice agent pipeline

Research and engineering to bring voice-to-voice response time low enough for natural phone conversations.

**<1 s**target voice-to-voice latency

**streaming**ASR → LLM → TTS

**barge-in**natural interruptions

## Pipeline

1.  Measure the latency budget of each stage
2.  Streaming ASR with early end-of-turn detection
3.  Small, fast model for the first response; large model for complex turns
4.  Streaming TTS with sentence-level chunking
5.  Load tests on real telephony

Figures show the typical order of magnitude for this approach compared with calling a large general-purpose model. Actual results depend on the task and data; we measure them on your data during the baseline phase.

[Discuss a project like this](https://aibyos.com/contact)

## Typical tooling

-   Pipecat / LiveKit
-   Whisper / Deepgram
-   vLLM
-   ElevenLabs / open TTS

## Related

[All research](https://aibyos.com/research)[Sovereign GPU compute](https://aibyos.com/gpu)[Fine-tuning](https://aibyos.com/services/fine-tuning)