Voice agent latency isn't mostly the LLM — turn detection can eat ~1 second
anandwana001 · reddit · 2026-11-23
The author (an Agora dev advocate, disclosed) breaks down where voice agent latency goes: real-time network 150–300 ms; VAD/turn detection 720–940 ms (640 ms silence window); STT to last word 400–700 ms; LLM first token 250–1000 ms; TTS first byte 100–350 ms.
The biggest hidden cost is deciding the user finished talking — wait too long and it feels slow, jump early and it interrupts. Agora's Conversational AI Engine addresses this layer with semantic turn detection (English/Chinese), interruption handling, and filler phrases, plus bring-your-own LLM/TTS/MCP or managed mode. Quickstart via npx skills add AgoraIO/skills in Claude Code or Cursor.
More from coding & agent
- Uncensored 26.4B model runs on a 12GB Mac at 85 tok/s with 8 parallel agents via Locals — SumitGup · 2026-11-23
- spaCy creator Honnibal shares a cheap-LLM ensemble recipe for building NER data with coding agents — honnibal · 2026-11-23
- 398M tokens, 98% cached: how one dev ran 12.5h of agent game dev on 3% of a $200 Claude plan — Yamapama · 2026-11-23
- Harvard paper: higher SWE-Bench scores didn't translate into more merged pull requests — rohanpaul_ai · 2026-11-23
- Solo dev ships AquaPaths puzzle game in 6 months — 9 of 10 commits co-authored by Claude Code — Pleasant_Present_160 · 2026-11-23
- A Five-Level Maturity Model for Safe Agent Interaction with Applications — crotalul94 · 2026-11-23