ConceptLatency
Response Latency
A key metric for the real-time performance of voice interaction systems, typically decomposed into first-packet latency (the time from when the user finishes speaking to when the system emits its first audio packet) and end-to-end latency (the total delay including network and inference). Human turn-taking in natural conversation averages about 200 milliseconds; beyond 500 milliseconds users noticeably perceive a pause.
Timeline (last 90 days)
Oct 24
If system latency exceeds 500 milliseconds, users will clearly perceive pauses and unnaturalness.
Unverified50%
Oct 18
Token consumption and response latency under long contexts directly affect the sense of immersion in an AI Galgame.
Unverified50%