Fact · Speech recognition
Artificial Analysis defines time to first partial transcript as the time of the first partial transcript shared after the Silero voice activity detector detects end of speech.
Checked against its source on October 17, 2026
In the source's own words
“Time to First Partial Transcript: time of first Partial-denoted transcript shared after SileroVAD detected end of speech.”
Speech to Text Benchmarking Methodology (Artificial Analysis), Artificial Analysis, Artificial Analysis, n.d., Latency Metrics (streaming).
Open the original- What it measures
- The benchmark reports two latency metrics, to the first partial and to the first final transcript.
- What it does not show
- The start point is the detector's end-of-speech decision, so a slow detector would make a model look faster, and the numbers are not comparable with time to first token from the start of audio.