Speech-to-Text Latency Benchmark

How fast are real-time Speech-To-Text APIs across the world? We measure time-to-first-transcript for different API providers with router.audio, so you can see the real-world latency your users in different regions will experience. Hover over a location to see the per-model breakdown.

We measure latency using a 4 second clip of the text Hey, how's it going, with a 138ms leading silence and around 2 seconds of trailing silence. The latency is measured from the start of audio and the leading silence duration is deducted from the measured latency. Partial transcript latency is not measured. This benchmark is updated daily.