Run 2 · 15 October 2026
V2, V3 and V3.5
100 trials per version · 300 attempts
The three-version comparison examines speech availability, startup time, full audio delivery and variation across repeated inputs.
Explore run 2Kokekun · benchmark archive
Translation and speech,
measured side by side.
Explore the recorded comparisons behind Kokekun’s development. Each run includes timing patterns, pronunciation guidance outcomes, saved audio and downloadable measurements.
Version identity: V2 (public release at test time) is the version labeled “main” or “Current main” in the saved figures and source records. These runs measured kokekun.com on 15 October 2026; they did not separately test the /v2 installation. A later change to the public site does not change these archived results.
Run 1 · 15 October 2026
V2 and V3
100 trials per version · 200 attempts
The earlier comparison explores where v3 gains time, where pronunciation guidance fails, and how patterns differ across input groups.
Explore run 1Each run uses 50 distinct inputs twice, in English and Japanese, with typed text and saved recordings. Timing measures received audio bytes; it does not measure audible onset or Bluetooth delivery. These are archived observations of the versions tested, not a live service-status page.