Choose the voices
The same twenty male and twenty female characters from the English study, selected using recorded usage.
The same forty characters, now across eight languages. Matched voices, matched words, and two speech models. Listen first, then reveal the blind review.
Matched inputs.
Reversed listening order.
Model names and judge scores start hidden. Your picks stay in this browser.
Preparing the listening room…
Try another name or reset the filters.
The same twenty male and twenty female characters from the English study, selected using recorded usage.
Each language uses one matched passage combining conversation, emotional delivery, and pronunciation. Both models receive the same reference voice, text, and settings.
Gemini 3.1 Pro hears anonymous A/B samples. Each comparison is repeated in reverse order to expose order-sensitive judgments.
One automated judge is not a human preference study. These passages do not establish quality across every dialect or live-call latency, reliability, retention, or conversion.