Voice Leaderboard

The Global Leaderboard ranks all AI voices based on their overall performance (ELO score) across all user votes in the Voice Arena. The Personal Leaderboard shows how *you* specifically ranked the voices based on the points you awarded during your participation.

ELO is a rating system used to calculate the relative skill levels of players (or in this case, voices) in head-to-head comparisons. A higher ELO score indicates a voice has performed better overall against other voices in the Arena, based on community votes.

Points on your Personal Leaderboard reflect how many times you voted for a specific voice as being better ('A is Better' or 'B is Better') during your Voice Arena sessions. It's a direct measure of your preference.

The confidence interval (e.g., ± value) associated with the ELO score gives a range within which we are statistically confident the 'true' ELO rating of the voice lies. A smaller interval generally means we have more data and are more certain about the voice's ranking.

'Appearances' indicates how many times a particular voice has been featured in a comparison round across the entire Voice Arena globally. More appearances generally lead to more reliable ELO scores.

We are gradually assigning memorable names (like 'Seren', 'Stone') to our voices. Some voices, especially newer or experimental ones, might still appear with their internal key (a unique code) until they receive an official name. The mapping helps associate the key with a known name where available.