Editorial preview: vendor recordings and hands-on verdicts are pending. What counts as evidence.

A small publication.
An audible standard.

I'm Hunter Smith, an AI engineer based in Melbourne. I'm building The Voice Bench for people who already know they need a voice tool and are deciding which one deserves their time and money.

The useful part of a voice review should be something you can hear. That means matching the scripts, keeping the awkward takes and explaining which limitations matter for a real project.

Why this exists

A feature list can tell you whether a control exists. It cannot tell you whether the performance works, whether a correction is repeatable or how many attempts a finished passage will take. This publication is designed to leave those details visible.

What is ready, and what is next

The first edition contains source-checked research notes, a published test method and a working listening interface. The vendor test recordings and hands-on verdicts are still pending. Demo audio is identified as such wherever it appears.

That distinction matters. Research can narrow a shortlist; only the actual audition should earn a listening score.

Contact the editor or report a correction. The contact page identifies whether the public mailbox is available.

How it will be funded

The intended business model is affiliate commission when a reader chooses a tool through a tracked link. There are no affiliate links configured in this edition. Future relationships will be disclosed on the affected pages and cannot change a score.

Read the method or read the affiliate and privacy policy.