Blog

Why we didn't build an accent selector

2026-07-23 · AutoCast

When we built out English as a target language for dubbing, the obvious feature to ship alongside it was an accent selector — a dropdown letting you pick American, British, or Australian pronunciation for your cloned voice. It tests well in mockups. It sounds like exactly the kind of control a creator dubbing for different regional audiences would want.

We didn’t build it. Here’s why.

The dropdown would have lied

No text-to-speech system available today — ours included — actually controls pronunciation at the accent level from a voice clone. What we can control is regional word choice and spelling: “elevator” vs. “lift,” “color” vs. “colour,” “truck” vs. “lorry.” That’s genuinely useful for a script meant for a UK or Australian audience. But it is not the same thing as the voice sounding British or Australian, and a dropdown labeled “Accent: British” implies the second thing, not the first.

We talked through shipping it anyway. The case for it was straightforward: creators want it, competitors imply they have it, and most viewers wouldn’t notice the gap between “sounds different” and “is engineered to sound American vs. British.” That’s exactly the problem. If it would have gone unnoticed, that’s because it would have worked as a placebo, not as a feature.

What we shipped instead

AutoCast’s English target options are labeled by what they actually change — vocabulary and spelling for US, UK, or AU audiences — not by an implied pronunciation promise we can’t back. It’s a smaller claim than “accent selector.” It’s also one we can stand behind if a customer asks us directly what it does, which is the bar we try to hold every feature description to.

This isn’t a story about restraint for its own sake. We’d ship a real accent-pronunciation control the day the underlying voice technology actually supports it — that’s a feature request we hear and take seriously. Until then, we’d rather sell you the feature that exists than the one that photographs well in a screenshot.

See supported languages for exactly what each target language variant changes, and creating your voice for how voice cloning works today.