What voice does sayit use to read words and passages aloud?

sayit's default reference voice is American English, generated by a neural text-to-speech pack, and you can switch it to a genuinely different British voice in settings — not just a relabeled version of the same audio. What's worth knowing honestly: today, the pronunciation you're actually scored against stays American English regardless of which voice you choose to listen to, because accent-matched scoring is a known, measured limitation rather than something quietly working under the hood.

30-second version: Pick American or British in settings and the voice reading passages and words back to you genuinely changes — real distinct voices, not a label swap. Australian, Canadian, and Indian options exist too, recorded as a preference for now rather than swapping in dedicated voices yet. Separately, and independent of which voice you picked: your own pronunciation is compared against American English targets across the board today. That's a real, currently-measured limitation, not a hidden one — sayit's own engineering notes call it out explicitly rather than quietly shipping a mismatch.

Is the British voice actually different, or just a label?

Genuinely different audio — the underlying voice pack ships distinct British voices alongside the American ones, with real differences in vowel quality and rhythm, not a relabeled American voice. When you switch your accent setting to British, what you hear for a "hear it correctly" playback changes accordingly.

What about Australian, Canadian, or Indian?

Those are real settings you can select, and they're recorded against your account, but today they don't yet swap in a dedicated voice the way American and British do — American and British are currently the two accents the voice pack ships genuinely distinct audio for. Selecting one of the others records a preference rather than changing the audio you hear right now.

Does switching my voice to British also change what I'm scored against?

Not today — and this is worth being direct about, because it's easy to assume otherwise. The target pronunciation your speech is compared against — the canonical phonemes behind every accuracy score — is generated in American English regardless of which playback voice you've selected. A non-rhotic British-style "car," correctly dropping the final /r/ the way real British speech does, can be measured against an American target that expects that /r/ to be there. This is a known limitation, not a silent one: it's documented internally as a real, measured source of unfair scoring for learners deliberately targeting a non-American accent, and turning on a fully accent-matched target has its own trade-offs that haven't yet been resolved cleanly enough to enable as a default.

Why not just turn on accent-matched scoring, then?

Because it isn't a clean win yet. The acoustic model that listens to your speech has its own American-leaning bias, so even a speaker genuinely producing non-rhotic British speech doesn't always get recognized as rhotic-free consistently — which means matching the target to British without also reliably reading the learner's actual accent from the audio can misalign scores in the other direction just as easily as it fixes them. Getting this right needs a better signal than a settings toggle, and that work isn't finished.

So does sayit penalize me for a legitimate non-American accent feature right now?

For accent features that are simply regional coloring within a phoneme's own territory — flapped /t/, vowel quality, general rhythm — no; those are tolerated by design, covered in how sayit handles accents. The specific, narrower gap described here is about targets that are structurally different, like whether a final /r/ is expected at all, and it's the one place sayit is upfront that American English is still the yardstick today, regardless of your chosen playback voice.

Voice and scoring target, honestly

Try it

Switch your voice to British in Settings and listen to a word on /pronounce — the audio itself is real and different. Just know that your own scoring target hasn't moved with it, yet.