Can our institute run sayit on our own servers, or offline?

No — sayit doesn't offer an on-premise or offline install. It's a hosted cloud service: recordings are uploaded to sayit's servers to be scored, for every account, on every plan, including institutional ones. If your institute is evaluating pronunciation software and specifically needs audio to never leave your own infrastructure, that's a real constraint worth knowing up front, honestly, before a pilot rather than after one.

30-second version: sayit's scoring engine runs server-side because the underlying models are the same for every learner and need consistent, server-grade compute — there's no lighter on-device mode and no self-hosted deployment option for institutions today. What an institute gets instead is control over the data terms: recordings processed for scoring and deleted unless you specifically opt into retention, no use of student recordings to improve sayit's models without consent, and per-student data export on request.

Why isn't there an on-premise option?

Because the scoring pipeline — the phoneme recognizer, the alignment engine, the language models behind generated content — is substantial enough that running it well needs the same server infrastructure for every account, not a smaller model that could reasonably run on a school's own hardware. Building and maintaining a genuinely separate on-premise deployment is a different product commitment than operating one hosted service well, and it isn't one sayit makes today.

If the audio has to leave our building, what actually protects it?

The data terms an institute agrees to, not the server's physical location. In practice: recordings are processed to generate scoring and, unless your institute specifically opts into longer retention, deleted afterward rather than kept indefinitely. Recordings are never used to improve sayit's underlying models without consent — the same consent mechanism that governs any individual account's training-data choice applies at the institutional level too. And your institute can request an export of its own students' data on demand, so nothing about the relationship locks your data inside a system you can't get it back out of.

Does this mean student data is treated the same as any individual user's?

The same core protections apply — recordings kept only as long as needed, never sold, deletable — with institutional terms layered on top for retention and export specifically, since a school or company needs those guarantees written into an agreement rather than left as a general product policy. The privacy page covers the general rules that apply to every account; institutional pilots add explicit retention and export terms on top of that baseline.

What if data residency (which region servers are in) is the actual concern?

That's a separate question from on-premise hosting specifically, and one worth raising directly if it's a hard requirement for your institute — it's the kind of detail that belongs in a pilot conversation rather than assumed either way from a general privacy page.

What this means in practice for an evaluating institute

Try it

If your institute is evaluating sayit and this changes the conversation, the honest next step is a direct pilot discussion rather than guessing from a marketing page — the classroom CMS and API for institutes pages cover what's actually built for institutional use today.