Is there an API for coaching institutes and edtech partners to integrate?

Yes — sayit exposes a speech-scoring API built specifically for coaching institutes and edtech partners who want to embed pronunciation feedback into their own product rather than send students to a separate app. Post a student's recording to one endpoint, and within seconds you get back a full scorecard: an overall score, per-word verdicts with the IPA of what the student actually said next to what they should have said, the specific sounds that went wrong with a plain-English fix, fluency, intonation, word stress, pacing, and, for IELTS or TOEFL prompts, an estimated band with examiner-style notes.

30-second version: The core call is POST /analyze — send audio, optionally with the target text the student was reading, and get back the same detailed scorecard sayit's own app renders. Exam-specific scoring, placement estimation, TTS for model audio, and per-student progress endpoints all sit alongside it. There's no speech-to-text "auto-correct" hiding behind the numbers — the phoneme recognizer has no language model, so it reports the sounds it actually heard.

What's the actual shape of a scoring call?

Multipart audio in, structured JSON out. If you send the passage the student was supposed to read alongside the audio, you get reference-mode scoring — per-word accuracy against that exact text, plus completeness. Leave the target text out and you get freeform scoring — the same phoneme-level accuracy, but on whatever the student actually said, the way an open IELTS or TOEFL answer works. The response includes per-word target and produced IPA, a named issue and a plain-language hint for anything flagged, fluency (words per minute, long pauses, filler words), an intonation reading, and a stress score — the same detail described in target versus produced and accuracy, fluency, read, and pace.

Can I score IELTS or TOEFL prompts specifically?

Yes — a dedicated exam-scoring endpoint takes the recording plus a prompt reference and returns the full analysis alongside an estimated band, examiner-style notes tied to specific observations in the recording, and a band-gap reading against a target score you name. Every estimate is explicitly labeled as not an official score — it measures the criteria an audio-only system can actually hear, described in full in the IELTS scorer and its honest limits.

What other tools does the API expose, beyond scoring a single recording?

Text tools for turning your own material into scored content — sending any passage through IPA and readability analysis, or generating model-answer audio to play back. A placement endpoint that takes a scored result and estimates a rough CEFR or exam level plus a first-language guess, useful for routing a new student to appropriately leveled material. And progress endpoints per student and, where an institute is provisioned, per class — a gradebook, cohort statistics, and rollups.

How do institutes actually integrate today?

Through a provisioned account your own server authenticates with, forwarding student audio to the scoring endpoints and rendering the results in your own interface — students never see sayit's branding, login, or pricing in that setup. A more granular, per-organization API-key model is a planned next step rather than something live today; the account-based integration path is what's fully supported right now. Per-student history in that setup is your responsibility to store, unless you'd rather sayit hold it, which the classroom console is built for directly.

What honest limits should I know before building on this?

Two, specifically: estimated exam bands are never official, and are labeled as such in every response. And latency runs several seconds per take through the scoring queue — this is not built for a real-time, sub-second response, so plan your UI around a short wait rather than an instant one.

The API at a glance

Try it

If you're evaluating an integration, get in touch to discuss a pilot — or see the classroom console if a ready-made interface, rather than a raw API, is the better fit for your institute.