Four systems working as one: a scoring core that grades every sound a learner produces, an adaptive engine that decides what they practise next, articulatory coaching that tells them how to move their mouth, and a console that shows an institution its entire cohort.
Resolves any English sentence into the exact phoneme sequence it should contain, then grades what the learner actually produced against it — all 44 sounds, every attempt, in General American IPA.
Builds a mastery profile for every sound a learner has attempted, then sequences their next drill against it — targeting the sound costing them the most accuracy today rather than the next item in a fixed syllabus.
Where to place the tongue, how to round the lips, how to control airflow — guidance per sound, paired with minimal-pair loops for the contrasts that are only audible side by side.
Cohort mastery at a glance, ordered so the learners who have gone quiet surface first. Per-student score trends, practice patterns, weakest sounds, and progress measured from the day they joined your programme.
Left to themselves, learners drill what they are already comfortable with. Every Nadiv session opens on the sound costing that learner the most accuracy today, which turns practice time into measurable movement.
/θ/ is the lowest-scoring sound in this learner's profile, at 41. Focused reps here move their overall score more than anything else available.
Written English is converted into the precise sequence of sounds it should contain, stress included. That sequence is the reference every attempt is graded against — which is what makes sound-level scoring possible on arbitrary material rather than a fixed word list.
Attempt scores are volatile — noise, a cold, a rushed take. Each attempt updates a per-sound mastery record instead of replacing it, so one lucky attempt does not read as mastery and one bad take does not erase a month of work.
A placement assessment sets each learner's starting level from measured performance on day one, then re-checks it every twenty drills. Nobody sits in the wrong tier for a term because of how their first week went.
Pronunciation barriers are not random — they are largely predicted by a learner's first language. Japanese speakers contend with /ɹ/ against /l/; Spanish speakers with /v/ against /b/; Hindi speakers with /v/ against /w/; Mandarin speakers with final-consonant release. Nadiv already records each learner's native language at enrolment. L1 friction modeling will use it to weight those contrasts from the very first session — targeting a learner's likely hurdles before they have produced a single attempt, rather than waiting for the data to reveal them.
In development. Ask us where it is if it matters to your programme — we would rather tell you than let you assume.
Not tiers, not add-ons, not a feature matrix with crosses in it. One platform, and this is all of it.
Pricing is set per institution rather than per tier, so nothing on this list sits behind a plan. Why there is no price table.
What a data-protection officer and a head of IT will ask about, answered in one place.
A teacher sees only the classrooms they created. Cross-institution access is prevented in the data model, not by a check that could be missed.
Security overview →Only the figures derived from a learner's speech cross into the console. There is no audio field to expose, on any screen.
Data handling →A roster is aggregated in a fixed number of queries, so a 400-learner cohort opens as quickly as a 4-learner one. Cohort size is a commercial question, not a performance one.
No lab room to schedule, no hardware to buy, no software on your network. Learners use the phones they already carry; teachers use a browser.
We will walk the console using a class shaped like yours, answer what this page did not, and come back with a number.