Elispeak's answer

Real-time pronunciation and fluency scoring on every utterance, not just transcription. Multiple AI tutor characters with distinct voice personas so conversation feels varied. Exam-mode practice tuned to TOEFL, IELTS, and CELPIP public rubrics with exam-safe band-descriptor framing (no fake scores). Sub-1.5-second feedback latency so the loop feels like speaking to a person, not waiting on an API.

Elispeak's answer

Other speaking apps either stop at transcription or ship one voice for every scenario. Elispeak gives per-utterance pronunciation, fluency, and grammar feedback at conversational latency (under 1.5 seconds end-of-utterance) and ships multiple tutor personas so a job-interview rehearsal feels different from a relaxed coffee chat. Exam modes use the published TOEFL / IELTS / CELPIP rubrics with safe band-descriptor framing — no fake scores, no bait-and-switch. And it works free: you can do your first real reps before ever seeing a paywall.

Elispeak's answer

Adult English learners at B1 to C1 who need speaking for real outcomes, not grades. Four main segments: (1) relocators and expats preparing for visa / customs / tenancy interviews, (2) remote job seekers rehearsing technical and behavioral interviews in English, (3) customer-facing professionals (support, sales, nursing) who read and write fine but freeze on the phone, and (4) TOEFL / IELTS / CELPIP candidates who need exam-mode drills between tutor sessions. Consistent pattern: they can compose a sentence on paper in 10 seconds but need 30 when the mic opens, and they want low-stakes reps that don't cost tutor-hour money.

Elispeak's answer

Traditional apps teach vocabulary and grammar well, then quietly hand speaking off to expensive tutors or unpredictable language-exchange partners. We kept hitting the same wall with our own users: they could read and write English comfortably but froze the moment they had to speak to a human. Scheduling a tutor for a 5-minute drill is absurd; going to a class in-person is a weekly event, not a daily habit. Elispeak started as a narrow experiment — can real-time speech-to-score feedback close the gap between thinking of a sentence and saying it? Once the latency and scoring felt right, we built AI tutor personas so every session felt like a real conversation, added exam modes for TOEFL / IELTS / CELPIP, and kept the free tier generous so learners could form the habit before the paywall.

Elispeak's answer

Streaming speech recognition with word-level timestamps and phoneme probabilities for pronunciation scoring. Phoneme-level alignment and edit-distance features for pronunciation scoring; stream-level features (pace, filler-word density, pause distribution) for fluency; an LLM pass scoped to published rubric criteria for content scoring. Neural conversational text-to-speech with a separate voice-persona config per AI tutor character. WebRTC for browser audio capture and low-latency round-trip. Stripe for billing, GA4 for attribution, and a pinned model version per exam mode with a held-out eval suite to catch regressions before upgrading.

Elispeak's answer

Elispeak is currently direct-to-consumer. The largest user cohorts are individual learners preparing for immigration-linked speaking exams (IELTS, CELPIP) and remote job interviews in English. We do not have named enterprise logos yet — the product is built around one-to-one practice, not a seat-based team subscription. Teacher and speaking-club partnerships are an active pilot channel for 2026.