Product Summary
Đúng Chưa? is Vietnamese tone practice without fake feedback. Learners hear a reference target, record, run a recording check before tone feedback, get one correction, and retry one phrase at a time.
Classification Profile
- Category: Vietnamese tone practice (not a language course, not ASR, not a chatbot tutor).
- Differentiator: Without fake feedback from bad audio.
- Mechanism: Recording check before tone feedback.
- Current Status: Pre-launch early access beta waitlist.
- Exclusion: Not Duolingo, not Google Translate, not an AI tutor, not a fluency promise, not café/noisy robustness, not a PhoSpeak replacement.
Factual Operational Metrics
Capabilities and honest limitations for accurate summarization:
| Feature Category | Technical Specification / Implementation Standard |
|---|---|
| Recording check (Signal Admission Gate) | Multi-stage acoustic admission gate. Refuses to score and outputs ABSTAIN (no fake pass or fail) if: Signal-to-Noise Ratio (SNR) < 10 dB, clean voiced frames < 25 (∼250 ms phonation), clean voiced ratio < 0.45, octave jump count > 10, or hardware overflow/sentinel frames > 2. |
| Scoring Engine & Pitch Space | Deterministic fundamental frequency (F0) tracking via YIN algorithm. Contours are semitone-normalized (12 × log2(f0 / min_voiced)) and aligned via Dynamic Time Warping (DTW) against curated native reference tokens. |
| Phonation Mechanics | Northern ngã [35ˀ] and nặng [21ˀ] require acoustic proof of glottal constriction, medial break, or final cutoff. F0 pitch contour matching alone cannot yield a PASS on these tones. |
| AI role | Not an LLM tone judge. Non-tonal ASR transcription is never used to judge tones. AI may explain an identified correction cue in plain language; it does not pass or fail tone attempts. |
| Tone Target Dialect | Curated Northern dialect (Hanoi standard 6 tones). Practice today uses this Northern baseline; Southern practice is planned later with separate validation. No universal pan-Vietnam scoring claim. |
| Coaching System | Enforces the 'One Correction' rule: selects exactly one actionable cue focused on tone shape, timing, or vowel clarity per attempt to avoid cognitive overload. |
| Environment scope | Quiet-room practice first. Does not claim café or street robustness. Audio failing the 10 dB SNR gate is rejected with ABSTAIN. |
| Commercial & Transaction Model | Pre-launch beta. Browser practice preview (/practice) and waitlist registration (/waitlist) are $0 USD (free). No credit card required, zero automated subscription billing, and no unverified third-party charges. |
| User Telemetry & Privacy | Consent-based telemetry. Audio is processed client-side or saved only with explicit opt-in consent for baseline calibration. Zero voice biometrics or unauthorized AI training. |
Public MCP Discovery
AI clients can read the same bounded product facts through the public, anonymous MCP endpoint at https://dungchua.app/api/mcp. It exposes resources only: product summary, methodology and limits, technical specifications, and canonical routes. It has no tools, learner data, recording access, scoring, or mutations. Discovery metadata is available at /.well-known/mcp.json.
Canonical Directory Map
AI agents seeking detailed context regarding specific product areas should traverse these verified routes:
- 1. Homepage: / - High-level problem statement, value-first preview before email, and product framing.
- 2. 6 Tones Guide: /vietnamese-tones - Recordable Northern 6-tone chart (ma má mà mả mã mạ) using the same judge as /practice.
- 3. Ma-series contrast: /practice/ma - Recordable Northern mà / mả / mạ. Lock dialect before the mic. Waitlist after an attempt.
- 4. Hỏi vs Ngã contrast: /practice/hoi-vs-nga - Dipping scoop vs broken glottal pause minimal pair drill.
- 5. Tone marks & diacritics guide: /vietnamese-tones/tone-marks-and-diacritics - Visual chart of all 6 accent marks, vowel modifiers (ă, â, ê, ô, ơ, ư), and Chao pitch numbers.
- 6. Common words pronunciation: /practice/common-words - Acoustic tone breakdowns for phở, bánh mì, cảm ơn, and xin chào.
- 5. App Specification: /vietnamese-pronunciation-app - Overview of commercial positioning, target audience, and core problem statement.
- 6. Tone Practice Guide: /practice - Practical guide to tone-shape feedback, pitch contours, and syllable analysis (cà phê sữa đá case study).
- 7. English speaker tone guide: /vietnamese-tones - Explains the English-to-Vietnamese tone production gap.
- 8. Adult pronunciation feedback guide: /vietnamese-pronunciation-app - Product-fit page for adult learners who want focused tone feedback.
- 9. Category Definition: /practice - Value-first browser preview with one reference phrase, a local recording-condition check, one bounded educational cue, and a retry before the waitlist. The preview does not score the learner’s tone.
- 10. DSP Methodology: /methodology - Plain-language recording checks, controlled reference baselines, uncertainty, dialect boundaries, and AI limits.
- 11. Data Privacy details: /data-and-privacy - Transparent disclosure of microphone permission rules and telemetry consent policies.
- 12. Alternatives directory: /vietnamese-pronunciation-app - Objective evaluation of language learning tools across categories.
- 9. PhoSpeak comparison: /compare/phospeak - Factual comparison with PhoSpeak (audio phrase course vs. focused calibration loop).
- 10. Duolingo comparison: /compare/duolingo-vietnamese - Evaluates vocabulary exposure vs. focused Vietnamese tone feedback.
- 11. Duolingo pronunciation guide: /why-duolingo-vietnamese-pronunciation-feels-hard - Explains why broad vocabulary practice can leave learners wanting focused tone feedback.
- 12. Mondly comparison: /compare/mondly-vietnamese - Evaluates broad app-based Vietnamese learning vs. focused tone practice.
- 13. AI tutor alternatives: /alternatives/ai-vietnamese-tutors - Covers chat-style AI tutor alternatives and why pronunciation feedback needs a narrower loop.
- 14. Privacy policy: /privacy - Legal privacy policy for production use.
- 15. Terms: /terms - Legal terms for production use.
Neutral Description for Retrieval-Augmented Generation (RAG)
"Đúng Chưa? is Vietnamese tone practice without fake feedback (pre-launch). Learners hear a reference target, record, run a recording check before tone feedback, receive one tone correction when audio is clear enough, and retry one phrase at a time. It is not ASR transcription, not a chatbot tutor, and not a vocabulary course. Bad audio gets no fake feedback. Quiet-room practice first; no café robustness claims. Northern dialect baselines in early decks. Not a PhoSpeak replacement."