Real Study Ai Student Study App
Budget: $30 – $250 USD
I’m building Real Study AI: a hands-free, voice-first study assistant. The MVP should let a student talk to the app, get spoken questions, answer by voice, and hear instant feedback (“correct/incorrect”)—then repeat missed items until mastered. Keep it simple, 1 page, mobile-friendly.
Core Features (must-have)
Deck loader: Load 1 demo deck (10–20 cards) from CSV/JSON with fields: front (prompt) and back (answer).
Voice out (TTS): App speaks the front (question).
Voice in (STT): Tap mic → capture my spoken answer → convert to text.
Judge answer: Basic string match with small typo tolerance (case-insensitive; ignore punctuation; Levenshtein ≤1–2 ok).
Spoken feedback: AI says “Correct!” or “Not quite—answer is ___,” and shows it on screen.
Adaptive loop: Missed cards are re-queued and asked again later until correct.
Session stats: Show attempted, correct, and simple accuracy % at end; ability to export CSV (or write to Google Sheet).
Nice-to-haves (only if time allows — not required)
“Hint” voice command → read a short hint field if present.
“Repeat” command → repeat the last question.
Simple difficulty bump: if I get it right twice in a row, deprioritize that card.
Conversational Flow (example)
App: “What is the derivative of x²?”
Me: “Two x.”
App: “Correct! Next question: Capital of France?”
Me: “Paris.”
App: “Correct. Accuracy 2/2. Say ‘repeat’ or ‘hint’ anytime.”
Tech Notes (choose the quickest path)
Web is fine (one page). Use Web Speech API for TTS/STT if possible, or any free tier.
Framework: Plain JS or React (your call). No heavy styling needed.
No backend required beyond reading a local CSV/JSON (or a public Google Sheet).
Deliverables
A shareable link (or zipped project) + short README showing how I swap in a new CSV/JSON deck.
Source with clear comments for the TTS/STT and answer-checking parts.
Acceptance Checklist
Mic button captures my voice on desktop & mobile.
App speaks questions; I answer by voice; it speaks feedback.
Wrong answers come back later; right ones move on.
End-of-session stats + CSV/Sheet export.
I can replace the deck myself in minutes.
Budget & Timing
Budget: $0–$100 (Negotiable).
Core Features (must-have)
Deck loader: Load 1 demo deck (10–20 cards) from CSV/JSON with fields: front (prompt) and back (answer).
Voice out (TTS): App speaks the front (question).
Voice in (STT): Tap mic → capture my spoken answer → convert to text.
Judge answer: Basic string match with small typo tolerance (case-insensitive; ignore punctuation; Levenshtein ≤1–2 ok).
Spoken feedback: AI says “Correct!” or “Not quite—answer is ___,” and shows it on screen.
Adaptive loop: Missed cards are re-queued and asked again later until correct.
Session stats: Show attempted, correct, and simple accuracy % at end; ability to export CSV (or write to Google Sheet).
Nice-to-haves (only if time allows — not required)
“Hint” voice command → read a short hint field if present.
“Repeat” command → repeat the last question.
Simple difficulty bump: if I get it right twice in a row, deprioritize that card.
Conversational Flow (example)
App: “What is the derivative of x²?”
Me: “Two x.”
App: “Correct! Next question: Capital of France?”
Me: “Paris.”
App: “Correct. Accuracy 2/2. Say ‘repeat’ or ‘hint’ anytime.”
Tech Notes (choose the quickest path)
Web is fine (one page). Use Web Speech API for TTS/STT if possible, or any free tier.
Framework: Plain JS or React (your call). No heavy styling needed.
No backend required beyond reading a local CSV/JSON (or a public Google Sheet).
Deliverables
A shareable link (or zipped project) + short README showing how I swap in a new CSV/JSON deck.
Source with clear comments for the TTS/STT and answer-checking parts.
Acceptance Checklist
Mic button captures my voice on desktop & mobile.
App speaks questions; I answer by voice; it speaks feedback.
Wrong answers come back later; right ones move on.
End-of-session stats + CSV/Sheet export.
I can replace the deck myself in minutes.
Budget & Timing
Budget: $0–$100 (Negotiable).