██╗ ██╗ ██████╗ ██╗ ██╗██╗ ██╗ ██╗ ██║ ██║██╔═══██╗╚██╗██╔╝██║ ╚██╗ ██╔╝ ██║ ██║██║ ██║ ╚███╔╝ ██║ ╚████╔╝ ╚██╗ ██╔╝██║ ██║ ██╔██╗ ██║ ╚██╔╝ ╚████╔╝ ╚██████╔╝██╔╝ ██╗███████╗██║ ╚═══╝ ╚═════╝ ╚═╝ ╚═╝╚══════╝╚═╝
AI-powered platform to practice and improve your presentation skills
VoxLy helps you become a better speaker by turning every practice session into structured, actionable feedback. Record or upload a presentation video and the platform automatically transcribes your speech, analyzes your body language, and generates AI-driven feedback with a per-session score, so you can track your progress over time instead of guessing what to improve.
Practicing presentations alone is hard: you can't watch yourself speak and judge your posture, gestures, and pace at the same time. VoxLy closes that gap by combining speech-to-text, computer-vision gesture detection, and an LLM coach into a single, simple workflow: upload a video, get a report.
- Video upload & processing: record or upload presentation videos (MP4, up to 100MB).
- AI transcription: automatic speech-to-text powered by OpenAI Whisper, with multi-language support (English/Spanish).
- Body-language analysis: computer-vision pose and hand tracking (MediaPipe) detects up to 4 people per video and flags issues like crossed arms, hands in pockets, fidgeting, face/hair touching, rigid arms, slouching, and hands behind the back.
- Automated scoring: a rule-based engine deducts points per detected gesture based on severity, duration, and frequency, producing a clear per-person score.
- Annotated video output: renders the original video back with pose/gesture overlays for visual, timestamped feedback.
- AI-generated feedback: GPT-based evaluation summarizes strengths, weaknesses, and suggestions for each session.
- Session management & progress tracking: create, organize, and revisit practice sessions, and follow your improvement across time.
- Authentication & personalization: secure JWT-based login, account settings, and password reset.
The backend is the core API: it handles authentication, sessions, and orchestrates calls to the AI service, which performs speech transcription and computer-vision gesture analysis and returns structured results (transcript, gesture events, score, annotated video). Video files are stored in Cloudflare R2, structured data in PostgreSQL, and the frontend talks to the backend over REST.
voxly/
├── backend/ # Java Spring Boot API (auth, sessions, evaluation, feedback, progress)
├── frontend/ # React + TypeScript SPA (dashboard, sessions, recorder, progress)
├── ai/ # Python/FastAPI service - transcription & body-language analysis
└── scripts/ # Dev & setup helpers
The SPA users interact with: dashboard, session list and detail, new-session recorder, progress charts, and account/personalization.
- Stack: React 19, TypeScript, Vite, Tailwind CSS, React Router.
The main API. Handles identity (JWT auth, OAuth2), session lifecycle, file uploads, evaluation/feedback orchestration, and progress aggregation.
- Stack: Java 21, Spring Boot, Spring Security, Hibernate/JPA, Flyway, PostgreSQL, AWS S3 SDK (Cloudflare R2), Quartz.
Processes uploaded videos: extracts audio for Whisper transcription, runs MediaPipe pose/hand tracking for gesture detection, scores the presentation, and can render an annotated output video.
- Stack: Python, FastAPI, OpenCV, MediaPipe, OpenAI API (Whisper + GPT).
| Layer | Technologies |
|---|---|
| Frontend | React 19 · TypeScript · Vite · Tailwind CSS |
| Backend | Java 21 · Spring Boot · Spring Security · Hibernate/JPA · Flyway |
| AI service | Python · FastAPI · OpenCV · MediaPipe · OpenAI (Whisper + GPT) |
| Data & storage | PostgreSQL (Neon) · Cloudflare R2 |
| Infra | Docker · Gradle |
- Java 21+
- Node.js 18+ and pnpm/npm
- Python 3.10+ and FFmpeg
- Docker (for local infrastructure, optional)
git clone https://github.com/Dexaroz/VoxLy
cd VoxLy
cp .env.example .env
# fill in .env with your credentials (DB, OpenAI, Cloudflare R2, JWT secret...)Run each service (see make help / Makefile targets):
make backend # Spring Boot API
make ai # FastAPI analysis service
make frontend # Vite dev serverSee ENV_SETUP.md for detailed environment configuration.
VoxLy: practice, get feedback, present better.