Skip to content

SherryNazifi/ReadyPresent

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ReadyPresent

AI-Powered Real-Time Presentation Confidence Coach

ReadyPresent is a full-stack AI system that analyzes live presentations and provides real-time feedback on delivery and speaking confidence. It combines computer vision and speech analysis to evaluate eye contact, pacing, and verbal habits, then generates a structured confidence score with actionable feedback.


Overview

Public speaking performance depends not only on content, but also on delivery. Speakers often lack objective feedback on:

  • Eye contact and engagement
  • Speaking pace
  • Filler word usage
  • Overall confidence

ReadyPresent addresses this gap by analyzing webcam and microphone input in real time and computing a live confidence score.


Features

  • Real-time eye alignment detection using facial landmarks
  • Live words-per-minute (WPM) calculation
  • Context-aware filler word detection
  • Weighted confidence scoring model
  • Session summary with improvement advice
  • Single-page application interface

How It Works

Eye Analysis

Webcam frames are sent to a FastAPI backend. MediaPipe Face Mesh extracts facial landmarks, and nose alignment relative to the screen center is used as a proxy for camera engagement. Exponential smoothing is applied to stabilize the score.

Speech Analysis

The Web Speech API captures transcript data in real time. The system calculates:

  • Words per minute (WPM)
  • Filler word frequency
  • Pause estimates

Filler detection includes both single-word and multi-word phrases, with contextual handling for ambiguous terms such as "like."

Confidence Model

The overall confidence score is computed using a weighted formula:

Confidence = 0.4 × Eye Score

  • 0.25 × WPM Score
  • 0.2 × Filler Score
  • 0.15 × Pause Score

All components are normalized to a 0–100 scale.


Architecture

Frontend:

  • HTML/CSS single-page application
  • Webcam frame capture
  • Speech recognition
  • Live metric updates

Backend:

  • FastAPI
  • OpenCV
  • MediaPipe
  • NumPy
  • REST endpoints:
    • /analyze
    • /speech_metrics
    • /confidence

Tech Stack

  • Python
  • FastAPI
  • Web Speech API
  • MediaPipe
  • OpenCV
  • NumPy
  • JavaScript
  • HTML / CSS

Running Locally

  1. Install dependencies:

pip install -r requirements.txt

  1. Start the server:

uvicorn main:app --reload

  1. Open in browser:

http://127.0.0.1:8000

Allow webcam and microphone access when prompted.


Future Improvements

  • Advanced AI-generated presentation feedback
  • Historical session tracking
  • Confidence trend visualization
  • Cloud deployment
  • PDF session report export

Authors

Shahrzad Nazifi, Joanna Liu, Jeng-Aun Chou, Chloe Le

UMass Amherst

About

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors