VISIONAID
An AI-powered Document Reader for Visually Impaired and Illiterate Users
AD-1 Project Report
Submitted by:
Nandini Behera
[Link] (CSE - AI & ML), 3rd Year, 1st Semester
Malla Reddy College of Engineering and Technology (MRCET)
Under the Guidance of:
[Guide Name]
Department of Computer Science and Engineering (AI & ML)
MRCET, Hyderabad
Academic Year 2025-2026
ABSTRACT
VisionAid is an AI-powered desktop application designed to assist visually impaired and
illiterate individuals in understanding printed or handwritten text documents.
The system extracts text content from scanned documents or captured images using Optical
Character Recognition (OCR) powered by Tesseract, analyzes it using Natural Language
Processing (NLP) via spaCy, and converts it into natural-sounding speech through Google
Text-to-Speech (gTTS).
The project provides an easy-to-use Streamlit-based interface, allowing users to upload or
capture images of printed material and hear the information read aloud instantly.
By leveraging open-source AI tools, VisionAid demonstrates how assistive technology can
empower individuals to access essential information independently.
CHAPTER 1: INTRODUCTION
VisionAid bridges the gap between digital information accessibility and the visually
impaired community. Many people still rely on printed documents, but individuals with
visual impairment or illiteracy face challenges in reading and understanding these
documents.
This project aims to solve this by building a simple AI-driven system that reads documents
aloud using OCR and TTS technologies.
1.1 OBJECTIVES
• To extract text from printed or handwritten documents using Tesseract OCR.
• To process extracted text for key information using NLP (spaCy).
• To convert processed text to speech using gTTS.
• To design a user-friendly Streamlit interface for accessibility.
• To provide a cost-effective assistive solution for visually impaired users.
1.2 SCOPE
The scope of VisionAid is limited to document-based text extraction and reading. It focuses
on English and regional language documents where OCR models are available.
Future versions can extend support for more languages, offline TTS, and Android-based
deployment.
CHAPTER 2: LITERATURE SURVEY
Existing solutions like Seeing AI (Microsoft) and Envision AI provide assistive tools for the
visually impaired, but they are often dependent on paid APIs or require smartphones with
internet access.
VisionAid, in contrast, focuses on open-source implementation using Python libraries like
Tesseract, spaCy, and gTTS for affordability and offline flexibility.
Research in assistive AI highlights the importance of accessibility, real-time feedback, and
minimal user interaction—all of which are key design considerations in VisionAid.
CHAPTER 3: SYSTEM ANALYSIS
3.1 EXISTING SYSTEM
Existing systems are often mobile-based, expensive, and dependent on cloud APIs. Users
must rely on third-party servers, risking privacy and requiring stable internet connections.
3.2 PROPOSED SYSTEM
VisionAid introduces a lightweight desktop application that uses local AI tools for text
extraction and speech output. It ensures privacy, low cost, and easy accessibility for all age
groups.
3.3 SOFTWARE REQUIREMENTS
• Python 3.10+
• Streamlit
• Tesseract OCR
• spaCy
• gTTS
• PIL (Pillow)
• pytesseract
3.4 HARDWARE REQUIREMENTS
• Processor: Intel i3 or higher
• RAM: Minimum 4GB
• Storage: 500MB free space
• Operating System: Windows 10 or Linux
CHAPTER 4: SYSTEM DESIGN
The VisionAid architecture consists of three main components:
1. OCR Module – Extracts text from document images.
2. NLP Module – Identifies and summarizes key entities.
3. TTS Module – Converts text to speech output.
The user interacts through the Streamlit interface, uploading an image or scanned
document. The system processes it sequentially through these modules and outputs audible
information.
CHAPTER 5: IMPLEMENTATION
VisionAid is implemented using Python and Streamlit for frontend UI and integrates OCR,
NLP, and TTS functionalities. The OCR engine (Tesseract) extracts text, spaCy performs NLP,
and gTTS converts text into speech.
Code Snippet (Simplified)
import streamlit as st
from PIL import Image
import pytesseract
from gtts import gTTS
import io, spacy
[Link]("VisionAid - AI Document Reader")
uploaded = st.file_uploader("Upload Image", type=["jpg","png"])
if uploaded:
image = [Link](uploaded)
[Link](image, caption="Uploaded Image")
text = pytesseract.image_to_string(image)
[Link]("Extracted Text:", text)
nlp = [Link]("en_core_web_sm")
doc = nlp(text)
if [Link]("Play Audio"):
tts = gTTS(text=text, lang="en")
mp3 = [Link]()
tts.write_to_fp(mp3)
[Link]([Link](), format="audio/mp3")
CHAPTER 6: RESULTS AND DISCUSSION
VisionAid successfully reads printed text from images and converts it into natural-sounding
audio output.
Testing was conducted on different document types, including typed letters and forms,
achieving an OCR accuracy of 90–95% for clear images.
The Streamlit interface ensures a simple, one-click interaction flow.
CHAPTER 7: CONCLUSION AND FUTURE SCOPE
VisionAid demonstrates the potential of AI in creating accessible assistive tools for
individuals with disabilities.
By combining OCR, NLP, and TTS, the application enables users to access textual content
effortlessly.
Future enhancements include multilingual support, offline text-to-speech (via pyttsx3), and
Android-based deployment for real-world usability.
REFERENCES
1. Tesseract OCR Documentation – [Link]
2. spaCy NLP Library – [Link]
3. Streamlit Framework – [Link]
4. Google Text-to-Speech (gTTS) – [Link]
5. Research Papers on AI for Accessibility (IEEE, Springer)