PDF Visual Translator

An advanced AI-powered document translation tool that leverages Google Gemini models to perform visual-aware translation. Unlike traditional OCR tools, this project preserves the exact layout, font styles, colors, and background of the original PDF while overlaying perfectly localized text.

---

🚀 Key Features

1. Dual-Stage AI Translation Engine

The heart of the application is a sophisticated two-step process:

• Phase 1 (Extraction & Reasoning): Uses Gemini Flash to perform high-precision OCR and contextual translation, respecting custom glossaries and linguistic rules.
• Phase 2 (Visual Reconstruction): Uses Gemini Pro (Visual Specialist) to erase original text and "redraw" the translated segments using pixel-perfect alignment and style matching.

2. Professional Quality Evaluation

• Automated Audit: Every translation is automatically evaluated by a secondary AI agent.
• Detailed Metrics: Scores across Accuracy, Fluency, Consistency, Terminology, Completeness, and Format Preservation.
• Optimization Suggestions: Context-aware feedback provided for every page to ensure professional-grade output.

3. Advanced Linguistic Controls

• Bilingual Preservation: Smartly keeps English headers, technical terms, and product names (e.g., "CONTENTS 目录" becomes "CONTENTS 目錄").
• Glossary Enforcement: Upload or paste custom term mappings to ensure industry-specific accuracy.
• Hong Kong Localization: Dedicated support for Traditional Chinese (Hong Kong) standards.

4. Enterprise Export Options

• Translated PDF: High-fidelity document mirroring the original layout.
• Comparison View: Side-by-side (Original vs. Translated) PDF for verification.
• Evaluation Report: Detailed JPEG/PDF reports containing AI scores and suggestions for each page.

---

🛠 Technology Stack

• Frontend: React + Vite + Tailwind CSS
• AI Backend: Google Gemini API (Pro & Flash)
• PDF Processing: PDF.js (Parsing) & jsPDF (Generating)
• Icons: Lucide React

---

📦 Getting Started

Prerequisites

• Node.js (LTS version recommended)
• Google AI Studio API Key (Get it here)

Installation

1. Clone the repository:

   git clone <repo-url>
   cd pdf-visual-translator

2. Install dependencies:

   npm install

3. Set up environment: Create a .env.local file or set the environment variable:

   GEMINI_API_KEY=your_api_key_here

Running Locally

npm run dev

Open http://localhost:3000 in your browser.

---

🔄 Translation Workflow

graph TD
    Start([Upload PDF/Image]) --> PDF{Is PDF?}
    PDF -->|Yes| Conv[PDF to Image Conversion]
    PDF -->|No| Process[Start Translation Engine]
    Conv --> Process
    
    subgraph Engine [Dual-Stage Engine]
        Process --> Step1[<b>Step 1: Reasoning</b><br/>Gemini Flash performs OCR,<br/>Terminology application,<br/>and Contextual Translation]
        Step1 --> Step2[<b>Step 2: Visual Editing</b><br/>Gemini Pro erases original text,<br/>redraws translated text,<br/>and matches styles]
    end
    
    Step2 --> Eval[<b>AI Evaluation</b><br/>Quality audit & Scoring]
    Eval --> Export([Export: PDF / Comparison / Report])

Loading

---

🧬 Project Architecture

graph TD
    UI[React Interface] --> PS[PDF Service]
    UI --> GS[Gemini Service]
    PS -->|Convert to Image| UI
    GS -->|Step 1: Reasoning| Flash[Gemini Flash]
    GS -->|Step 2: Edit| Pro[Gemini Pro Visual]
    UI -->|Export| jsPDF[jsPDF Engine]

Loading

---

🛡 License

This project is licensed under the MIT License.