AI HTR TEXT EXTRACTOR

Multi-Provider AI Text Extractor

A Python utility that extracts text from images and PDF files using multiple AI providers (Google Gemini, OpenAI GPT, and Anthropic Claude).

Features

• Multi-Provider Support: Automatically detects and uses available AI providers based on your API keys
• Flexible File Support: Handles images (PNG, JPG, JPEG, WEBP, BMP, GIF, TIF, TIFF) and PDF files
• Customizable Prompts: Specify custom extraction prompts for different use cases
• Model Selection: Choose specific models for each provider
• Auto-Save: Automatically saves extracted text to timestamped files
• PDF Multi-Page: Processes all pages in PDF files with per-page extraction

Prerequisites

Required Python Packages

pip install python-dotenv pillow pdf2image google-generativeai openai anthropic

Additional Requirements

• For PDF processing: Install Poppler utilities
  • Windows: Download Poppler and set --poppler_path or add to PATH
  • macOS: brew install poppler
  • Linux: sudo apt-get install poppler-utils

API Keys Setup

Create a .env file in the same directory with your API keys:

GOOGLE_API_KEY=your_gemini_api_key_here
OPENAI_API_KEY=your_openai_api_key_here
ANTHROPIC_API_KEY=your_anthropic_api_key_here

You only need to provide keys for the providers you want to use.

Usage

Basic Usage

# Extract text using all available providers
python multi_extractor.py document.pdf

# Extract text from an image
python multi_extractor.py image.png

Advanced Usage

# Use a specific provider
python multi_extractor.py document.pdf --provider openai

# Use a specific model
python multi_extractor.py image.jpg --provider gemini --model gemini-1.5-flash-latest

# Custom extraction prompt
python multi_extractor.py receipt.png --prompt "Extract all prices and item names from this receipt"

# Windows with custom Poppler path
python multi_extractor.py document.pdf --poppler_path "C:\poppler\bin"

Command Line Arguments

• file_path: Path to the image or PDF file (required)
• --provider: Specify AI provider (gemini, openai, anthropic)
• --prompt: Custom extraction prompt (default: general text extraction)
• --model: Specific model name (when using single provider)
• --poppler_path: Path to Poppler bin directory (Windows only)

Default Models

• Gemini: gemini-2.0-flash
• OpenAI: gpt-4o
• Anthropic: claude-3-5-sonnet-20241022

Output

The tool creates timestamped text files for each provider:

• Format: {filename}_{provider}_extracted_{timestamp}.txt
• Example: document_gemini_extracted_20241225-143022.txt

Supported File Types

Images

• PNG, JPG/JPEG, WEBP, BMP, GIF, TIF/TIFF

Documents

• PDF (multi-page support)

Error Handling

• Gracefully handles missing API keys
• Continues processing with available providers if one fails
• Provides detailed error messages for troubleshooting
• Skips unsupported file types

Example Output

Auto-detected providers: gemini, openai. Using default models.

--- Processing with: GEMINI (Model: gemini-2.0-flash) ---
  Processing PDF page 1/3 with gemini...
  Processing PDF page 2/3 with gemini...
  Processing PDF page 3/3 with gemini...

--- Extracted Text (using GEMINI) ---
[Extracted text content here]

Extracted text from GEMINI saved to: document_gemini_extracted_20241225-143022.txt

--- Processing Complete ---

Tips

1. For receipts/invoices: Use custom prompts like "Extract all line items, prices, and totals"
2. For forms: Try "Extract all field labels and their values"
3. For handwritten text: Some providers perform better than others; try multiple
4. Large PDFs: Processing may take time as each page is analyzed separately

Troubleshooting

• Import errors: Ensure all required packages are installed
• PDF conversion fails: Check Poppler installation
• API errors: Verify API keys and rate limits
• No providers detected: Check .env file configuration

---

This software is freely distributed under the BSD 3-clause OSI license. Please see the LICENSE file for more information.