New Features

Direct PDF Upload Mode

The HyperFlow LLM PDF Transformer service now supports Direct Upload Mode, a streamlined processing option that sends PDFs directly to LLMs with native PDF support, bypassing the alternative page-by-page image rendering pipeline.

Key Benefits:

• Up to 3-5x faster processing compared to image rendering mode
• Reduced computational overhead with direct PDF processing
• Batch processing support for multiple related documents
• Optimized for modern LLMs like Claude, GPT-4, and Gemini that support direct upload and processing of PDF files.
• But see limitations below.

Note that most of these PDF-capable services use an internal page-to-image rendering and OCR, some with text extraction, that is similar to the page-by-page image mode available in this transformer service.

If you wish to use an LLM service that supports image OCR, but not PDF direct upload, use the Image Rendering Mode in this service.

How It Works:

• PDFs are sent directly to compatible LLMs without image conversion
• Multiple PDFs can be processed together in configurable batches
• LLMs analyze the entire document structure rather than individual page images
• Should maintain all document structure and formatting information

Enhanced Processing Mode Selection

Two Processing Modes Available:

1. Image Rendering Mode (Default)
   • Traditional page-by-page processing with high-resolution image rendering
   • Best for scanned documents, complex layouts, or when embedded image extraction is needed
   • Supports page range selection for partial document processing
2. Direct Upload Mode (New)
   • Modern streamlined processing for text-heavy documents
   • Faster processing with lower resource usage
   • Batch processing capabilities for multiple related documents
   • Note: See limitations section below for feature trade-offs