Transcribe Bank Statements to Markdown with Gemini Vision AI

This n8n workflow provides an innovative method for extracting data from bank statement PDFs using multimodal LLMs like Gemini Vision AI, offering a superior alternative to conventional OCR tools. It handles complex layouts, multi-column tables, and non-standard PDF formats with high accuracy, converting documents directly into structured Markdown without preprocessing hassles.

The process begins by importing a PDF bank statement from Google Drive. The PDF is then converted to a series of JPG images (one per page) using the self-hostable Stirling PDF tool, which returns a ZIP file. n8n's Decompress node extracts the images, and a Sort node ensures proper page ordering.

Subsequently, Gemini Vision AI processes the images to transcribe and format the content into clean Markdown, capturing transactions, balances, and tables precisely. Benefits include reduced complexity, lower costs compared to premium OCR services, and customizable output schemas. Ideal for financial data analysis, auditing, or integrating bank data into reports and dashboards.

Use cases span personal finance tracking, business accounting automation, compliance reporting, and AI-driven document processing pipelines. It's particularly valuable for handling messy PDFs where OCR fails, saving hours of manual cleanup.