Turn bank statements into usable data without manual typing. The flow pulls a PDF from Google Drive, converts pages to images, transcribes each page to markdown with a vision model, and extracts only the deposit rows. It is built for finance teams that need fast, accurate data for reconciliation and reporting.

The run starts on click. A Google Drive node downloads the PDF. An HTTP request sends the file to a PDF service that returns a zip of images. The zip is extracted, images are listed and sorted by file name to keep page order, then resized to speed up AI processing. A language model converts each page image to markdown, the pages are combined, and an information extractor pulls structured deposit rows. This design handles scanned PDFs and messy layouts better than basic OCR.

Use a Google Drive account and a Google Gemini API key. For privacy, point the HTTP request to your own PDF service. Expect a clean table of deposits ready for review or export, saving hours of manual work. Ideal for monthly bookkeeping, cash posting, and audit support.

• Google Drive download pulls a specific statement file by ID for consistent inputs.
• PDF to image conversion returns one image per page inside a zip for reliable page handling.
• Zip extraction and list building create a clean array of page images for batching.
• Sorting by file name keeps page order so tables and dates stay aligned.
• Image resize reduces size to speed up model calls without losing needed detail.
• Vision model transcribes each page to markdown to preserve tables and columns.
• Aggregate step combines all page transcripts into one document for analysis.
• Information extractor filters only deposit rows into structured data ready for export.

• Reduce manual entry from hours to minutes by automating deposit capture
• Improve data accuracy by up to 90 percent versus hand typing
• Handle scanned statements that text extractors miss
• Process multi page PDFs in a single run with ordered results
• Connect Google Drive and a vision model for end to end processing

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with Google Drive and Google Gemini. See the Tools Required section above for links to create accounts with these services.
3. In the n8n credentials manager, open the Google Drive node. In the Credential to connect with dropdown, click Create new credential, then follow the on screen steps to sign in with Google Drive OAuth and grant file read access.
4. Open the Gemini nodes. In the Credential to connect with dropdown, click Create new credential, then follow the on screen steps to add your Google Gemini API key. Name the credential clearly, for example Gemini Prod.
5. In the Google Drive node, confirm the file mode is set to ID and paste the statement file ID you want to process.
6. Open the HTTP Request node. Check the URL points to your PDF to image service. For sensitive data, use your own hosted service URL. Leave multipart form settings as provided.
7. Run the workflow with the manual trigger. Confirm the HTTP step returns a zip and the Extract Zip step creates binary images.
8. Open the Code and Sort nodes. Confirm the output shows a list of images ordered by file name so pages are in sequence.
9. Check the Edit Image node is set to resize by percent. If you see timeouts, reduce the percent further to speed up processing.
10. Run again and view the Transcribe to Markdown node output. You should see readable markdown with tables where deposits appear.
11. Open the Information Extractor node result. Verify only deposit rows are returned. If the model misses rows, tweak the prompt or increase the model context if available.
12. Troubleshooting: If the Google Drive download fails, verify the file ID and access permissions. If the HTTP step errors, confirm the service URL and that the server is reachable. If pages are out of order, adjust the sort to use natural page numbering.