Paper Receipts: The Blind Spot of Personal Finance
Even in an increasingly digital world, paper bills remain ubiquitous: dinners at cafes, pharmacy bills, fuel station printed slips, and local grocery receipts. Manually typing 15 items or calculating tax totals after a long dinner is painful.
How Gemini 2.5 Flash Vision Powers Real-time OCR
When you send an image to the XpenseDiary WhatsApp bot, the image is passed to Google's multimodal Gemini 2.5 Flash Vision model. The model understands spatial hierarchy on thermal printouts and handwritten receipts:
- Merchant Name: Identifies the store or restaurant header.
- Date Detection: Finds printed transaction timestamps.
- Grand Total: Distinguishes sub-totals, CGST/SGST taxes, and discount lines to extract the true final payable amount.
- Category Mapping: Maps pharmacy bills to Health & Medical, restaurant receipts to Food & Dining, etc.
Sub-2 Second Turnaround on Chat
Within 1.5 to 2 seconds of sending a photo on WhatsApp, you receive a verified confirmation breakdown with merchant name, amount in rupees, and category tagging.