Skip to main content

Image Recognition

Take or select a photo, and AI automatically recognizes bill information.

Image Recognition

How to Use​

  1. Open the recording page
  2. Tap the camera icon
  3. Take a photo or select from gallery
  4. AI automatically recognizes the bill
  5. Confirm and save

Supported Scenarios​

  • Shopping Receipts - Supermarket, mall receipts
  • Food Delivery Orders - Uber Eats, DoorDash screenshots
  • Transfer Records - Payment app transfer screenshots
  • Invoices - Electronic or paper invoices

Recognized Content​

AI will attempt to identify:

  • Amount
  • Merchant name
  • Transaction time
  • Purchase items

Multiple Transactions per Image​

Since 3.2.3, the mobile AI flow supports detecting multiple transactions from a single image (previously only the first was kept). Bill screenshots, supermarket receipts, monthly Alipay / WeChat statements with many rows all yield a draft list of N transactions:

  • Confirm / tweak amount, category, account, tags and note per row
  • The original image is attached to every one of the N transactions, so the source is reachable from any of them
  • Tap "Skip" on any row to drop it — only the rows you want get saved

Tips​

  • Use clear, legible images
  • Include complete amount information
  • Avoid glare or shadows

Manual Adjustment​

AI recognition may not be 100% accurate. Please verify and adjust manually before saving.

Auto Tag​

When using image recording, the system automatically adds an "Image Recording" tag to the transaction for easy filtering and statistics.

Auto-Save Attachment​

The image used for recognition is automatically saved as a transaction attachment. You can view the original image in the transaction details.

Web Image Recording​

After signing in to BeeCount Cloud on the web, the desktop browser supports direct image paste:

  1. On any page, press ⌘K / Ctrl+K to open the command palette
  2. Press ⌘V / Ctrl+V to paste an image (a screenshot in the clipboard, or drag-drop)
  3. The default action switches to "AI bill (image)" → Enter
  4. A dialog shows N transaction drafts; edit category / account / tags / note per row
  5. Click "Save selected" to commit them in one batch

The server uses a vision LLM (Zhipu GLM-4V Flash etc., bind it in AI Config) to parse, and all N transactions share the original image as an attachment. On failure, the LLM raw output is shown for debugging.

Same as mobile, recognized transactions are auto-tagged with "AI" + "Image".