Photos of tables in. Clean Excel rows out.
Snap a printed table, a stack of receipts, or a screenshot. Vecbase reads every cell, lines up the currencies and dates, drops the duplicates, and drops an editable Excel into your Drive.
A peek at what you get
Page 1 of 2·Reimbursement sheet
Page 2 of 2·Summary by category
From messy photo to clean cells
You snap the receipt at a weird angle in a coffee shop. The Agent reads it, normalizes it, and saves an .xlsx your finance team can actually open.
What you snap
IMG_2841.HEIC
Tilted · glare · handwritten total · coffee stain
14 receipt images
Agent · in sandbox
ocr_imageReducto · 14 images
run_pythonpandas · dedupe 3 rows
run_pythonnormalize ccy + dates
run_pythoncategorize + write xlsx
What lands in your Drive
expense-report-may-2026.xlsxcleaned
DateMerchantCat.USD
- 2026-05-02Singapore AirlinesFlight953.05
- 2026-05-02Marina Bay SandsHotel711.18
- 2026-05-04ANA · NH 802Flight446.10
- 2026-05-05Devsumi TokyoConf.247.76
- 2026-05-06Sukiyabashi JiroMeal353.38
- Total · 14 rowsUSD4,287.12
expense-report-may-2026.xlsx
24 KB
Currencies in USD · dates ISO · categories tagged
The cleanup rules you actually need
OCR is the easy part. The reason your spreadsheet usually still needs an hour of human cleanup is everything below — Agent applies all of it before the file lands.
Currency normalize
USD · EUR · JPY · SGD → reporting currency at receipt-date FX.
Date format
"3/5/26" · "5 May 2026" · "2026年5月5日" → all become 2026-05-05.
Dedup near-identical rows
3 removed · fuzzy match on (date + merchant + amount), with the removal log.
Categorize
taxi / flight / hotel / meal / conference / software / other — tagged per row.
Detect handwriting
2 amounts flagged for review — confidence < 90% gets a yellow tag, never silent-filled.
Sum by category
Totals row + a Summary tab so Finance sees the rollup without rebuilding it.
Every low-confidence cell is flagged · never silent-filled
How it works
Step 01
Upload your photos
JPG, PNG, HEIC, or a scanned PDF — all fine. Send a whole batch if you have one; Vecbase handles them at the same time.
Step 02
Tell Vecbase how clean you want it
One sentence is enough: "categorize and convert to USD", "stitch the long table back together", or "just lift the cells, don't touch anything else".
Step 03
Open the Excel
Vecbase reads each image carefully, lines up the numbers, and saves the workbook to your Drive. Usually 20–60 seconds for a batch of 10–15 photos.
Why Vecbase for this
Bad photos are fine
Crooked, glare on half the page, coffee stains, handwritten amounts — Vecbase straightens, cleans up, and reads what it can. Anything it isn't sure about gets flagged for you to check, not silently guessed.
Currencies and dates already lined up
USD, EUR, JPY, CNY all converted to your reporting currency using the rate on the day of the receipt. "3/5/26", "5 May 2026", "2026年5月5日" — all become one consistent date format. Categories assigned the same way every time.
Duplicates actually removed
When you photograph the same receipt twice or your long-table photos overlap, Vecbase spots the duplicates (same date + merchant + amount) and removes them — and tells you how many it took out.
Lands neatly in your Drive
The Excel sits next to the original photos in your Drive — share by link, hand it to Finance, or pipe it into your accounting tool. No more hunting for last quarter's receipts.
Frequently asked
JPG, PNG, HEIC (iPhone), and scanned PDFs (single or multi-page). Phone snapshots, screenshots of web tables, scanner outputs, photographed whiteboards all work. Camera EXIF orientation is read so sideways phone photos right themselves.
Get yours in under 90 seconds
Sign in, hand it over to the Agent — the finished file lands in your Drive.