How it works
From a photo to a filed, checked record
Five steps, and a person checks the middle one. That's the honest version of automated extraction: fast where the model is confident, and stopped for a human where it isn't.
-
Upload a photo or PDF
A single receipt, a multi-page statement, or a scanned invoice. Images and PDFs up to 25MB. The file goes straight to encrypted object storage, never to the disk of the server handling your request.
-
Google Gemini reads it
Your file is sent, through Deplyra's own routing gateway, to Google's Gemini vision model. It returns merchant, date, currency, subtotal, tax, total and every line item as structured data, not a guess dressed up as one: anything it can't read is marked as low or missing confidence rather than invented.
-
You review it
The extracted fields sit on a review screen next to the original image. You fix whatever the model got wrong. Nothing is treated as final until you've looked at it.
-
It's filed and searchable
Once reviewed, the document lives in your library, filterable by merchant, currency, date range and review state.
-
Match it to your statement
Line up the receipt against the matching line on a bank or card statement you've added, so the record is checked, not just stored.
The honest part
What we won't do
We won't pretend a guess is a fact
Optical reading of a receipt is never perfect, especially on a blurry photo or a faded thermal print. Where confidence is low, the field is marked low, not silently filled in.
We name where your file goes
This tool works by sending your file to a server and on to a model provider. The vendor's name, what's sent, how long it's kept, and how to delete it are on the data page, in plain sentences.