Image to Excel
A photo of a document becomes a spreadsheet you can sum — one credit each.
VisionParse turns a photographed document into an Excel workbook — the header fields on one sheet, every line item on another, with amounts still stored as numbers.
The reading is the AI extraction we measure elsewhere; what this page adds is the trip into cells. Five documents from our corpus went through the real pipeline and out as workbooks, then were opened with a spreadsheet library and compared cell by cell against the committed ground truth: 115 of 115 values landed in the right cell, with every quantity and amount still typed as a number rather than text.
- AI tier · 1 credit a document
- JPG · PNG · WEBP · PDF
- Photos auto-compressed · PDF to 25 pages
- Deleted within 1h
What this tool does, in numbers.
- ACCURACY
115 of 115 cells correct across five benchmark documents, checked by opening each workbook with a spreadsheet library.
- PRIVACY
The workbook is built in your browser from fields already on screen, so your document never travels a second time to become a file.
- LIMIT
Amounts and quantities stay numbers; a business card returns a contact file instead, because a card is not a grid.
- SPEED
A photographed receipt returns as a workbook in about two seconds, arithmetic checks included.
- COST
One AI credit a document, whether it carries three line items or thirty.
- ACCESS
Needs an account with AI credits: every paid plan includes them monthly, or a one-time pack adds them without a subscription.
- RETENTION
The uploaded photo is deleted within an hour, the extracted fields within 30 days, and nothing trains a model.
Where the 115 of 115 cells, five documents comes from.
Every figure on this page traces to a dated run on hardware we name. Nothing here is a vendor estimate.
Measured 2026-08-06
- Claim
- Extracted values survive the trip into spreadsheet cells — including staying numbers, which is the whole point of asking for Excel.
- Method
- AI extraction tier (gemini-3.5-flash-lite) into a browser-built .xlsx, one credit. Full methodology.
- Sample
- Three invoices, a delivery note and a goods received note from the benchmark corpus, each run through the production extraction pipeline, written to .xlsx by the same code the browser runs, reopened with the openpyxl library and compared value by value against the ground truth committed beside each sample.
- Measured
2026-08-06on Contabo Cloud VPS 8 — 8 vCPU AMD EPYC, 24 GB RAM, Ubuntu 24.04.4- Limitations
- This measures the fields the schemas describe, not arbitrary table capture: a page of unlabelled columns is a different problem, and the free table-to-CSV tool is the honest starting point for it. Handwritten documents and non-English layouts are unmeasured for spreadsheet output.
- Contact
- VisionParse@senithu.lk
What this tool gets wrong
Numbers arrive as numbers, not as text that looks like numbers. Every quantity and amount across the five documents came back typed as a number, so a total column sums the moment the file opens. A spreadsheet full of values you cannot add is a screenshot with extra steps, so this is checked explicitly rather than assumed.
This reads documents, not arbitrary tables. The workbook is built from a document schema — invoice, receipt, purchase order, delivery note, goods received note. Hand it a page of unlabelled columns and it will do its best with the wrong instrument. For a plain bordered table, the free table-to-CSV tool is the honest place to start, and it publishes its own column-detection limit.
One credit covers 25 pages; the AI engine reads up to 35 MB a document. Both ceilings are ours and both are stated rather than discovered: a longer PDF should be split, and a document past 35 MB is refused with our number instead of a provider error. Photographs are compressed in the browser first, so a phone picture never runs into either.
Three steps, no account.
Photograph the document
Straight from your phone — JPG, PNG, WebP or PDF. Large photos are compressed in the browser before they are sent.
Say what it is
Invoice, receipt, purchase order, delivery note or goods received note. The choice decides which fields come back.
Download the workbook
Two sheets: the document’s own fields, and the line items as a grid you can sort and sum. JSON is there too if you would rather have it.
What this tool accepts.
| Property | Free tool |
|---|---|
| Engine | AI extraction tier, 1 credit per document |
| Measured result | 115 of 115 cells across five documents (2026-08-06) |
| Output | Excel .xlsx with a Document sheet and a Line items sheet, plus JSON |
| Where the file is made | In your browser — the fields are not sent anywhere to become a spreadsheet |
| Number handling | Quantities and amounts written as numbers, not text |
| Input formats | JPG, PNG, WEBP, PDF — up to 25 pages a credit, 35 MB a document |
| Access | Account with AI credits — any paid plan, or a one-time pack |
| File retention | Originals deleted within 1 hour; results within 30 days |
When to use the API instead.
This page converts one file now. To convert files from your own software, the same engine runs over HTTP from Starter upward.
# same engine, from your code curl -X POST https://api.visionparse.app/v1/ocr \ -H "x-api-key: $VISIONPARSE_KEY" \ -F "file=@image.jpg" → 202 { "jobId": "c0df8e2c…" }
Use the API when
You convert more than a few files, convert on a schedule, or need results inside another system. Batch, webhooks and bounding boxes are API-only.
Frequently asked
Will the totals actually add up in Excel?
Yes, and it is checked rather than hoped: every quantity and amount in the benchmark came back stored as a number, so a SUM over the column works the moment the file opens. Values stored as text are the usual failure of converted spreadsheets, which is why we measure for it.
Does my document get uploaded twice?
No. The document is read once, and the workbook is then assembled in your browser from the fields already on your screen. Nothing makes a second journey to a server to come back as a file.
Can it convert any table in a picture?
It reads documents it has schemas for — invoices, receipts, purchase orders, delivery notes, goods received notes — and puts their line items in a grid. A page of unlabelled columns is a different job; start with the free table-to-CSV tool, which states its own column-detection limit honestly.
What about a photo taken at an angle?
Measured on the receipt scanner: the same slip faded, creased, photographed at an angle under uneven light and shrunk to a 600-pixel crop returned all fourteen fields correctly every time. The reading is the same engine here.
Why is this not free?
Turning a picture into labelled fields is the AI engine’s job and it costs money per call — free OCR gives you a wall of text, not a spreadsheet. So it runs on credits: included monthly with every paid plan, or bought once in a pack.