DEV Community: Tori Cj

How to Extract a Table From an Image Into Excel (5 Methods That Actually Work)

Tori Cj — Thu, 02 Jul 2026 16:19:09 +0000

You've got a table — in a screenshot, a chart, a scanned PDF, a slide someone exported as a PNG — and you need it in Excel. You try to select it. Nothing highlights. Copy-paste gives you nothing, because to your computer that table isn't data: it's a picture of data.

If you've ever found yourself re-typing numbers off your screen cell by cell, this post is for you. Here are five methods to extract a table from an image into Excel, with the honest pros and cons of each.

Method 1 — Excel's built-in "Data from Picture" (mobile)

Microsoft quietly shipped this in the Excel mobile app and it's genuinely useful when it works.

Steps:

Open the Excel app on your phone (iOS/Android).
Tap Insert → Data from Picture.
Photograph your screen (or pick the image from your gallery).
Review Excel's guesses, fix flagged cells, tap Insert.

Pros: free, built into Excel, no extra tools.
Cons: mobile-only — there's no desktop equivalent, so you end up photographing your own monitor, which is as silly as it sounds. It also struggles with dense tables, merged cells, and low-contrast screenshots, and every mis-read cell needs manual review on a phone keyboard.

Best for: paper documents you can photograph, occasional one-offs.

Method 2 — The Google Drive OCR trick

Google Docs has hidden OCR: upload an image to Drive, right-click it, Open with → Google Docs, and Google appends the recognized text under the image. Copy that text into Sheets and use Data → Split text to columns.

Pros: free, works on desktop, surprisingly good character recognition.
Cons: it outputs a wall of text, not a table. Column boundaries are lost, so you spend your saved time splitting text back into columns and re-aligning rows. Multi-line cells and numbers with thousands separators get mangled.

Best for: paragraphs of scanned text. Weak for real tables.

Method 3 — A drag-select browser extension (the quick one-off king)

This is the method I use daily, mostly because I got tired of the other four and built it: Snap to Data, a Chrome extension.

Steps:

Press Alt+S on any page (or click the extension icon).
Drag a box over the table — it can be an image, a chart, a scanned page, anything visible.
Get back plain text, markdown, or a CSV, and paste it straight into Excel or Google Sheets.

That's the whole workflow: drag, copy, paste. It reads the pixels, so it doesn't matter whether the table is real HTML, an image, a canvas chart, or a screenshot in a PDF viewer.

Pros: works on desktop right where the table is (your browser), keeps rows and columns aligned, outputs paste-ready CSV, handles images and scans — exactly where the copy-paste approach dies.
Cons: Chrome-family browsers only; the free tier covers 5 snaps a month (Pro is $9.99/mo if you live in spreadsheets).

Best for: the "I just need THIS table in Excel, now" moment.

Method 4 — Dedicated OCR services (for bulk jobs)

If you have fifty scanned PDFs of tables, one-off tools stop making sense.

Tabula (free, open source) — excellent for native PDF tables; it does not OCR scanned images.
Online OCR converters (Smallpdf, Adobe's online tools, OCR.space, etc.) — upload the file, download an .xlsx. Quality varies wildly with scan quality.
Cloud OCR APIs (Google Vision, AWS Textract) — the industrial option if you can script; Textract in particular returns real table structure.

Pros: built for volume, some preserve table structure well.
Cons: setup friction, per-page pricing on the good ones, privacy questions when uploading sensitive documents to random converter sites.

Best for: batch processing, recurring pipelines.

Method 5 — Re-typing it by hand

The fallback nobody wants. For a 5×4 table it's honestly fine — two minutes and you're done, no tools. For anything bigger it's slow and, worse, error-prone: transposed digits in a table you'll base decisions on are far more expensive than any tool.

If you catch yourself doing this weekly, you're the exact person methods 3 and 4 exist for.

Comparison

Method	Works on images?	Keeps table structure?	Desktop?	Free?	Best for
Excel "Data from Picture"	✅	Mostly	❌ mobile only	✅	Photographed paper docs
Google Drive OCR	✅	❌ loses columns	✅	✅	Scanned prose, not tables
Snap to Data extension	✅	✅	✅	5/mo free	Quick one-off tables in the browser
OCR services / Textract	✅	Varies–✅	✅	Partly	Bulk / recurring jobs
Manual retype	✅	✅	✅	✅	Tiny tables only

The bottom line

To extract a table from an image into Excel: for a quick one-off in your browser, a drag-select tool like Snap to Data gets you a clean CSV in seconds; for bulk scanned documents, use a proper OCR service; and if you're on your phone with a paper document, Excel's own Data from Picture is right there. Whatever you pick — stop re-typing data off your screen. Life's too short.

Built something similar or found a better workflow? I'd genuinely like to hear it — I built Snap to Data because of this exact headache and I'm still improving the messy-table handling.

How to Extract Data from Receipts with an API (Node.js & Python)

Tori Cj — Thu, 25 Jun 2026 22:10:46 +0000

If you've ever tried to pull structured data out of a receipt or invoice, you know it's deceptively hard. A photo of a receipt is just pixels. Plain OCR turns it into a wall of text — but you still don't have fields. You want the merchant, the date, each line item with its price, the tax, and the total, in a clean shape your code can use.

This guide shows how to go from a receipt image (or PDF) to structured JSON in a single API call — with copy-paste examples in cURL, Node.js, and Python.

Why not just use raw OCR?

Traditional OCR (Tesseract and friends) gives you text, not data. You'd still have to write and maintain brittle parsing logic to find the total, separate line items, handle different receipt layouts, currencies, and languages. That parser breaks constantly as new formats show up.

A purpose-built receipt OCR API skips all of that: it returns a fixed, predictable JSON schema, so you never write parsing code.

The approach

We'll use the Receipt & Invoice OCR API on RapidAPI. You send an image URL, a base64 image, or a base64 PDF, and you get back structured JSON. There's a free tier (50 calls/month) so you can follow along at no cost.

Step 1 — Get an API key

Go to the API listing on RapidAPI.
Subscribe to the free Basic plan.
Copy your X-RapidAPI-Key.

Step 2 — Your first call (cURL)

curl -X POST 'https://receipt-extraction-api.p.rapidapi.com/v1/extract' \
  -H 'content-type: application/json' \
  -H 'X-RapidAPI-Key: YOUR_KEY' \
  -H 'X-RapidAPI-Host: receipt-extraction-api.p.rapidapi.com' \
  -d '{"image_url": "https://receipt-extraction-api.onrender.com/sample.png"}'

Step 3 — Node.js

const res = await fetch("https://receipt-extraction-api.p.rapidapi.com/v1/extract", {
  method: "POST",
  headers: {
    "content-type": "application/json",
    "X-RapidAPI-Key": process.env.RAPIDAPI_KEY,
    "X-RapidAPI-Host": "receipt-extraction-api.p.rapidapi.com",
  },
  body: JSON.stringify({
    image_url: "https://receipt-extraction-api.onrender.com/sample.png",
  }),
});

const { data } = await res.json();
console.log(data.merchant.name, data.total); // "BLUE BOTTLE COFFEE" 21.26

Step 4 — Python

import os, requests

resp = requests.post(
    "https://receipt-extraction-api.p.rapidapi.com/v1/extract",
    headers={
        "content-type": "application/json",
        "X-RapidAPI-Key": os.environ["RAPIDAPI_KEY"],
        "X-RapidAPI-Host": "receipt-extraction-api.p.rapidapi.com",
    },
    json={"image_url": "https://receipt-extraction-api.onrender.com/sample.png"},
)

data = resp.json()["data"]
print(data["merchant"]["name"], data["total"])  # BLUE BOTTLE COFFEE 21.26

The response schema

Every response has the same shape — every field is always present, and missing values are null, so you never have to write defensive checks:

{
  "document_type": "receipt",
  "merchant": { "name": "...", "address": "...", "phone": "..." },
  "date": "2026-06-25", "time": "09:14", "currency": "USD",
  "line_items": [
    { "description": "Cappuccino", "quantity": 2, "unit_price": 3.50, "total": 7.00 }
  ],
  "subtotal": 16.75, "tax": 1.51, "tip": 3.00, "total": 21.26,
  "payment": { "method": "visa", "card_last4": "4242" }
}

Working with PDFs and uploaded files

Instead of image_url, you can send a base64 image (with media_type) or a base64 PDF:

{ "pdf_base64": "JVBERi0xLjQ..." }

{ "image_base64": "iVBORw0KGgo...", "media_type": "image/png" }

This is handy when the file lives on the user's device and isn't publicly reachable by URL.

Where this is useful

Expense and reimbursement apps — auto-fill expense reports from a photo.
Accounting and bookkeeping tools — turn receipts into ledger entries.
Personal finance apps — categorize spending from receipts.
Spend analytics — structured line-item data for reporting.

Wrapping up

Extracting data from receipts used to mean training a model or babysitting an OCR-and-regex pipeline. With a dedicated receipt OCR API you make one request and get clean, typed JSON you can use immediately.

You can try it free here: Receipt & Invoice OCR API — and there's a live demo (image → JSON) at receipt-extraction-api.onrender.com.

If you build something with it, I'd love to hear about it — and what receipt edge cases I should handle next.

I built a Receipt/Invoice JSON API (here's how, and how to use it)

Tori Cj — Thu, 25 Jun 2026 20:54:33 +0000

Every expense tracker, accounting tool, and bookkeeping app eventually hits the same wall: turning a photo of a receipt into usable data. Raw OCR gives you a blob of text, not structured fields — and building a reliable parser on top is a rabbit hole.

So I built a small API that does one thing well: receipt or invoice in (image or PDF), clean structured JSON out.

What you get back

Every field, every time (null if missing) — so you can skip defensive parsing:

{
  "document_type": "receipt",
  "merchant": { "name": "...", "address": "...", "phone": "..." },
  "date": "2026-06-25", "time": "09:14", "currency": "USD",
  "line_items": [
    { "description": "Cappuccino", "quantity": 2, "unit_price": 3.50, "total": 7.00 }
  ],
  "subtotal": 16.75, "tax": 1.51, "tip": 3.00, "total": 21.26,
  "payment": { "method": "visa", "card_last4": "4242" }
}


curl -X POST 'https://receipt-extraction-api.p.rapidapi.com/v1/extract' \
  -H 'content-type: application/json' \
  -H 'X-RapidAPI-Key: YOUR_KEY' \
  -H 'X-RapidAPI-Host: receipt-extraction-api.p.rapidapi.com' \
  -d '{"image_url": "https://example.com/receipt.jpg"}'

No problem — that's **totally normal** and not your fault. HN restricts "Show HN" for brand-new accounts (anti-spam). It's **not blocked forever** — once your account has a little history/karma (comment thoughtfully on a few threads over a week or two), Show HN opens up. **Park HN for later.** Your post did *not* go out, so nothing to undo.

Let's keep rolling. 👇

---

## 🟢 POST #3 — dev.to (article)

**Why:** no restrictions for new accounts, it's evergreen, and it ranks in Google — a long-tail traffic source that keeps working.

**Go to:** [dev.to/new](https://dev.to/new) (sign in with GitHub — instant)

**Title:**

I built a Receipt/Invoice → JSON API (here's how, and how to use it)


**Tags** (type these in the tags field):

api, showdev, ai, webdev


**Body** (paste this whole block — it's Markdown, which dev.to uses natively):

So I built a small API that does one thing well: receipt or invoice in (image or PDF), clean structured JSON out.

What you get back

Every field, every time (null if missing) — so you can skip defensive parsing:

{
  "document_type": "receipt",
  "merchant": { "name": "...", "address": "...", "phone": "..." },
  "date": "2026-06-25", "time": "09:14", "currency": "USD",
  "line_items": [
    { "description": "Cappuccino", "quantity": 2, "unit_price": 3.50, "total": 7.00 }
  ],
  "subtotal": 16.75, "tax": 1.51, "tip": 3.00, "total": 21.26,
  "payment": { "method": "visa", "card_last4": "4242" }
}

Using it

curl -X POST 'https://receipt-extraction-api.p.rapidapi.com/v1/extract' \
  -H 'content-type: application/json' \
  -H 'X-RapidAPI-Key: YOUR_KEY' \
  -H 'X-RapidAPI-Host: receipt-extraction-api.p.rapidapi.com' \
  -d '{"image_url": "https://example.com/receipt.jpg"}'

It also accepts image_base64 (+ media_type) or pdf_base64.

Why a fixed schema

The annoying part of receipt parsing isn't the OCR — it's getting consistent, typed output you can trust. This API always returns the same shape, with null for anything it can't find, so your code never has to guess.

Try it

Live demo (watch an image turn into JSON): https://receipt-extraction-api.onrender.com

Free tier (50 calls/mo) + paid plans: https://rapidapi.com/DenisCj1/api/receipt-extraction-api

I'd love feedback — especially edge cases: handwritten receipts, multi-page invoices, non-USD currencies. What would make this genuinely useful in your stack?