You've got a table — a screenshot from a PDF report, a photo of a printed spreadsheet, a chart from a research paper — and you need it back in Excel, not as a static image. Re-typing rows and columns by hand is the worst.
This guide shows the fastest way to extract tables from images into editable Excel, CSV, or Google Sheets — without paying for enterprise software.
The Problem with Most OCR Tools and Tables
Plain OCR turns tables into wall-of-text. You lose row/column boundaries, headers blend into data, and the result is unusable in Excel.
Worse, some AI tools default to LaTeX output for tables. LaTeX is great for academic papers, but useless if you just want to paste data into a spreadsheet. (Search query "提取图片文字 不要用 latex 编译" — "extract image text but not LaTeX" — shows this is a real pain point.)
The solution: AI OCR that's table-aware and outputs plain text or markdown tables.
Method 1: img2txt.app (Recommended for Tables)
img2txt.app uses Gemini's vision model, which understands tabular structure natively.
Steps:
- Screenshot or photograph the table — keep all rows visible in one image
- Upload to img2txt.app/image-to-text
- The output preserves rows and columns as a markdown table (pipes
|) or tab-separated text - Copy the result
- Paste into Excel — it auto-splits into rows/columns
Tips for best table OCR:
- Crop tightly to the table edges — extra margin can confuse the model
- High-contrast screenshots (light background, dark text) work best
- For very wide tables, capture at the highest resolution your screen supports
No LaTeX, no PDF parsing libraries, no signup — drag, drop, copy, paste.
Method 2: Microsoft Excel "Data from Picture"
Excel has a built-in feature (mobile + desktop) that reads tables from photos.
On Excel mobile (iOS / Android):
- New blank sheet → tap Insert → Data from Picture
- Take a photo or pick from gallery
- Review and accept the parsed data
On Excel desktop (Microsoft 365):
- Data tab → From Picture → Picture from File or Picture from Clipboard
- Excel parses and asks you to review each row
Strengths: Goes straight into Excel. Weaknesses: Mobile is more reliable than desktop; struggles with merged cells and rotated tables.
Method 3: Google Sheets via Google Drive
Less direct, but works:
- Upload the image to Google Drive
- Right-click → Open with → Google Docs
- Docs runs OCR and extracts the text
- Manually paste into Sheets — alignment may need fixing
Strengths: Free, unlimited. Weaknesses: Table structure is often lost; only worth it for simple tables.
Method 4: Specialized Tools (Nanonets, Rossum)
For enterprise volumes of invoices, receipts, and forms, dedicated table-OCR services like Nanonets and Rossum offer training on your specific document layouts.
Strengths: Best accuracy for repetitive document types. Weaknesses: Paid, complex setup. Overkill unless processing thousands of similar documents per month.
Comparison
| Tool | Speed | Output Format | Free Quota | Best For |
|---|---|---|---|---|
| img2txt.app | Seconds | Markdown / TSV | 3–6/day | Mixed content, screenshots |
| Excel "Data from Picture" | Minutes (with review) | Native Excel | Free with M365 | Phone-captured tables |
| Google Docs OCR | Multi-step | Plain text | Free, unlimited | Quick & dirty |
| Nanonets / Rossum | Trained pipelines | Custom | Paid | High-volume invoices |
Tips to Avoid Common Table OCR Errors
- Capture each table separately. Multiple tables in one image confuse the structure parser
- Avoid skewed photos. Take the picture straight-on; perspective distortion breaks column alignment
- Merged cells are the hardest case. If your table has many merged cells, expect to manually fix a few
- Re-paste as Values in Excel. When pasting, use "Paste Special → Values" or "Match Destination Formatting" to avoid weird formatting
- For accounting tables: spot-check totals after pasting. OCR is reliable but never zero-error on numbers
Where Table OCR Earns Its Keep
- Reports & research papers — grab data tables instead of retyping
- Old printed forms — digitize legacy paperwork in minutes
- Screenshots of dashboards — capture data from BI tools that don't expose CSV exports
- Foreign-language tables — combine with to OCR + translate column headers
Summary
Extracting a table from an image to Excel used to require enterprise software. In 2026, a free AI OCR tool gets you 90% of the way for free, in seconds. For most users, is the easiest path: drop the image, copy the markdown/TSV output, paste into Excel.
For enormous, repetitive workloads — invoices, receipts, forms — specialized tools earn their cost. For everyone else, the AI route wins.
FAQ
Can OCR extract tables from images?
Yes. AI OCR tools like understand table structure — rows, columns, headers, merged cells — and return clean tabular text you can paste straight into Excel or Google Sheets.
How do I avoid OCR output coming back as LaTeX?
Some AI tools default to LaTeX for tables, which is hard to paste into spreadsheets. img2txt.app outputs plain text tables (tab-separated or markdown) by default — copy-pasteable, no LaTeX compilation needed.
Will the rows and columns stay aligned after OCR?
On clean printed tables, yes — alignment is preserved. Merged cells and very dense tables sometimes need light cleanup. Always spot-check before relying on the data.
Can I extract a table from a screenshot of a webpage?
Yes. Screenshot the table, drop it into img2txt.app, and the structure is preserved. This works well for sites that block copy-paste or render tables in canvas/SVG.
Is it free to convert image tables to Excel?
Yes, with limits. img2txt.app allows 3 free conversions per day for anonymous users, 6 for registered users. For very heavy batch table extraction, Pro plans offer unlimited usage.
