imissfiles56 tools
  1. Home
  2. Guides
  3. How to get a table out of a screenshot

How to get a table out of a screenshot

A screenshot, a photo, a table stuck inside a PDF. How to turn each one into rows and columns you can actually use.

The table is right there on the screen. You can read every number. You cannot select any of it. It came as a picture in a chat, a dashboard that only exports images, a slide, a report someone photographed and emailed. Retyping it is slow and you will make mistakes. This guide is the other route: extract the table from the image, let the rows and columns be rebuilt, then check the result. It works the same whether you started with a screenshot, a photo or a scan.

All of the tools linked here do that in your browser. The image is not sent anywhere.

First: can you just copy it?

Before you reach for recognition, check whether the table is really text. Try to select a single cell. If the text highlights, it is text, and copying it straight into a spreadsheet will always beat reading it from pixels. That is often the case on web pages and in PDFs that were exported rather than scanned. For a PDF with selectable text, PDF to Excel and PDF to CSV read the text directly and skip recognition entirely.

If nothing selects, the table is an image, and the rest of this guide applies.

Extracting a table from an image: pick the output first

You wantUseBest for
A spreadsheet a person will openImage to ExcelA real .xlsx that opens in Excel, Google Sheets, Numbers and LibreOffice
Plain data from a screenshotPNG to CSVImports, databases, pandas, R, any script
Plain data from a phone photoJPG to CSVThe same, tuned for camera JPEGs
Any image format as CSVImage to CSVWebP, GIF, BMP and anything else
Words with positions, for codePNG to JSON or Image to JSONEvery word, its bounding box and confidence score

If you are not sure, choose Excel. It is the easiest to check by eye, and every spreadsheet can save it as CSV afterwards.

Capture it properly

The quality of the image decides the quality of the table far more than the tool does. Recognition matches letter shapes, and column detection relies on words lining up. Both suffer from the same few things.

Screenshot, do not photograph the screen. A screenshot is exactly the pixels your display drew: perfectly straight, no glare, no moiré. A photo of a monitor has all of those problems at once. If the table is on your own screen, the screenshot key is always the better option.

Capture at full size. A table grabbed while zoomed out, or a screenshot that was shrunk before it reached you, has too few pixels per letter. Zoom in until the table fills the screen, then capture. Enlarging a small image afterwards does not help, because resizing invents no detail.

Crop to the table. Menus, headers and sidebars become stray rows. Use Crop image to cut down to just the grid first.

If it has to be a photo, shoot square on. A page photographed at an angle has columns that converge down the page, and that breaks column detection even when every word is recognised correctly. Hold the phone parallel to the paper, fill the frame, and use even light without your own shadow across it.

Keep PNG if you have the choice. PNG is lossless. JPEG compression softens the edges of letters, which is exactly what recognition depends on. Most screenshots are already PNG.

Run it and read the preview

Drop the image into the tool you picked. Tesseract recognises the text in the tab and keeps the position of every word. Words that share a line become a row. Column boundaries come from the horizontal positions where words keep starting down the page. You then see a preview of the grid before anything downloads.

Do not skip the preview. Scan three or four rows against the original. You are looking for:

  • Merged columns. Two narrow columns read as one. Usually caused by tight spacing or a right-aligned column of numbers under a left-aligned header.
  • Split rows. Text that wrapped onto a second line inside one cell turns into an extra row.
  • Misread digits. A 5 read as an S, an 8 as a B, a decimal point lost. These look perfectly plausible in a spreadsheet, which is what makes them dangerous. Check totals if the table has them.

Fixing a handful of cells in the output is normal. If most of the grid is wrong, the problem is almost always the image, and a better capture will do more than any setting.

Can Excel do this itself?

Yes, if you have Microsoft 365. Excel has a feature called Data from Picture: on the Data tab, choose From Picture, then either Picture From File or Picture From Clipboard. On a phone, the Excel app has the same thing under Insert. It shows a preview, lets you fix cells it was unsure about, and inserts the table into your sheet. If you already pay for Microsoft 365 and work in Excel all day, it is the quickest route and worth knowing about.

The tools in this guide are for the other cases: you are on a computer without Microsoft 365, you want CSV or JSON rather than a workbook, or you would rather the image never leave your device. Google Drive can also read an image if you open it as a Google Doc, but that uploads the picture to your Drive and gives you a document, not a spreadsheet.

Tables inside PDFs

A scanned PDF is a stack of pictures. If the table you need is on a scanned page, pull the page image out with Extract PDF images, crop it to the table, and run it through the tools above. For a scan you mainly want to search rather than turn into cells, OCR a PDF adds a text layer to the whole document instead.

What this will not do

It will not read handwriting. Tesseract is trained on printed type, and a handwritten grid is the hardest case there is. It will not turn a chart into numbers, because bars and lines are not text; it will pick up the title and labels and nothing else. And it will not understand merged header cells that span several columns. Those land in one column and need moving by hand. For more on where recognition is reliable and where it quietly fails, see what OCR can and cannot do.

Common questions

Can I turn a screenshot of a table into a spreadsheet?

Yes. Add the screenshot to Image to Excel for an .xlsx workbook, or PNG to CSV for plain comma-separated data. The text is recognised in your browser and the rows and columns are rebuilt from where each word sat. Nothing is uploaded.

Should I choose Excel, CSV or JSON?

Excel when a person is going to open the file. CSV when it is going into a script, a database or an import. JSON when code needs every word with its position and confidence score.

Why are two of my columns merged together?

Columns are inferred from where words consistently begin down the page. Merged cells, wrapped text and right-aligned numbers under a left-aligned header all confuse that. Fix the affected rows after import, or crop to the part of the table that is cleanly aligned.

Can it read the numbers off a chart?

No. It reads text, so it will pick up the title, axis labels and legend, but the bars and lines themselves are not text. You need the table the chart was drawn from.

Is a screenshot better than a phone photo?

Yes, by a wide margin. A screenshot is machine-drawn and perfectly aligned, with no angle, shadow or lens distortion. If you can screenshot the table instead of photographing the screen, do.

Does the image get uploaded?

No. Tesseract runs as WebAssembly in the tab, and the spreadsheet, CSV or JSON is built on your device. Open the Network panel while it runs and no request carries the image.

Can I convert a screenshot of a table to Excel for free?

Yes. Image to Excel is free, needs no account and has no daily limit. Drop in the screenshot, check the preview, and download a real .xlsx. Excel's own Data from Picture feature does the same job, but it needs a Microsoft 365 subscription.

How do I copy a table from an image into Excel?

You cannot copy it directly, because the table is pixels, not text. Extract it first: run the image through Image to Excel, download the workbook, and copy the cells from there into the sheet you are working in.

Tools on this page