Skip to content

Free tool

Image to CSV. No account needed.

Photograph a table, a receipt, or a printed report. AI reads it and returns structured, editable CSV data. Supports JPG, JPEG, PNG and HEIC.

What you get

Ungated, and it deletes your file.

This tool is the front door to Mammoth. Use it on its own, or carry the extracted data into a pipeline that runs every week.

  • No account required

    Upload and extract straight away. Your file is processed and deleted within an hour.

  • Reliable on clean tables

    Tables, receipts and printed reports come back as usable rows. Messy scans and mixed formats usually work too — have a look at the output before you rely on it.

  • One extractor, every format

    The same engine reads images, PDFs and statements. Pick whichever page matches the file in front of you.

Try it on the image in front of you

The AI finds tables, receipts and text-heavy regions in a JPG, PNG or HEIC, then returns clean rows and columns you can open in Excel.

Doing this every month? Turn it into a pipeline that runs itself.

How to convert an image to CSV

Works on a photo, a screenshot or a scan. The clearer the picture of the table, the less there is to check afterwards.

  1. 1

    Upload the image

    JPG, JPEG, PNG or HEIC. A screenshot is usually easier than a photo — no perspective, no shadow, no focus to get wrong.

  2. 2

    Let it find the table

    The AI locates tables, receipts and text-heavy regions in the image rather than reading it top to bottom, so surrounding text does not end up as rows.

  3. 3

    Check the columns

    Look at the header row and any column where a stray mark could change a digit. This is the step people skip and then regret; it takes a few seconds on a small table.

  4. 4

    Download the CSV

    Open it in Excel, Sheets or Numbers, or import it straight into whatever wanted the data. No account, and the image is deleted within an hour.

Questions people ask

I want to put image links into a CSV, not pull a table out of one.
Different job, and this is not the tool for it. If you are adding a column of image URLs to a product feed or a Shopify import, you want the file the images live in — upload them to your store or host, copy the URLs, and paste them into a column of the spreadsheet you already have. This page does the opposite: it reads a picture of a table and gives you the numbers that were in it.
Can I do this in Python instead?
Yes, and if you already work in Python it is a reasonable route: pytesseract or an OCR API for the reading, pandas to shape the result. It is worth it when this is a repeating job inside something you are already building. It is not worth it for one image, and the fiddly part is not the OCR call — it is the table structure, which is what you would end up hand-correcting anyway.
Could I just ask ChatGPT?
For a small, clear table it will often do a decent job, and it can hand you a CSV. Two things to watch. A language model will fill a cell it cannot read confidently rather than leaving it empty, and that is the failure you least want in a number. And it has no fixed limit you can rely on — the same prompt on a bigger image can quietly stop partway. Check the output either way; on anything you are going to calculate from, check it twice.
How accurate is it?
Good on a clear photo or screenshot of a printed or on-screen table; less good on a skewed phone shot, a low-resolution image, or a table with merged cells and no ruled lines. We do not publish an accuracy figure, because we have not measured one on images, and a number measured on clean structured documents would not be honest here. Check the output against the picture before you rely on it.
What happens to my image after I upload it?
It is processed and deleted within an hour, and no account is required, so there is nothing to sign up for and nothing kept.
I need a spreadsheet, not a CSV file.
A CSV is a spreadsheet file — Excel, Sheets and Numbers all open it on a double-click, and Save As gives you .xlsx if a workbook is what you need. If Excel specifically is where this is heading, our JPG to Excel page is the same extractor written for that job.
This is a monthly job, not a one-off.
Then the converter is the wrong shape of answer. Doing it once is two minutes; doing it every month and joining the results is a pipeline. Mammoth is the platform underneath this tool — the same extraction, plus the cleaning and joining recorded once, so next month's images land in the same table without anyone rebuilding the steps. Mammoth can watch a folder and add each new batch to the same table as it arrives, or collect the files itself on a schedule.

Extracting once is useful. Extracting every week is the point.

Bring the output into Mammoth to clean it, join it to everything else, and have it delivered on a schedule.

  • 14-day Pro trial
  • No credit card
  • Viewers always free