PNG to CSV converter

PNG to CSV accepted the receipt image, but the tested output collapsed the receipt into a one-column CSV instead of preserving Qty and Amount columns.

Converted with AI Deep Read · input deleted within 24 h
Free: 5 conversions a day · 5 MB each
Higher limits →

From real runs

What you get — and what changes

Element	Result	What happens
CSV file	Kept	The API returned a valid text/csv file, so the pair works as a downloadable CSV conversion.
Receipt columns	Lost	Item, Qty, and Amount did not survive as CSV columns in this run; only the first-column text appeared.
Item names	Kept	Product, tax, coupon, and total labels were present as separate records in the one-column output.
Amounts	Lost	The PNG CSV output omitted 18.90, 12.49, -4.20, 2.24, -3.00, and 26.43 from the visible records.
Payment details	Changed	Payment details were reduced to labels such as Card, Auth, and Note instead of retaining their full values.
Encoding	Kept	The CSV was plain UTF-8 text with newline-separated records and no binary workbook packaging.
Runtime	Kept	The PNG-to-CSV run completed in 6.4 seconds, produced 168 bytes, and used one Deep Read credit.

CSV file Kept
The API returned a valid text/csv file, so the pair works as a downloadable CSV conversion.
Receipt columns Lost
Item, Qty, and Amount did not survive as CSV columns in this run; only the first-column text appeared.
Item names Kept
Product, tax, coupon, and total labels were present as separate records in the one-column output.
Amounts Lost
The PNG CSV output omitted 18.90, 12.49, -4.20, 2.24, -3.00, and 26.43 from the visible records.
Payment details Changed
Payment details were reduced to labels such as Card, Auth, and Note instead of retaining their full values.
Encoding Kept
The CSV was plain UTF-8 text with newline-separated records and no binary workbook packaging.
Runtime Kept
The PNG-to-CSV run completed in 6.4 seconds, produced 168 bytes, and used one Deep Read credit.

Sample conversions

Receipt table PNG

Input: PNG receipt with columns for Item, Qty, and Amount, including a returned binder, sales tax, coupon, and total.
Output: CSV contained lines such as NORTH PIER MARKET, Item, Printer paper A4, USB-C cable 2m, and TOTAL. Amounts were missing.

Small table screenshot

Input: PNG table where surrounding prose and headings are less important than the row-and-column data.
Output: Use XLSX if this pair produces a single-column fallback. XLSX preserved the same receipt's numeric values more reliably.

A real PNG → CSV run

We converted our PNG sample (219 KB) with AI Deep Read under AI Deep Read on 2026-09-11. It took 11 s and produced a 364 B CSV file, 100% smaller than the input. The engine raised 1 warning, listed below.

Input · sample.png · 219 KB

1240 × 1754 px, 8-bit, colour type 2

Width: 1,240
Height: 1,754
Encoding: 8-bit, colour type 2

Output · sample.csv · 364 B

12 data rows × 1 column; header: Feedback Atlas — Q3 activity trail

Rows: 12
Columns: 1
Header: Feedback Atlas — Q3 activity trail

AI Deep Read read 1 page; 1 credit per page

First lines of the output

csv

Feedback Atlas — Q3 activity trail
Prepared for: Board of Directors · 12 Sep 2026
**Headline numbers**
Metric
ARR
Net revenue retention
Support CSAT
Refunds
**What went well**
• Enterprise pipeline doubled after the SOC 2 report shipped.
• Support CSAT held at 94% while ticket volume rose 31%.
• Two renewals slipped into October. Owner: VP Sales.
Signed

Is your file a scan?

PNG-to-CSV uses 1 Deep Read credit for the image, then our table engine writes CSV from the model's table text. On the tested receipt, that second step produced a one-column fallback, so treat CSV as a draft extraction.

Options and when to change them

extractTablesAsGfm: Keep this enabled; if the model does not emit a clean pipe table, CSV may collapse to a single column.
preserveLayout: Try it when a PNG table has blank cells or multiple sections that need spatial context.
languageHint: Set it when OCR-sensitive labels are in a known language and short column names are easy to confuse.

Do it yourself

Tesseract + pandas 2.3

bash

tesseract receipt.png stdout --psm 6 > receipt.txt
python3 - <<'PY'
import pandas as pd
rows = [['Item','Qty','Amount'], ['Returned binder',1,-4.20]]
pd.DataFrame(rows[1:], columns=rows[0]).to_csv('receipt.csv', index=False)
PY

Manual row review is required; the tested API run shows why CSV extraction from receipt images must be checked.

PNG to CSV by API or MCP

The HTTP API and the MCP server run the exact engine this page uses (AI Deep Read); the response carries the same warnings the web converter shows. Check support first with GET /api/v1/pairs?from=png&to=csv.

curl

curl -X POST https://filetypeconverters.com/api/v1/convert \
  -H "Authorization: Bearer $FTC_API_KEY" \
  -F "file=@input.png" -F "from=png" -F "to=csv" \
  -F "deepRead=true" \
  --output output.csv

API reference

When this is the wrong conversion

If amount columns matter, use /convert/png-to-xlsx and export CSV from the reviewed workbook.
If the PNG contains paragraphs rather than tables, /convert/png-to-md or /convert/png-to-txt will better match the content.
If the table has merged headers, create a spreadsheet manually from OCR notes instead of trusting one-pass CSV.

PNG to CSV questions

Why did my PNG receipt CSV lose amounts?

The tested receipt produced a valid CSV, but the amount column did not materialize. Use XLSX for the same image when numeric recovery matters.

Can PNG to CSV still be useful?

Yes for simple label lists or tables where the output can be reviewed quickly. Do not import the CSV automatically until you compare columns against the source image.

Should I enable table extraction?

Yes. The CSV route depends on a clean pipe-table transcript before serialization. If the result is one column, rerun with layout preservation or switch to XLSX.