Upload & extract
Upload a PDF, image, or document and Flitch extracts the structured data into a dataset.
For a file with no structure to walk: a PDF report, a scanned table, an invoice, a screenshot. Flitch reads it, finds the tables and figures, and turns them into a dataset.
Anything with a structure already written down goes through file upload, which parses it outright. That is most things, including several that do not look it: a GPS route, a KML of placemarks, a feed, an archive, and a browser's network recording.
Upload
Open the form
Go to Data → Add → Upload & extract.
Add a file
Drop in a PDF, image, or document.
Review the extracted data
Flitch reads the file, pulls out the tables and figures it finds, and shows a preview. Confirm the rows and columns.
Save
The extracted data becomes a dataset.
File types
| Extension | Type | Notes |
|---|---|---|
.pdf | Reports, statements, invoices. Tables and figures are read off the page. | |
.png, .jpg, .jpeg, .gif, .webp | Images | Screenshots, photos, scans. Text and tables are read. |
.docx, .txt, .md | Documents | Structured content is extracted. |
Everything else goes through file upload, which parses it directly and costs nothing to run. Check its list before reaching for extraction: a file that can be parsed will be read exactly, where extraction is a reading and worth checking.
Privacy
Your file is processed to extract its data and stays in your workspace. It is deleted when you delete the dataset.
Troubleshooting
Nothing extracted. The file had no structured data Flitch could read, or the scan quality was too low. Try a clearer copy.
Wrong columns. Extraction inferred the shape. Open the dataset to fix column types or values.