Everything runs on this server. No content goes to OpenAI, Anthropic, Google or any other service. No ad networks, no cross-site tracking.
As little loss as possible
Numbers, names and tables stay exactly as they appear in the original. Nothing is summarised or rewritten.
Deleted automatically
Files and results disappear on their own after 24 hours — or right away, if you prefer.
What people use this for
Everyday cases, no jargon.
Photographed an invoiceLine items you can copy instead of retype.
Scanned a delivery noteQuantities and part numbers as a table.
Contract as a PDFFind individual clauses with a text search.
Photographed a price listA table ready for your spreadsheet.
Lecture notesText for summaries and flashcards.
Old letter from the filing cabinetSearchable instead of just an image.
Excel spreadsheetAs Markdown for tools that can't read XLSX.
The conversions that matter
What survives the trip, and what doesn't.
PDF to Markdown
Invoices, contracts, manuals. Headings, paragraphs and tables carry over. With multi-column layouts the reading order can get scrambled.
Photo to text
A photographed document becomes text you can copy. A straight, sharp photo gives the best result.
Word to Markdown
Headings, paragraphs, lists and tables arrive structured. Comments, footnotes and text boxes are lost.
Excel to Markdown
Every sheet becomes a Markdown table with a header row. Cell formats, formulas and charts are not carried over.
Scan to text
Even an old, yellowed letter becomes searchable. Poor scan quality or handwriting lowers recognition.
PowerPoint to Markdown
Each slide becomes a heading and a bullet list. Graphics, animations and notes are lost.
What goes in, what comes out
Klartext accepts these file types:
BMP
DOCX
HTML
JPEG
JPG
MD
PDF
PNG
PPTX
TIF
TIFF
WEBP
XLSX
You always get two files: .md to read and reuse, .json with the full structure for other programs.
The same limits for everyone
Technical fair-use limits keep the service stable for everyone. They exist to prevent overload — not to sell you anything.
Common questions
Is Klartext really free?
Yes. There are no tiers, no payment and no locked features. All you need is a free account.
What is Markdown, anyway?
A simple markup language for text: headings, lists and tables are marked with ordinary characters like # or |. Readable as plain text and, in note apps, wikis or on GitHub, nicely formatted.
What is the JSON file for?
It holds the same content broken down for machines — pages, blocks and table cells captured individually. Handy for your own programs, automations or databases, where Markdown would be too coarse.
Are my documents sent to an AI provider?
No. The entire conversion runs on our own server. No content is sent to OpenAI, Anthropic, Google or any other AI or OCR service.
Are my documents stored?
Only temporarily, for processing. Uploaded files and results are deleted automatically after 24 hours, and you can remove any job by hand before that. Klartext is not a storage service.
Is the result error-free?
No. The conversion is automatic and can contain mistakes with poor scans, handwriting or nested tables. Markdown also can't represent every layout — the JSON file is the structure-faithful version. Check results before you rely on them.
How well does Klartext read handwriting?
Honestly: badly. Text recognition is trained on printed text. Handwritten notes, signatures or hand-filled forms often come out wrong or not at all — for print and good scans it works reliably.
Does it work with phone photos?
Yes. Sharpness and resolution decide: a straight, sharp, well-lit photo gives usable results, a blurry or dark one trips up the text recognition.