Less manual retyping
Invoice review
Turn photographed invoices and PDFs into text that finance teams can check, copy, and move into their existing process.
Made and operated in EuropeScans in, usable text out. Under European law.
For teams that handle invoices, receipts and contracts: upload a PDF or image and get copyable text, no account needed. Operated by a German company, DPA available.
Trusted by companies of all sizes and industries
- Operated by a German company
- GDPR · DPA available
- ISO 27001-certified data centres
Input: Source file
Loading the sample document…
Or test your own document
Drop an image or PDF here
JPEG, PNG, WebP, or PDF · maximum 5 MB
Choose fileOutput: Structured data
The structured JSON response will appear here.
10 requests left in 24hFiles are processed for this result and are not added to chat history.
Put the text to work
Less copying. Fewer broken handoffs.
OCR is useful when the result can move straight into the work that follows. Copy the extracted text, check it against the source, and send it to the right person or system.
Faster handoffs
Customer operations
Read forms, letters, and screenshots without switching between a document and a blank response field.
Useful text from old scans
Searchable archives
Recover copyable text from scans so older documents can feed search, summaries, and internal knowledge work.
A clear review step
Document checks
Compare the extracted text with the source before names, dates, totals, or contract wording move downstream.
Estimate your cost
What would it cost to process your documents?
Production OCR runs through the Inference API on Qwen 3.8 27B, billed per token like any chat completion — no flat per-page fee. Pricing scales with pages, not documents: a 10-page contract costs about 10x a one-page receipt. Estimate a rough cost below.
Pages / month
$0.00
$0.00 per page
1001,00010,000100,000
Estimate assumes 1,500 input and 600 output tokens per page (an illustrative token budget), at Qwen 3.8 27B's current Inference API pricing ($0.42/1M input, $2.55/1M output). Count each page of a multi-page PDF separately. Higher-resolution images and denser pages cost more — this is a starting estimate, not a quote.
Frequently asked questions
How accurate is the OCR?
Accuracy depends on the source: printed text with clear resolution and lighting extracts best. Always compare names, dates, totals, and other important details against the original document before you rely on the result.
Which file types and languages are supported?
The demo accepts JPEG, PNG, WebP, and PDF files up to 5 MB. Text recognition is not limited to a single language — try a sample in your own language to see how the result looks.
Does it work with handwriting?
The demo is built for printed and typed text, such as invoices, forms, and screenshots. Handwriting can be recognized in some cases, but accuracy is lower than for printed text — treat handwritten results as a starting point, not a final source.
What happens after 10 requests in 24 hours?
The public demo allows 10 requests per 24 hours so it stays available to everyone without requiring an account. To keep extracting text and continue the work — summaries, comparisons, follow-up questions — use OCR inside LLMBase Chat.
What happens to my files? Is anything used for training?
Customer content is not used to train, fine-tune, evaluate, or improve AI models. Demo files and extraction results are not added to chat history, and file contents are checked before processing starts.
Do you support business use, GDPR requirements, or a DPA?
Yes. LLMBase is operated by a German company with privacy terms written for European businesses, and provides a Data Processing Agreement for business customers. See the privacy policy for details.
Continue with LLMBase Chat
Extract the text, then work with it
Upload documents in AI Chat to summarize content, compare files, and ask follow-up questions without starting over in another tool.
Review names, dates, totals, and legal wording against the source before relying on extracted text.