Skip to main content

Upload Documents to Chat

Upload photos or PDFs of receipts, invoices, and invoices. The assistant automatically extracts data with OCR.

Supported Formats​

FormatMax sizeUse cases
PDF25 MBInvoices, quotes, payrolls
JPG10 MBPhotos of receipts, tickets
PNG10 MBScreenshots, scanned documents

How to Upload a Document​

From the Chat​

  1. Click the paperclip icon (📎) in the input bar
  2. Select your file (PDF, JPG, PNG)
  3. The assistant processes the image
  4. It shows you the extracted data

Option: Drag and Drop​

Drag your file directly to the chat:

  1. Open the assistant's chat
  2. Drag the image or PDF from your folder
  3. Drop it into the chat window
  4. It loads automatically

What the Assistant Extracts​

OCR automatically recognizes:

From Receipts/Invoices​

  • Concept: what you bought
  • Amount: how much it cost
  • VAT: if it's itemized
  • Date: when
  • Vendor: from whom you bought it
  • Category: Material, Services, Travel, etc.

Example 1: Office Receipt​

You upload a photo of the Staples receipt:

"3 packs of A4 paper - €18.50"

The assistant:

  • Extracts concept: "A4 Paper"
  • Quantity: 3
  • Amount: €18.50
  • Category: Office supplies
  • Date: Feb 8, 2025

It asks you: "Should I create an expense?"

Example 2: Invoice from a Vendor​

You upload an invoice you received from your vendor:

Invoice:

Electricidad - Febrero 2025
Total: €145,30 + 21% IVA = €175,81

The assistant:

  • Detects: Service expense (electricity)
  • Amount: €175.81 (with VAT)
  • Category: Services
  • Date: February 2025

Example 3: Client Invoice​

You upload a PDF invoice that you issued and the client paid:

The assistant:

  • Extracts invoice data
  • Document number
  • Concept and amount
  • Suggests marking it as "Paid"

Flow After Upload​

  1. The assistant processes the document (1-3 seconds)
  2. Shows the extracted data in editable format
  3. Questions: "Should I create an expense of €45.00 for Material?"
  4. You confirm or correct the data
  5. It is automatically created in Frihet

If something is not accurate, edit directly before confirming:

  • Change concept
  • Adjust amount
  • Change date
  • Reassign category

Special Cases​

Invoice with Multiple Lines​

If you upload an invoice with 5 different concepts:

The assistant:

  • Extracts each line
  • Asks if you want to create one expense with the total
  • Or if you want to create 5 separate expenses (if they are different categories)

Recommendation: Respond naturally

  • "Create a single expense with everything" → €185.30 total
  • "Separate Material and Services" → 2 expenses

Receipt with Unclear Data​

If the photo is blurry or tilted:

  • The assistant attempts to extract anyway
  • Marks doubtful data as "Requires review"
  • Allows you to edit them manually

Document in English or Different Language​

The assistant understands:

  • Spanish ✅
  • English ✅
  • French ✅
  • German ✅
  • Portuguese ✅

If it's in another language, it translates it automatically.

Incomplete Document​

If critical information is missing (date or amount):

The assistant asks you:

  • "When was this expense?" → You enter the date
  • "What is the total amount?" → You edit the amount

Tips for Better Extractions​

Quality Photos​

✅ Good lighting — no shadows ✅ Complete document — everything fits in the photo ✅ Right angle — avoid tilted photos ✅ Decent resolution — not pixelated

❌ Very dark photo ❌ Cropped document ❌ Very tilted photo ❌ Blurry photo

Information Order​

If possible:

  • Make the concept visible
  • Ensure the amount is clear
  • Include the date
  • Show the vendor/company

Cases Where OCR Fails​

Cash register tickets:

  • Very small letters
  • Degraded thermal paper
  • Intertwined numbers

Solution: Edit manually afterward. The assistant will allow you to.

Creating Expenses vs Invoices​

The assistant automatically suggests:

TypeWhen
ExpenseIt's a receipt for something you bought
InvoiceIt's a document you issued to a client
Received invoiceIt's an invoice from a vendor

If it's not sure, it asks:

  • "Is it an expense or an invoice?"
  • You answer and it processes correctly

Document Privacy​

  • Documents are processed with local OCR
  • They are not saved on third-party servers
  • The extracted data is saved in your Frihet account (encrypted)
  • You can delete the original document afterward

From Settings → Privacy you can:

  • Delete old uploads
  • Disable OCR if you prefer

Limitations​

The assistant cannot:

  • Process Word or Excel files (only PDF/JPG/PNG)
  • Extract from very deteriorated images (< 100px wide)
  • Recognize handwriting (manual writing)
  • Read QR codes or barcodes

Next section: Integrations