Blog
Explore here ↓Comparisons & benchmarks
What Odoo's built-in OCR did to a ₹61,250 Indian vendor bill, tested today
We ran Odoo's IAP OCR on a real Indian vendor bill (Nice Packers and Movers, ₹61,250 total). The Bill it created came back at ₹6.00, with one fake line item and the bill number parsed from the vendor's address. Verbatim output, screenshots, and the same bill through our pipeline.
Xero/Hubdoc got the bill total right. The tax came back ₹0. We tested on a real Indian GST invoice.
We ran Hubdoc (the OCR bundled with Xero Business Edition) on a real Indian vendor bill. The total came through correctly at ₹61,250. The vendor was 'No Contact', the tax was zero, and the eight line items were collapsed into one. This is the most misleading failure mode of the three bundled OCRs we tested, because it looks clean.
What Zoho's terms say, and what its Advanced Autoscan actually did to a real ₹61,250 bill
A line-by-line read of Zoho Receipts' terms of service and Zoho's corporate privacy policy, plus a field-by-field test of what Zoho Books' Advanced Autoscan extracted from a real Indian vendor bill. The Bill it posted was ₹4,250 short of the printed total, with the GST line silently dropped.
Can ChatGPT read invoices? ChatGPT vs a purpose-built OCR for Indian GST bills
An opinionated comparison of pasting Indian GST bills into ChatGPT or Gemini versus using a purpose-built OCR pipeline: where the chat approach works, where it silently fails, and when a CA should pick which.
India OCR accuracy benchmark: how the major tools handle real GST bills
We ran real Indian GST bills, including a handwritten invoice and a Devanagari-language bill, through the major OCR tools and scored raw text, structured GST fields, Tally-ready output and India pricing. Here are the results.
Handwritten & thermal
Digitizing the messy bills: handwritten chits, faded thermal, and the petrol-pump pile
What actually works when you have to get handwritten kirana bills and faded thermal receipts into a GST-ready Excel sheet, with our own test results on real handwritten and thermal documents.
Thermal receipt to Excel: why faded petrol and retail bills defeat generic OCR
The thermal-paper problem explained for Indian CAs: why heat-printed petrol-pump and retail receipts fade and break ordinary OCR, what our own thermal test showed, and how to set the job up so the small print survives.
International invoices
Facture vers Excel et Sage: ce qu'une OCR doit vraiment lire sur une facture TVA française
Comment sortir le numéro de TVA fournisseur et client, le HT, la TVA et le TTC d'une facture française vers Excel ou Sage, avec un test direct sur une vraie facture TVA Paris-Lyon.
Factura española a Excel y a Holded: el tiquet que te dio el cliente no es una factura, y un buen OCR debería decirlo
Pasamos un tiquet térmico real de El Corte Inglés en Madrid por nuestra pipeline. Leyó todos los campos con alta confianza y marcó el documento como no deducible, que es justo la parte que la mayoría de OCRs pasan por alto. Pensado para asesores fiscales y contadores en España, México, LatAm y pymes hispanohablantes en EE. UU. que cargan a Holded, Aspel, Alegra o QuickBooks.
Regional language (India)
How to get data out of a Hindi invoice without typing it all by hand
Why phone scanners and ChatGPT mangle handwritten and regional-language bills, and a faster way to pull vendor, GSTIN and tax amounts into Excel.
Pulling invoice data out of Indian-language bills, language by language
A working CA's guide to getting vendor, GSTIN and tax fields out of Hindi, Marathi, Gujarati, Kannada, Telugu and mixed-script bills into Excel, and which problems are language-specific versus universal.
Why OCR scrambles your invoice columns: linearization and reading order, explained
The reason general OCR turns a four-column bill into one garbled line is a step called linearization. Here is what reading-order detection is, why Indian-language tables make it harder, and how to tell if a tool gets it right.