# How to add document OCR to TallyPrime > Source: https://netray.co/integrations/tally-document-ocr > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Integrations ## Short answer Run OCR and field extraction on scanned bills, then check them against Tally's ledgers and GST rules and import vouchers via XML after a reviewer confirms low-confidence fields. Tally does not read documents itself, so OCR is a pipeline in front of it. ## System TallyPrime ## Capability document-ocr ## Intro A large part of Tally's workload in trading and manufacturing firms is paper: purchase bills, delivery challans, expense receipts and bank slips. Clerks type these in, and errors in quantity, rate or GSTIN show up weeks later during GST reconciliation. OCR on its own only turns pixels into text. The value comes from extracting the right fields, validating them against Tally's masters and tax rules, and giving the reviewer a quick way to fix doubtful ones. The pipeline should treat confidence as a first-class setting, because indirect cost sits in the entries that look right but are not. ## Use Cases ### Handwritten and printed purchase bills **Description:** Extract supplier, bill number, date, line items and tax from paper bills photographed on a phone or scanned in batches. ### Delivery challan to receipt note **Description:** Read challan numbers and quantities to prepare a receipt note against the purchase order. ### Expense receipts **Description:** Capture fuel, travel and petty cash receipts into payment or journal voucher drafts with the right expense ledger. ### GSTIN and HSN verification **Description:** Check extracted GSTINs against format rules and HSN codes against stock item masters. ### Cheque and deposit slips **Description:** Read cheque numbers, dates and amounts to prepare receipt or contra entries for matching with bank statements. ## System Surface - Voucher Import Data via the XML interface: purchase, payment, journal and receipt note vouchers with ledger and inventory entries. - Master Export Data to fetch current ledgers, stock items, units and GST classifications for matching. - Voucher attachments: recent TallyPrime releases support attaching documents to vouchers, though the way to supply them through the API varies, so confirm for your version and fall back to a linked file path in narration. - Batch and godown fields on inventory entries, relevant when challans carry batch numbers. - TDL fields to carry an OCR reference id or confidence flag on the voucher for later audit. ## How It Works ### Standardise capture **Detail:** Define how documents arrive: scanner folder, phone photos or email. Add basic image fixes like deskew and contrast before OCR. ### Run OCR and layout parsing **Detail:** Use an engine that returns word positions and confidence. Supplement with a layout-aware model for tables and totals. ### Extract the fields **Detail:** Map to party, number, date, lines, taxes and totals. Cross-check line sums against the stated total. ### Match to Tally masters **Detail:** Fuzzy-match names to ledgers and items from the exported master lists, and store the supplier-specific aliases that reviewers confirm. ### Apply confidence thresholds **Detail:** Auto-accept only high-confidence fields. Route anything under the threshold, or any failed arithmetic check, to review. ### Import and archive **Detail:** Send approved vouchers to Tally, store the image and the extracted data, and record the voucher number. ## Example Flows ### Photographed bill from a supplier **Trigger:** Purchase staff uploads a phone photo to a shared folder **Steps:** - Image is cleaned and OCR runs. - Extraction finds bill number, date, three lines and CGST/SGST. - Totals fail a check by a small amount; the doubtful line is highlighted. - Reviewer corrects the rate and approves; voucher is imported. **Result:** The entry is created in a minute or two, and the corrected rate improves the supplier's alias data. ### Batch of challans at month end **Trigger:** Scanner drops a multi-page PDF **Steps:** - Pages are split into documents. - Challan numbers and quantities are matched to open purchase orders. - Unmatched challans go to a review list. **Result:** Receipt notes are prepared in bulk, with exceptions visible rather than hidden. ## On Prem OCR and extraction can run entirely inside the office. Open OCR engines and an open-weight vision or language model hosted on a local GPU machine can handle printed bills, and Tally stays on the LAN. This matters when bills reveal pricing and customers. Keep scans in a local store with access limited to accounts staff. Handwritten documents are harder, so expect a higher review share and measure it before promising savings. ## Security - Store scans in an access-controlled location and link them to vouchers by reference. - Do not send unredacted bank details and GSTINs to external OCR APIs unless your policy and agreements allow it. - Keep a full audit trail: image, extracted values, reviewer changes and import result. - Limit the import service to creating draft or specific voucher types, not altering existing ones. - Retain originals for statutory retention periods set by your advisor. ## Options ### Cloud invoice OCR with Tally export **Fit:** Firms comfortable with a hosted service and standard invoice layouts. **Effort:** Days to set up; verify voucher mapping and data residency. ### Open-source OCR plus custom extraction **Fit:** Teams with development capacity and a limited set of suppliers. **Effort:** Four to eight weeks to reach dependable accuracy. ### Partner build with Netray **Fit:** Mixed document types, local-only processing, and review tooling tied to Tally. **Effort:** Scoped per project; includes a tuning period on real documents. ## Pitfalls - Arithmetic validation matters more than OCR confidence; a confident wrong digit still passes a character-level score. - Handwriting, stamps over text and faint thermal receipts drop accuracy sharply. - Auto-creating new ledgers from OCR text clutters the chart of accounts; require approval for new masters. - Tax splits differ between intra-state and inter-state bills; place of supply must drive the ledger choice. - Without a feedback loop the same supplier errors repeat every month. ## Faq ### Does TallyPrime have built-in OCR? **Answer:** Not for reading scanned bills into vouchers as a general feature we can rely on. Tally lets you attach documents to some records, but extraction and posting need an external pipeline that outputs voucher XML or files for import. ### How accurate is OCR for Indian GST bills? **Answer:** Printed computer-generated bills can be read well, while handwritten or low-quality scans need much more review. Measure accuracy on a sample of your own bills before deciding thresholds, and expect lines and tax splits to need most corrections. ### Should the OCR system post directly into Tally? **Answer:** Only for narrow cases with clean history. Most firms get better results with a review step, then import. A misread amount posted automatically is costly to find later in GST reconciliation. ### Can this run without sending bills to the cloud? **Answer:** Yes. Local OCR and a locally hosted open-weight model can extract fields from printed bills, with Tally on your network. Plan GPU or CPU capacity for batch days, and test the engines on your own documents. ### How do I link the scan to the voucher? **Answer:** Where your TallyPrime version supports document attachment through its interface, attach it. Otherwise store the file in a document system and put a reference id in the narration or a custom field so auditors can find the source. ## Search terms tally ocr, tallyprime invoice ocr, scan bills into tally, tally purchase bill ocr, ocr to tally voucher xml, tally document capture, tally gst invoice ocr, automate tally data entry from scans --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json