Work

Zonov Vista: An Enterprise Vision Intelligence Model for Multilingual Document Understanding and Structured Information Extraction

Year: 2026

Type: article

Abstract: Zonov Vista is a general-purpose document intelligence and vision model developed by Zonov.ai. It parses printed, handwritten and mixed documents in English, Hindi and nine further Indian languages directly from the page image into structured JSON, without templates or a separate OCR stage, and copies every value exactly as it appears on the page. The model has about 8 billion parameters, runs on a single GPU, and does not retain documents after processing. The research presents the model's document processing capabilities, including multilingual OCR, layout understanding, contextual analysis, handwritten text recognition, and structured data extraction. Zonov Vista is intended for diverse enterprise applications, including invoices, receipts, forms, applications, financial statements, certificates, reports, and other business documents. This paper presents its technical approach, evaluation methodology, benchmark results, and potential applications in intelligent document processing.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.23974491

Authors: Kumar, Ankit, Rajpoot, Akash, Chawala, Arvind, Panwar, Dr. Parul, Bansal, Lavi

Language: English

Topic: Handwritten Text Recognition Techniques

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7225311064

FWCI:

Cites:

Cited by:

Related to: 10