A barcode scanner has spent decades doing one thing perfectly: beep, and the value is in your system.

Ask it to read the characters engraved on a part, printed on a label or stamped into plastic, and it has nothing to offer. A scanner is a reader, not a reader of text. It decodes a defined symbology — it cannot interpret an arbitrary character image. Whenever a customer needs a new character set, a new layout or an extra inspection point, the traditional approach runs out of road.

Scanner-OCR V2 adds that missing brain to the scanner you already own. Your scanner stays the reliable eye; Scanner-OCR is the brain that reads what it sees. Pull the trigger and the recognised text goes straight into the active field — the same gesture your team already uses for barcodes.

Everything runs on the local PC. No network, no cloud, no image ever leaves your factory.

Why scanner-side OCR doesn’t scale

Two approaches dominate today, and both hit a wall on the factory floor.

OCR built into the scanner’s own hardware is easy to deploy but frozen at the moment of manufacture. The vision model is fixed; it cannot follow your changing requirements (new characters, new layouts, more inspection points), and it cannot adapt to different industries.

Vision / multimodal large models are capable, but rarely practical on site: they need heavy compute, depend on the network (a data-compliance risk when images leave the plant), add latency that misses production takt time, and are complex and expensive to deploy.

Hardware OCR is too rigid; large-model OCR is too heavy. Neither lands cleanly on a real production line.

Scanner-OCR V2 takes the third path: a light, on-device model on the PC you already have.

What’s new in V2

V1 was a bespoke build — one customer, one codebase. V2 is a general platform: one product for every scenario.

2.0.2 — active capture by trigger pull

The current release captures on trigger pull: squeeze the trigger and the scanner returns the image immediately. This replaces the earlier polling approach and is noticeably more responsive — the capture happens when you decide, not on a timer.

🔒 Fully offline

The recognition engine runs locally. Images and business data never leave the plant, there is no network dependency, and it satisfies manufacturing data-security and compliance expectations by default.

👀 WebOCR view for operators

Two zero-training modes for the people on the line — they only ever look at the result:

• 🟢🔴 Status light — green breathe for OK, fast red flash for NA; visible from five metres
• 🖼️ Marked image — the original with colour-coded ROI boxes and text boxes
• 🎨 Colour output — matched fields in green, unmatched text in amber, no result in red

⚡ Background mode — as fast as scanning a barcode

This is the part that changes the workflow. Results no longer sit on a screen — they are typed into whatever window has focus: Excel, ERP, your in-house system.

1
2

Pull trigger → capture → recognise → data lands in the sheet
              ↑ ~350 ms end to end, as seamless as a barcode scan

• Focus injection — simulates keystrokes, supports CJK and symbols, Tab navigation and auto-Enter
• Device feedback — green/red sound and light, audible even through ear protection
• Data trail — records are logged to SQLite and exportable to Excel; fully traceable
• Tray resident — minimises to the system tray, starts with Windows, keeps running when the window closes

🎯 Rule configuration — how one product covers every scenario

Most OCR projects become “one system per customer” because the rules are hard-coded. Change the product layout and you need a developer, a rebuild and a redeploy. V2 makes the rules visual configuration: an operator can set up a new scenario in about two minutes.

1. Feature template — tell the software where to look

1
2
3

Pick a sample image → drag a box over the target (ROI) → learn features → save as template
                                        ↓
Each run: locate the region automatically → recognise only what's inside the box

• Drag-to-box on a sample image, multiple regions supported (serial, prize, date…)
• One-click feature learning (ORB keypoints) stored in the template
• Only the target region is processed — 8–10× faster, and the result is clean
• Three-level fallback: template match → feature match → whole-image, so a shifted target still works and the line keeps running

2. Regex rules — extract only the fields you want

1
2
3
4

Rule:     serial
Regex:    [A-Z0-9]{4}-[A-Z0-9]{4}-[A-Z0-9]{4}
Text:     "Ticket 5465-2270-8599 — winner!"
Result:   "5465-2270-8599"

Where to look (template) + what to take (regex) = a precise result. Templates and rules are configured independently, so changing one never breaks the other.

Onboarding a new customer is no longer a development project — it’s a configuration step.

Reuse what you already have

Scanner-OCR does not require a new device. It runs on a standard Windows PC and works with the scanner you already own.

Mounted on a stand, an existing handheld scanner becomes a fixed-mount OCR station. Because the scanner stays in place, you get repeatable framing without an operator holding it — and you avoid buying dedicated vision hardware. Existing devices, already paid for, turn into OCR scanners in minutes.

Supported devices

Scanner-OCR V2 currently supports Brady IPS scanners on the Gen7 platform:

In general, models ending in 0 are wired and models ending in 2 or 5 are wireless (1991 and 2105 are wireless).

Wired is faster than wireless — and it matters here

OCR streaming is bandwidth-hungry. Wireless scanners are limited by the Bluetooth link, which caps how quickly images can be transferred, so recognition is slower on a cordless device. For the fastest capture-to-result cycle — and for fixed-mount stations — a wired scanner is the better choice.

Using a scanner that isn’t on the list? We support additional models through custom development — contact us with the model.

Architecture

1
2
3
4
5
6
7
8

Brady IPS scanner
      │  USB (serial / HID)
      ▼
Electron desktop app (single .exe, double-click to run)
      ├─ vision-server: device control · image capture · rule storage
      ├─ engines: v5 WebGPU / v6 tiny WebGPU / v6 tiny CPU (local)
      ├─ output: focus injection · device feedback
      └─ data: SQLite records · Excel export

How it compares

Pricing

Optional add-on

All licenses are one-time — no subscriptions, nothing to renew. Full details on the Scanner-OCR pricing page.

Free 30-day trial

Try Scanner-OCR with your own scanner and your own parts for 30 days, free. Trials are arranged by contacting us today; an online order system for trials and licenses is coming soon.

Barcode quality grading (optional add-on) — please read

The optional barcode grading module scores 1D and 2D codes against the ISO/IEC 15416 / 15415 method, producing an A–F result.

Requirements and limits

• Platform: Windows 10/11, 64-bit. No GPU required.
• Scanner: a supported Brady IPS Gen7 scanner (wired recommended for speed). Non-listed models via custom development.
• Connection: USB (serial or HID).
• Interface language: Chinese today; the English build arrives in October 2026.
• Barcode grading is reference-grade only — not a compliance certification (see above).
• Fully offline: no backend, no database server, nothing leaving the machine.

💬 Questions or a custom requirement?

This global site is brand new, and our online tools are currently in Chinese. Tell us what you need and we’ll set up a fast, tailored solution for you — an English version is coming in the next few weeks.

📧 sales@happylab.me · 🌐 global.happylab.me

HappyLab builds practical AI and browser-based tools for industrial device workflows.