Cognex VisionPro: PC-Based Vision Software, Versions and Selection
VisionPro vs In-Sight: PC-Based Vision or Smart Camera
VisionPro vs In-Sight: when a plant should choose PC-based vision or a smart camera, compared on throughput, cameras, custom code, upkeep and IT lifecycle.
VisionPro vs In-Sight is the choice between running vision on a computer you maintain and running it inside a camera that Cognex sells as one device. For most single-station inspections in a plant, we would start with the smart camera and move to the PC route only when a specific requirement forces it. This page sets out the requirements that force it, on the same dimensions for both sides. It is a child of the VisionPro page and sits in the Cognex technology hub. Facts are as of November 2026, from Cognex documentation and releases.
VisionPro vs In-Sight: the decision in one picture
VisionPro vs In-Sight on the same dimensions
| Dimension | In-Sight (smart camera) | VisionPro (PC-based) |
|---|---|---|
| Where it runs | In the camera, with the processor inside the device | On a Windows PC, with cameras attached |
| Cameras per processor | One camera per device | Several cameras can share one PC |
| Camera choice | The In-Sight family | Cameras the PC route supports, including GigE Vision devices |
| Custom code | Within the tool environment (EasyBuilder or Spreadsheet) | Programming library, with Designer as a graphical option |
| Who typically maintains it | Trained plant or controls staff | A vision engineer or developer |
| Licensing | Cognex licences in the camera software | Cognex security dongle on the PC |
| Controls integration | Industrial Ethernet built into the camera | Through your application code |
| Platform to patch | The camera firmware and Vision Suite | Windows, drivers, GPU, dongle and the application |
| Deep learning | Edge AI tools on the camera | VisionPro Deep Learning on a GPU |
These are our general characterisations, consistent with how Cognex positions its products. Check each cell against current Cognex documentation. See In-Sight smart cameras and PLC integration for the camera side.
How each driver plays out in a plant
Throughput. Both routes can be fast enough for most packaging lines, and neither is faster by nature. Cognex publishes In-Sight 3800 frame rates of 125 frames per second at 1.6 megapixels and 18 at 16 megapixels in mono, so resolution costs speed on the camera route. Cognex also documents a trigger interval of at least 1.5 times the job execution time for the 3800 (a 23.1.1 known issue), which is the useful sizing rule: measure how long your job takes and set the trigger rate from that. On the PC route, speed depends on the processor, the network and whether a GPU does deep learning. In our view neither route replaces a test with your parts at your line speed.
Multi-camera. This is the PC route’s strongest argument. If one product needs six views, a single PC serving six cameras can replace six processors. The trade is that one failed PC takes six views down together, so the PC needs a recovery plan. Our reading of the 2026 In-Sight 6900 vision controller, which Cognex launched on 28 April 2026 with an NVIDIA Jetson module, is that Cognex is also offering a PC-free route for heavier loads. We could not verify its camera count, so confirm before relying on it.
Custom code. If the application needs logic that tool-based configuration cannot express, such as complex part tracking, database lookups or an unusual operator workflow, the programming library route fits. The cost is that the application becomes software you own. Our test: if you cannot name who will read this code in five years, do not write it.
Maintainability by plant staff. A smart camera job is a file edited with a Cognex tool. A trained controls technician can often change a tolerance. In-Sight documentation shows the camera route adding plant-facing features such as audit logging (Vision Suite 24.3.0), HTTPS and elevated access (25.1.0), and MQTT connectivity and FTP or SFTP server (25.3.0) (release history). A PC application has whatever its builder wrote.

IT lifecycle. A PC brings an operating system with its own support dates, a GPU and drivers, and a dongle. The VisionPro Deep Learning 4.2 notes show the coupling: a minimum GPU capability and driver version, no support for Windows Server 2016, and a known issue with a Windows update. Treat the PC as a controlled asset with a patching policy agreed with IT. The OT network assessment service is where that placement is usually worked out.
Cost of ownership drivers
Cognex does not publish VisionPro or In-Sight prices, so we give drivers, not numbers.
| Driver | Smart camera route | PC-based route |
|---|---|---|
| Hardware | A camera per station | Cameras plus an industrial PC, possibly with a GPU |
| Licences | Software and tool licences per camera | Dongle and tool licences per PC |
| Engineering | Configuration time | Configuration plus application development |
| Spares | A spare camera per type | Spare camera, PC image, dongle and drives |
| Upkeep | Firmware changes | Patching, driver control and application support |
| Validation (regulated sites) | The camera version and job | The whole PC stack and the application |
Two illustrative cases
Illustrative case one: a label check on a filling line. One camera looks at a pack at a fixed station. The result is pass or fail, sent to the PLC, and a controls technician will adjust tolerances at changeover. Every answer in the decision tree points to the smart camera: no special camera, no need to share a processor, plant staff maintain it, and the line has no spare space for a PC. Choosing the PC route here would add a computer and a licence part for no benefit.
Illustrative case two: a six-view inspection on a pharmaceutical-style packaging machine. Six cameras inspect different faces of a carton, with results merged into one decision and one record per unit, and a vision engineer at the integrator writes the combination logic. The first two questions point to the PC route. The cost is that the PC and the application must be validated and controlled as a unit. A written patching policy, a backup of the application and its dongle, and a named owner for the code are conditions of choosing it, not extras.
Questions to ask before choosing
- How many cameras does each product need, and could one fail without stopping the others?
- Is there a camera, sensor or 3D device we need that only the PC route supports?
- Who will change a job at changeover, and what training do they have?
- Who owns any custom code, where is the source kept, and who can rebuild it?
- Which Windows version and patch policy will IT agree to on the line?
- What is the plan for a failed PC and a failed dongle?
- What has Cognex stated in writing about the lifecycle of the software we pick?
Where it gets hard
- The smart camera route can hit a ceiling. Many cameras per product, or logic the tool environment cannot hold, forces a rethink late in a project.
- The PC route hides cost in labour. Hardware is the visible price. The recurring price is patching, rebuilds and the person who understands the code.
- Vendor positioning changes. We found no Cognex statement on Designer’s future or an end-of-life notice for VisionPro, and Designer releases are tied to specific VisionPro versions (the 4.5.0 notes state compatibility with VisionPro 9.21). A choice made on today’s lineup should survive both.
- Mixed estates are common. Many plants run both, and the cost is two skill sets and two upgrade paths.
If you want an independent opinion before a purchase, our owner’s engineer service is built for that. Related reading: edge solutions in industrial automation.
Related: VisionPro · PatMax · In-Sight smart cameras · Cognex hub
Frequently asked questions
Should I use VisionPro or In-Sight?
Use In-Sight when one or two cameras per station and a PLC handshake cover the job and plant staff will maintain it. Consider VisionPro when one computer must serve many cameras, the application needs custom logic, or the camera or 3D hardware is only available on the PC route. The deciding factor is usually who maintains it, not image quality.
Is a PC-based vision system faster than a smart camera?
Not by default. Speed depends on the image size, the tools and the processor. Cognex publishes frame rates for the In-Sight 3800 of 125 frames per second at 1.6 megapixels and 18 at 16 megapixels in mono. A PC can offer more processing, but you must size it, and a fast PC with a slow network gains nothing.
Does PC-based vision cost more than a smart camera?
It depends on the number of cameras and the lifetime. A PC system adds the computer, Windows, a Cognex dongle licence and engineering time. A smart camera system repeats the camera at every station. Cognex does not publish VisionPro prices, so get written quotes for both and include upkeep, spares and validation.
Can a plant maintain VisionPro without a developer?
It depends on how the application was built. Standard tools set up by a vision engineer can be maintained by trained staff. A custom application needs someone who can read its code and rebuild it. In our view, if no one in the plant or at the integrator will own the code for years, choose the smart camera.
Is there a middle option between a PC and a smart camera?
Cognex launched the In-Sight 6900 vision controller on 28 April 2026, powered by NVIDIA Jetson technology, and states up to 157 TOPS of AI performance, as a PC-free option that runs In-Sight software. We could not verify how many cameras it supports or its other details, so confirm with Cognex before treating it as a multi-camera answer.
References and further reading
- VisionPro PatMax RedLine documentation (VisionPro 9.5) · Cognex documentationAn example of the VisionPro 9.x documentation set.
- Cognex Deep Learning release notes, version 4.2 · Cognex documentationPC requirements: GPU, driver and Windows support for the deep learning side.
- Cognex launches In-Sight vision controller powered by NVIDIA · Cognex via PR NewswireThe 6900 as a PC-free option.
- In-Sight Vision Suite release history · Cognex documentationWhat In-Sight cameras gained release by release.
- Machine vision standards · EMVAGigE Vision, GenICam and the standards behind multi-vendor cameras.