Recognise
Identify items, icons, controls, and visual details.
RO-ER Visum
Give your Mac the gift of sight.
Visum is a line of visual intelligence models, made to run on your Mac. Visum models can recognise items, find things on your screen, follow what changes, and work together to understand complete tasks.
With the Visum Developer app or CLI, you can build, train, test, and publish your own specialised Visum models to do what you specifically need, or fit into your own software using the SDK.
Specialised models for complete visual understanding
The Visum lineup covers a vast array of visual capabilities. Each can be used where it is needed, whether alone, in a solution built using Visum Developer, or through an SDK integration into your apps.
Find
Locate them inside a complete screen or image.
Follow
Notice when they move, appear, disappear, or change.
Understand
Bring the results together into a useful answer.
Act
Use Visum Confector to carry out approved actions on your Mac. Confector is turned on manually and only uses the permissions you grant.
See it in context
Visual intelligence for real tasks
Read a displayed order
See an order panel, identify the requested items, and return a complete structured order.
Monitor a workflow
Follow a visual process, notice each stage, and report when the work changes.
Find complicated controls
Locate buttons, controls, windows, and other elements inside a busy interface.
Recognise specialist objects
Teach Visum about objects, icons, or visual details unique to your work.
Watch for a change
Notice when a visual state appears, disappears, moves, or changes.
Perform an approved action
With permission, Confector can turn a visual result into an action on your Mac.
How Visum works
01
Your screen or video
A visual task begins with what you are looking at.
→
02
Local Visum visual intelligence
Specialist models inspect the visual information on your Mac.
→
03
A useful answer—or an approved action
Visum returns something you can use.
Designed for local visual intelligence
Made for your Apple Silicon Mac
Visum specialist models run locally, so they can process visual information without a permanent internet connection. Your workflow data and Mac information is kept on device.
The minimum required processor is a base Apple M1, running MacOS 14 or higher. All Apple Silicon generations are supported, but Intel Macs are currently unsupported.
- Internet access may be needed for downloads.
- Internet access may be needed for publishing, updates, and cloud conversation.
- Internet access may be required if using the Visum AI Skill. RO Company can not guarantee your data remains on device if using the integration. Please read that company's privacy policy for more information on how they use your data.
The Visum Developer app
Teach Visum what matters to you.
Visum Developer is the complete workspace for creating, training, testing, and packaging Visum solutions. Record a task, mark the important objects and changes, and let Visum Developer prepare the underlying models.
FOR DEVELOPERS
Build Visum into your own software
Use the Visum Developer app or the Visum CLI to build, train, test, and package your models. After which, use the Visum SDK to integrate your Visum model into your app.
Visum Developer workflow
01
Capture
Use the in app tools to easily record or capture your screen, to then annotate.
02
Teach
Mark relevant regions, objects, states, and expected results.
03
Train
Visum Developer creates the required specialist models.
04
Test
Replay unseen examples and check the complete result.
05
Coordinate
The built in tools bundle the specialised models and coordinator model together.
06
Package
Export one finished solution for an app or workflow.
Visum Models Library
Download visual skills created by others
The Visum Models Library will let you download official RO Company models and approved community models, and then install them directly into Visum Player. Compatible models can be opened in Visum Developer for development work where their rights permit it.
The library is not available yet. Every public listing must be approved by RO Company before publishing begins.
OfficialCommunityLuxVexillumPeritusComplete solutions
Visum models
The specialist models inside Visum
Most users do not need to choose these models manually. Visum Developer determines which capabilities a solution requires, and then prepares the right specialists.
Recognise
Visum Lux
Identifies a prepared item or crop from examples.
Locate
Visum Vexillum
Finds multiple visual elements inside a larger screen or panel.
Follow
Visum Peritus
Tracks states, relationships, and meaningful changes over time.
EXECUTOR
VISUM CONFECTOR
Confector transcends the three main Visum editions, and works alongside them using Procurator to control your computer. Confector requires explicit permissions and must be turned on manually.
ORCHESTRATOR
VISUM PROCURATOR
Procurator is the model that is injected into bundles automatically, it understands the objectives and receives inputs from the main models, which it can convert into outputs for Confector to act upon.
Compare the technical details
| Capability | RO-ER Visum Lux | RO-ER Visum Vexillum | RO-ER Visum Peritus |
|---|---|---|---|
| Prepared-crop recognition | Core responsibility | May use Lux-compatible matching | May consume upstream matches |
| Produces a visual embedding | Yes | Optional attachment | May consume embeddings |
| Matches user-defined labelled references | Yes | Optional through Lux compatibility | Not its primary task |
| Finds multiple elements in a larger panel | No | Core responsibility | May consume or extend element observations |
| Bounding boxes and coordinates | No element detection | Core output | Preserved in structured state |
| Segmentation masks or polygons | No | Optional capability | May consume them |
| OCR or visible-text extraction | No | Optional capability | May consume OCR; language support is not implied |
| Full-screen understanding | No | Bounded visual-region analysis | Core responsibility |
| Temporal tracking | No | No | Core responsibility |
| Stable element tracks | No | No | Yes |
| State-change recognition | No | No | Yes |
| Element relationships or scene graph | No | Not a core guarantee | Yes |
| Workflow-stage output | No | No | Candidate capability |
| Natural-language questions | No | No | No |
| Grounded visual answers | No | No | Structured state rather than prose answers |
| Clarification and abstention | Structural rejection only | Errors, warnings, and uncertainty | Unknown or uncertain state handling |
| Proposed plans | No | No | No |
| Clicking, typing, or application control | No | No | No |
| Default cloud requirement | None intended | None intended | None intended |
| Typical runtime package | Model Pack or Reference Pack | Model Pack | Model Pack |
| Current public release status | Not released | Not released | Not released |
Visum Confector technical details
Input
An authorised instruction received through Visum Procurator.
Output
A permitted action passed to the application or computer.
Control boundary
Confector must be enabled manually and limited by explicit permissions.
Visum files
Visum Project
.visumproject
Editable Visum Developer workspace containing a complete solution and potentially several models.
Visum Set
.visumset
Prepared examples used for training, validation, or testing.
Visum Pack
.visumpack
One independently usable trained model or Lux reference pack.
Visum Bundle
.visum
Several separate models packaged with Procurator coordination so they operate as one solution.
Choose how you want to use Visum
The Visum Apps Suite
Visum Player
Run finished Visum models and bundles on your Mac.
Visum Developer
Create, train, test, coordinate, and package your own visual solutions.
Visum AI Skill
Put your agent into a special mode that downloads the Visum CLI and builds your Visum models for you.
Frequently asked questions
What is the difference between Visum Player and Visum Developer?
Visum Player runs finished visual intelligence models. Visum Developer is the creation environment for building, training, testing, coordinating, and packaging specialised models.
Do I need to train models myself?
No. You can run finished models in Visum Player, or create your own in Visum Developer when you need something specialised.
Can I download models that are already trained?
Yes, when the Visum Models Library opens. Every public listing will be reviewed and approved by RO Company.
Can one project contain several models?
Yes. Visum Developer can bundle several specialist models with a Procurator coordination model.
Does Visum send my screen to the cloud?
Visum specialist models run locally. Your workflow data and Mac information is kept on device.
Can I use Visum inside my own application?
Yes. Build and package your model with Visum Developer or the CLI, and then integrate it into your software using the Visum SDK.
What does Procurator do?
Visum Procurator coordinates the objectives, inputs, and outputs of the specialist models inside a bundle.
Can Visum control my computer?
Visum Confector can use the result of a Visum solution to control your computer. It must be turned on manually and only works with explicit permissions.
Which Macs are supported?
Visum requires a base Apple M1 processor or newer and MacOS 14 or higher. Intel Macs are currently unsupported.
What are Visum Packs and Visum Bundles?
A Visum Pack is one independently usable trained model or reference pack. A Visum Bundle contains several separate models with Procurator coordination so they operate as one solution.