Stated plainly: engineering drawings often carry a customer’s product structure. This page sticks to facts — which actions are local, which need a network, which are metered and which you opt into.
Preview, dimension detection, ballooning and export all run on your Windows machine.
It only runs when you select it for a job, it is metered per run and usage is visible.
Scope: the TuMo Vision Windows client and the official website. The client UI may change between versions, but the data boundary principles do not.
The table below lists processing location and network requirements per function. The "drawing content uploaded" column describes whether the drawing itself leaves your machine.
| Step | Processed on | Drawing uploaded | Metered |
|---|---|---|---|
| Import and preview | Your machine | No | No |
| Dimension detection (YOLO / CPU / GPU) | Your machine | No | Consumes detection quota |
| OCR extraction (local mode) | Your machine | No | Consumes detection quota |
| OCR extraction (cloud mode) | DeepSeek vision model (cloud) | Yes, in this mode only | Metered per run |
| Balloon numbering and leader lines | Your machine | No | No |
| Export ballooned PDF / Excel / report | Your machine | No | No |
| Drawing revision comparison | Your machine | No | No |
Outside cloud OCR, drawing content does not leave your machine. Sign-in, quota lookup and updates need a network connection but do not involve drawing content. If your compliance review needs item-by-item confirmation, contact us for a written statement. Cloud recognition is provided by DeepSeek, a third-party model provider, with visible scope and usage.
Saying clearly when data leaves the machine is more useful than repeating the word "secure".
All detection and export run on your machine and work offline.
No setup required — this is the path after installation.
When local OCR is not stable enough for certain drawings, you can choose cloud OCR per job (currently the DeepSeek vision model).
Choose the cloud model (DeepSeek) in the OCR settings; if unsure, evaluate with local mode first.
Not a security whitepaper — four statements about product behaviour.
These describe product behaviour, not a third-party security certification. We do not claim certifications and will not borrow certification marks; where compliance documentation is needed, we provide written material for your scenario.
The scope changes the data handling. Validate with the default setup first, then decide whether you need a more closed deployment.
Local detection and export for individual engineers and small teams.
200 free detections, then monthly or annual membership.
Multiple devices are quoted by device count and concurrency, with shared account and activation management.
We confirm usage scale and main drawing types first.
When drawings cannot leave the intranet, we can evaluate private OCR and offline models.
Assessed and quoted per environment.
No. Preview, dimension detection, OCR extraction (local mode), ballooning and export all run locally. Data leaves your machine only when you choose cloud OCR .
Local detection, review, numbering and export work offline. Sign-in, quota lookup, updates and cloud OCR need a network connection.
Cloud OCR currently uses the DeepSeek vision model (DeepSeek-Flash by default, switchable per drawing type in the client). It needs a network connection and a signed-in account, and is metered per run. If third-party recognition is not acceptable, use local mode or contact us about intranet and offline options.
Per run, metered. You can see the scope before a job starts and check usage in your account; if you do not choose cloud OCR, no such charge is generated.
Data compliance depends on your environment: where drawings come from, what the intranet allows, what your customers require. Send us the question and we answer for your scenario, not with a generic script.
Weekdays 9:00 – 18:00 CST, email answered within 24 hours.
Downloading the trial requires no drawings and no setup. Validate detection and export locally, then decide the next step.