How does in-call OCR work?

During a live session the agent captures the frame — or drags a box around just the label — and OCR runs on that region within a second or two.

Region targeting matters: cropping to the plate rather than the whole engine bay dramatically improves accuracy on small, angled, or reflective text.

For equipment audits and multi-unit walks, batch mode queues several captures and returns all extractions together, so a technician reading twenty asset tags does it in one pass.

It is stored on the inspection record next to the source frame, so anyone can verify the reading against the image it came from.

From there it can be copied, included in the exported report, or pushed into your CRM, DMS, or asset system through webhooks and the API.

Clean, well-lit labels read reliably; damaged, dirty, or heavily angled plates need a retake or a manual correction, and the agent can edit any extracted value inline.

Remote zoom and flashlight control exist largely to make these captures readable in the first place.

In-call OCR in Virtual Inspection Pro extracts text from a frame you capture or a region you select during the call. It can bring VINs, serial and model numbers, meter readings, licence plates, and labels into the record, ready to copy, share, or push to your system. Treat that output as a transcription to verify, not proof of the object's identity, condition, or compliance.

Before capturing, ask the participant to stop moving, improve the lighting, and hold the camera square to the label. Keep the full identifier visible, including prefixes, suffixes, and leading zeros. If reflections obscure characters, change the viewing angle rather than guessing. Select a region that includes the relevant field but excludes nearby text that could cause confusion. Never ask someone to enter an unsafe area or remove a protective cover solely to obtain a clearer image.

For VIN OCR video workflows, review the extracted string against the visible VIN before using it for an insurance file or lending record. Pay particular attention to similar-looking characters and cropped ends. NHTSA's VIN Decoder provides vehicle information associated with a submitted VIN; it is a separate reference check, not a Virtual Inspection Pro feature. A plausible decoder result does not establish that the camera showed the intended vehicle.

For serial number extraction inspection tasks, distinguish the model number from the unique serial number and any separate asset tag. Ask the participant to show the whole item, then its identification label, so the reviewer can understand the relationship. Compare the verified identifier with the work order or asset register. If records disagree, flag the mismatch for human review rather than changing the record merely to make it match.

Consider a property team checking a utility meter during a tenant handover. The agent first asks for a wider view showing which meter is being inspected, then a steady close-up of the display and meter identifier. After meter reading OCR extracts the digits, the agent compares them with the image and confirms the unit, decimal position, and displayed register before updating the handover record.

A number alone can be misleading: a digital meter may cycle through several screens, and a label may show a meter serial number rather than consumption. Ask the participant to wait for the required screen and capture it again if necessary. Record the observation date and relevant context through your established documentation process. If the display is damaged, the register cannot be identified, or a reading would drive a disputed charge, refer it to a qualified reviewer instead of treating OCR output as final.

Define who checks extracted text, which destination field receives it, and what happens when characters remain uncertain. Before copying, sharing, or pushing text to another system, confirm the case or property reference and preserve identifier formatting. Avoid collecting unrelated personal information visible beside the target label. Follow your organization's access, retention, and documentation procedures; OCR extraction itself does not establish those controls.

The International Code Council's remote virtual inspection resources provide general guidance for planning remote inspections, not a guarantee that every inspection can be completed remotely. Human review is appropriate for ambiguous characters, conflicting records, and consequential decisions. Arrange an in-person visit when safe access, camera visibility, or the required inspection scope prevents reliable remote assessment. Reading a label cannot replace a physical test, establish concealed conditions, or determine whether equipment meets applicable requirements.