Automatic face and plate blurring can cover most frames in typical footage, but publish-ready redaction often fails in the margins: fast motion, partial occlusions, reflections, dense crowds, or unusual angles. Manual redaction is what turns an “almost safe” export into a defensible release, because it lets teams cover the identifiers that automation misses and stabilize masks across the frames where exposure tends to happen.
Manual redaction in video is the process of adding and tracking masks by hand to hide identifying elements in footage when automated face blurring or license plate blurring does not catch everything. It is a core technique in visual data anonymization used before publishing photos or videos to reduce the risk of identifying individuals or vehicles.
Automated models are strong on clear, frontal faces and well-lit license plates. They are weaker with motion blur, low light, partial occlusions, unusual angles, or when people wear helmets, masks, or sunglasses. Additional objects frequently expose identity in footage, including company logos on clothing, distinctive tattoos, name badges, or text on computer screens. These items are highly context-dependent and are not reliably detected by general models, which is why manual redaction remains a core part of many publishing workflows.
In line with this reality, on-premise software such as Gallio PRO automatically detects and blurs faces and license plates. It does not detect logos, tattoos, name tags, documents, or screens automatically. It does not anonymize entire body silhouettes. It does not perform real-time anonymization of streams. Instead, it provides a built-in editor for fast manual redaction of additional elements within photos and videos.
To avoid repeating the same EU GDPR versus UK GDPR comparison tables across the series, this section focuses on practical compliance anchors and includes US considerations. Under EU and UK frameworks, a face in an image or a license plate in a video can be personal data if a person is identifiable directly or indirectly [1][2]. Regulators highlight necessity, proportionality, and transparency for CCTV and broader video processing, and redaction is widely used as a practical minimization measure when content is shared or published [3][4].
In the United States, there is no single nationwide equivalent to EU GDPR, but publishing or sharing footage that reveals identifiable individuals can still create risk under state privacy regimes, biometric laws where applicable, consumer and employment disputes, and common-law privacy claims. A least-disclosure approach is therefore a practical baseline: share only what is needed, reduce identifiability, and document what you did and why [5].
Obligations for license plate blurring vary by jurisdiction. In many European countries, blurring plates in publicly published images is a common risk-reduction practice. However, it is not accurate to treat it as universally mandatory across Western Europe. Whether it is required depends on context, purpose, audience, and identifiability under applicable rules and guidance.
In Poland the situation is not uniform. EU-level guidance and CJEU jurisprudence support treating plates as personal data when identifiability exists, while some Polish administrative court rulings have indicated that a license plate number may not be personal data in certain contexts. This remains context-dependent and publication risk should be assessed case by case [4][5].
Obligations to anonymize faces may arise from GDPR or UK GDPR where identifiability exists, and separately from national civil-law image-right rules where applicable. In Poland, a commonly cited set of exceptions to the requirement to obtain consent for dissemination of a person’s image under the Act on Copyright and Related Rights includes: a commonly known person photographed in connection with public functions, a person as a detail of a larger whole such as a public event or gathering, or a person who received agreed remuneration for posing. These exceptions are jurisdiction-dependent and their scope must be evaluated in context.
Manual work scales when it is guided by automation and a purpose-built editor. The workflow below starts with automatic face blurring and license plate blurring, then focuses manual effort on edge cases and secondary identifiers, which is typically the fastest path to publish-ready output.
Gallio PRO provides a built-in editor for these steps. It automatically blurs faces and license plates and supports manual redaction of other visual elements. Gallio PRO does not collect logs containing face or license plate detections and does not collect logs with personal or sensitive data.
Manual redaction becomes manageable when teams concentrate effort on the frames that actually drive re-identification risk. The tactics below are commonly used to reduce time without lowering protection quality.
In Europe, publishing images with visible license plates is commonly handled by blurring as a practical risk-reduction approach. Whether blurring is legally required depends on the circumstances, including the purpose of publication, audience, additional context, and whether a person becomes identifiable using reasonably likely means.
In Poland, guidance from EU bodies supports treating plates as personal data when identifiability exists, while certain court decisions have held that plates may not be personal data in some contexts. Because the assessment is context-dependent, many organisations choose to blur plates in all publicly shared visuals to lower risk and avoid disputes, especially for marketing and PR use.
Teams that need repeatable, on-premise processing for both photos and videos can check out Gallio PRO. For specific workflows or procurement, you can contact us.
Visual data anonymization is strongest when masks are stable, sufficiently large, and irreversible. Several checks help prevent edge-case exposure: verify masks on reflections in mirrors and windows, increase blur strength when motion blur creates thin edges, and watch for reappearance after occlusions. Consistency across frames often matters more than the visual style of the mask, because the goal is to make identification impractical with reasonably likely means in context [1][4].
Gallio PRO is on-premise software for anonymizing photos and videos. It automatically blurs faces and license plates. It does not detect logos, tattoos, name tags, documents, or screens automatically, and these can be redacted in manual mode using the built-in editor. It does not anonymize entire body silhouettes, and it does not perform real-time anonymization or video stream anonymization. To evaluate fit and performance in a given environment, you can download a demo.
Manual work is needed when automated detection misses faces or plates due to angles, occlusions, or low light, and when other identifiers appear, such as logos, tattoos, name badges, documents, or on-screen data.
Speed depends on video length, motion, scene changes, and density of identifiers. With pre-detection, tracking, and keyframes, many teams reduce manual time significantly, but actual gains remain context-dependent.
Yes. Gallio PRO processes both and provides one editor and workflow for face blurring, license plate blurring, and manual masks on other visible items.
No. It focuses on faces and license plates and does not anonymize full silhouettes. It does not provide real-time anonymization or video stream anonymization.
In some jurisdictions, there are statutory exceptions to consent for dissemination of a person’s image, including a commonly known person in connection with public functions, a person as a detail of a larger whole such as a public event, or a person who received agreed remuneration for posing. Applicability is jurisdiction- and context-dependent.
Use auto-detection for a baseline, then add manual masks to missed faces or plates and track them. Prioritize close-ups and edge frames at cuts, and export with irreversible blur or pixelation.
No. Gallio PRO does not collect logs containing face or license plate detections and does not collect logs with personal or sensitive data.
