Sighthound, Inc.

10/19/2026

Most records backlogs come down to arithmetic.

An hour of bodycam footage can take hours to redact by hand. Someone has to scrub through the video, covering faces and plates as they appear, then check the audio for names and addresses. Multiply that by dozens of pending requests, and a small records team can fall months behind.

To the public, that delay can look like an agency withholding footage. Inside, it's often a handful of people working through more video than they can reasonably handle.

Automatic redaction takes much of that repetitive work off their hands. Redactor detects faces, plates, IDs, and screens across footage, while reviewers check the results, correct anything missed, and handle sensitive audio. It also maintains audit logs of processing activity, giving teams a record to reference alongside their release documentation.

Records requests come with deadlines. The public sees whether those deadlines are met, even when they can't see the work behind them.

Try the hosted demo, link in the bio.

10/14/2026

A license plate only helps when the camera can read it.

Dirt, glare, or a bad angle, and the plate read is gone. So is the match that depended on it.

Sighthound Retriever, powered by ALPR+, also reads make, model, color, and generation (MMCG). With the plate unreadable, you still get a vehicle profile to work from, like a silver Honda CR-V.

It narrows the list. Your team makes the call. Works with fixed, in-vehicle, and dashcam cameras, at the edge or on your own server.

Request a technical fit review: https://www.sighthound.com/products/alpr?utm_source=linkedin&utm_medium=organic_social&utm_campaign=alpr&utm_content=mmcg_missing_detections

10/12/2026

A camera records whatever is in front of it. At an incident, that often includes a family on their porch, a kid waiting for the bus, and whoever wandered over to see what was going on. None of them did anything except be nearby.

If that footage goes out in a public records release, those people can end up exposed unless someone covers their identities first. Their afternoon overlapped with a police call, and that's the only reason they're in the file.

This part of redaction is easy to undervalue because it doesn't help close a case. What it does is keep an agency's duty to the public from turning into a privacy problem for people who never asked to be involved.

Redactor finds and covers faces, plates, IDs, and screens across a clip, giving reviewers a chance to check and adjust the results before export. It handles audio too, with searchable transcripts that help reviewers find and redact sensitive spoken details, including addresses.

Try the hosted demo: www.redactor.com/free-trial

10/02/2026

It’s Cybersecurity Awareness Month, and one security question deserves more attention:

Where does the data behind the camera actually go?

When video and plate reads are pushed into a centralized environment, that system becomes a more valuable target. More data in one place can mean more exposure if that environment is compromised.

Edge processing changes that architecture.

With Sighthound Retriever, analysis runs on agency-controlled hardware. There’s no required cloud and no default central pool of plate reads. Data can stay within the agency’s own environment, governed by its own retention, access, and sharing policies.

Passwords, access controls, and patching still matter. But security also starts earlier, with system design.

A useful principle is simple:

Don’t centralize data you don’t need to centralize.

Talk to our team about edge deployment:
sighthound.com/products/hardware

09/30/2026

Two questions decide most of your privacy posture, and neither is on a typical spec sheet: where is my data processed, and whose laws reach it there?

If a plate read or a piece of evidence video has to travel to a server somewhere to become useful, then wherever that server sits, and whoever has legal authority over it, is now part of your data's story. A great privacy policy can't override a court with jurisdiction over the company holding the data.

Processing at the edge changes the answer. The work happens on the device, in your environment. The data doesn't take a trip to somewhere with different rules. It stays where your policies actually apply.

For a public agency handling sensitive footage, that's the difference between controlling your data and hoping the vendor's arrangements hold.

Talk to our team about edge deployment:

Run AI workloads like ALPR+ on the edge. Sighthound’s compute node and smart camera deliver fast, offline performance for field deployments.

09/24/2026

A month ago we published how we think vehicle recognition should be built: processed locally, data held by the agency, retention it controls, and no central network pooling everyone's movements.

Since then the rules have started catching up, fast. A federal bill aimed at license plate surveillance. A governor ordering statewide safeguards. An industry body publishing responsible-use principles. The throughline across all of it is the same short list: retention limits, purpose limits, access controls, and audit trails.

We're not going to pretend any single vendor is exempt from where this lands.

The whole category is about to be held to a higher bar, and that's fair. But a system already built around local control and customer-held data doesn't have to scramble when those requirements arrive, because the requirements describe how it already works.

Regulation was never the threat to responsible vehicle recognition. It's the thing that finally rewards it.

Read our approach:

Sighthound publishes its approach to responsible vehicle recognition: edge processing, local control, and redaction that keep ALPR's public-safety value while addressing privacy, data-sharing, and retention concerns.

09/10/2026

License plates can be misread, obscured, or missing, and still return a hit.
When it does, you have a string match, not a vehicle match.

That's the difference between a plate reader and a vehicle classification system.
One flags a string. The other verifies the vehicle.

Make, model, color, and generation (MMCG) don't come off with a screwdriver.
When they match the alert, the read holds. When they don't, the false hit gets filtered
before anyone rolls on it.

That's how Retriever turns a plate lead into a confirmation.

Learn more about this application: https://www.sighthound.com/products/alpr
Or read how vehicle classification works: https://www.sighthound.com/blog/vehicle-classification-models-mmcg

09/04/2026

A plate reads clean most of the time. When it doesn't — dirt, distance, a bad angle — the vehicle's own details are what let someone resolve it without pulling footage.

Make, model, color, and generation (MMCG) don't come off with a screwdriver. They stay in the frame.

That's what Retriever reads, powered by ALPR+. From a single camera, running on the edge or your own server.

For parking enforcement, access control, and fleet work, MMCG on cars and trucks is what keeps the record usable when the plate can't be trusted.

Request a technical fit review: https://sighthound.com/products/alpr?utm_source=linkedin&utm_medium=organic_social&utm_campaign=alpr&utm_content=mmcg_highway

08/28/2026

A redacted clip is not weaker evidence. An undocumented one is.

Courts don't ask if you redacted. They ask if you can prove what you redacted, and why. Authentication is not one technical test. Rule 901 asks: is this what you say it is? A court weighs custody, metadata, and hashes. A redacted file still holds when the original is preserved and the source-to-export path is documented.

Redactor is the redaction step: auto-detect seven object types, redact heads not faces (no identification), run offline and air-gapped, export a full record for the packet.

The technical side is redaction. The legal side is documentation.

Try the hosted demo: https://www.redactor.com/free-trial?utm_source=linkedin&utm_medium=organic_social&utm_campaign=court_ready_workflow&utm_content=redactor_carousel

Or read how courts authenticate video: https://www.redactor.com/blog/how-courts-determine-audio-video-evidence-authenticity

08/21/2026

Your gate doesn't need a person. It needs a reading.

Most distribution centers lose 30–90 seconds per truck to manual plate checks. Someone walks out, reads the plate, checks the list, waves the driver through. The barrier is fast. The identification step isn't.

ALPR+ moves that step to the approach. A fixed camera reads the plate plus make, model, color, and generation (MMCG) as the truck rolls up, all in real time, at the edge. It checks your watchlist or scheduling system through the API before the cab reaches the arm.

Recognition runs on-device. Plate data never leaves your site.

The stop becomes a formality.

Try it on your gate: https://sighthound.com/products/alpr/demo?utm_source=linkedin&utm_medium=organic_social&utm_campaign=alpr_plus&utm_content=social_post