Sighthound, Inc.

Sighthound, Inc. AI redaction, license plate recognition, and edge hardware, built for privacy, security, and speed.

A plate reads clean most of the time. When it doesn't — dirt, distance, a bad angle — the vehicle's own details are what...
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

A redacted clip is not weaker evidence. An undocumented one is.Courts don't ask if you redacted. They ask if you can pro...
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

Your gate doesn't need a person. It needs a reading.Most distribution centers lose 30–90 seconds per truck to manual pla...
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

A patrol that passes every few hours misses three revenue killers: ICE-ing, overstays past the charging session, and ses...
08/20/2026

A patrol that passes every few hours misses three revenue killers: ICE-ing, overstays past the charging session, and session limits nobody enforces.

Your cameras already watch these bays.

Point them at an edge engine that reads plates and advanced vehicle attributes, and a session limit stops living on paper. Overstays trigger alerts in real time. Permit mismatches surface automatically. The system runs on your infrastructure, Sighthound provides the detection, you keep the rules, your data stays local.

Swipe for where manual enforcement breaks and what replaces it.

An interview recording has three people in it. One of them is a complainant whose voice can't be released.Until now that...
08/12/2026

An interview recording has three people in it. One of them is a complainant whose voice can't be released.

Until now that meant muting long stretches of the track, or working through the audio by hand and hoping nothing was missed.

Redactor 7.3 adds speaker diarization. It detects and separates who is speaking, lets a reviewer label and correct the speakers, then redacts one person while the rest of the recording stays audible.

How cleanly speakers separate depends on the recording, overlapping voices, background noise and mic placement all affect it, and every result needs review before export. We're refining it over the next few releases and would rather say that now than let anyone find out mid-workflow.

Also in 7.3: a rebuilt speech engine, a redesigned transcript editor, and work time tracking for teams that report review effort.

Details: https://www.sighthound.com/news/redactor-7-3-speaker-diarization

Releasing 10,000 hours of video in 30 days is a capacity problem, and counting files hides it.Three kinds of hours decid...
08/11/2026

Releasing 10,000 hours of video in 30 days is a capacity problem, and counting files hides it.

Three kinds of hours decide the date: footage, machine processing, and human review. Detection finishes fast; reviewers are the real limit, and no faster software fixes a review queue that isn't staffed to the hours.

Teams that hit the date model those hours, pilot on real footage, and take the hardest files first, so bottlenecks surface in week one, not week four.

Redactor runs the bulk workflow across several object classes, on-prem or air-gapped, with your people owning the final release.

Start a free trial: redactor.com/free-trial?utm_source=linkedin&utm_medium=organic&utm_campaign=redactor&utm_content=bulk-video-30-day-deadline

Stop Redacting the Same Footage Five TimesOne busy street frame may contain several kinds of sensitive data: faces, lice...
08/06/2026

Stop Redacting the Same Footage Five Times

One busy street frame may contain several kinds of sensitive data: faces, license plates, IDs, and screens. Most compliance teams use three separate tools to redact them all: exporting, uploading, checking, exporting again. Every handoff is another file leaving your control.

Redactor handles it in one pass.

Smart Redaction automatically detects and redacts seven classes: heads, people, vehicles, license plates, IDs, screens, and documents. One tool. One export. Done.

Security teams: Redactor runs fully offline and air-gapped on Windows, Linux, or Docker. Your video never touches external servers.

Get hands-on for 24 hours. Test it on your own footage, no signup required.

Start free: redactor.com/free-trial

A license plate only helps when the camera can read it. Dirt, glare, a bad angle, and the read is gone.Most vehicle dete...
08/04/2026

A license plate only helps when the camera can read it. Dirt, glare, a bad angle, and the read is gone.

Most vehicle detection treats the plate as the whole identity. When it fails, so does the detection.

Make, model, color, and generation (MMCG) closes that gap. ALPR+ reads the vehicle itself, so a silver Honda CR-V still reads as one when the plate isn't legible.

For parking and enforcement teams, that's a lead instead of a dead end.

Try the online demo and test it against your hardest plates:
https://www.sighthound.com/products/alpr?utm_source=linkedin&utm_medium=organic&utm_campaign=alpr&utm_content=mmcg_missing_detections

Every parking charge, permit, and dispute comes down to one question: which vehicle was there, and when?Most lots answer...
07/30/2026

Every parking charge, permit, and dispute comes down to one question: which vehicle was there, and when?

Most lots answer it with guesswork. Attendants patrol to catch what they can. Drivers claim they paid; there's no record either way. Permits checked windshield by windshield. Payment, permit, and enforcement data never reconcile.

More patrols don't close the gap. The problem was never effort. It's the record.

ALPR+ reads the plate at entry and exit, timestamps everything, flags violations in real time, and backs the plate up with make, model, color, and generation. One camera. Complete trail. No disputes.

Try a free demo: sighthound.com/products/alpr/demo

Can you redact footage before it leaves the platform?" sounds like a feature request. Listen closer and it's a procureme...
07/22/2026

Can you redact footage before it leaves the platform?" sounds like a feature request. Listen closer and it's a procurement question: where does our media live, who touches it, and will the output hold up under a FOIA request?

A detection model answers none of that. Detection is the commodity: heads, plates, IDs, screens, all trained, all shipped. The decisions that actually shape your product are architectural. Does redaction run as a background job? Inside the customer's network? In front of a reviewer who never leaves your UI?

Pick wrong and you've built the right feature on the wrong boundary. Pick right and the regulated deals stop stalling at security review.

Proving it costs you one afternoon: download the Redactor Agent Toolkit, run Docker Compose, and let your coding agent drive the first API call end to end: https://dev.sighthound.com/redactor/reference/agent-toolkit/

Address

587 East State Road 434, Suite 1063
Longwood, FL
32750

Opening Hours

Monday 9am - 5pm
Tuesday 10am - 6pm
Wednesday 9am - 5pm
Thursday 10am - 6pm
Friday 9am - 10pm
Saturday 10am - 6pm

Telephone

+14079745694

Alerts

Be the first to know and let us send you an email when Sighthound, Inc. posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Sighthound, Inc.:

Shortcuts

Share