Upload a ride. Get a report. Download what you need.
Three steps for the cyclist. Seven steps for the pipeline. You only have to do the first one.
Upload
Drag in a clip, or your whole SD card dump. Uploads resume if your wifi blips. Your card can be wiped the moment the upload finishes.
Auto-analyze
The pipeline finds every vehicle, measures every pass, identifies the close calls, and reads the plates. A notification lands when it's done.
Download
Open the ride, pick the incident, click Download. A police-ready PDF, annotated frame, distance, vehicle, plate, GPS, timestamp.
How close did they actually get?
Upload a ride and find out to the inch. Free, 120 minutes a month, no card.
The pipeline, step by step.
What “auto-analyze” actually does, in seven stages. No setup, no calibration, no manual review on your end.
Upload your ride
Drop in any clip from a forward, helmet, chest, or rear-facing camera. We probe the file, pick the right camera and mount preset automatically, and queue it for analysis.
Find every vehicle
A computer-vision detector runs across the footage and tracks each vehicle through the frames it appears in. Cars, trucks, motorcycles, vans, anything that crossed your camera gets picked up and followed.
Measure how close
For every tracked vehicle, we estimate the side gap to the rider in feet. The math uses the road surface as a reference and adapts to your camera and mount, so the same number means the same thing whether you ride with a handlebar GoPro or a seatpost cam.
Filter the noise
Parked vehicles you rode past and oncoming traffic that wasn't actually close get classified out. What's left is the set of vehicles that genuinely passed you, bucketed by clearance into Severe, Near, and Safe.
Identify the vehicle
For every non-safe pass we run a vision model that returns make, model, year range, color, and body type. The model picks the best angle it has, not just the closest frame, so distant and side-on shots still get a clean read.
Read the plate
Plate reading uses several candidate frames and takes the consensus, so a single blurry frame can't hallucinate characters. When the plate isn't legible from any angle, we say so, no guesses.
Build the report
The ride lands as a timeline with a chapter for each pass. Click a chapter to seek the video; toggle vehicle boxes to outline each detected vehicle and its measured clearance on top of the footage; open analytics for the totals. Every non-safe pass is one click from a downloadable incident PDF.
On your footage
One toggle puts the measurement on the video.
Every vehicle the pipeline tracked is outlined where it actually appeared, with the class it was identified as and the clearance it left you. Nothing is painted on the road that wasn’t measured on a vehicle.
The pipeline does the calibration for you. You don’t measure anything, mark anything, or tell us where the road is.
No GPS on your camera? Drop in your phone track.
Helmet cams, chest cams, dashcams — most don’t record GPS. Doesn’t matter. Drop in a Strava export or any GPX from your phone, and we line it up with the video timeline second-for-second.
Every incident gets pinned to the exact spot on a map. Every report gets an exact street address and the wall-clock time the pass happened. Officers don’t have to take your word for any of it.
Upload your phone track
Strava GPX, Apple Health route, Garmin Connect, or any standard GPX file.
We align by timestamp
Your video’s recording time and the GPX track are matched to the second, no manual nudging.
Map view + exact addresses
Every incident gets reverse-geocoded. The incident PDF gains a map, a pin, and a street address.
Be on it the day it ships.
Start free now, phone-track sync reaches every account the day it lands, no plan change required.
The questions riders ask first.
- Do I need to calibrate my camera or mark the road?
- No. The pipeline probes your file, picks the right camera and mount preset automatically, and uses the road surface itself as its reference. You never measure anything, mark anything, or tell it where the road is, so the same clearance number means the same thing on a handlebar GoPro or a seatpost cam.
- What cameras does it work with?
- Any forward, helmet, chest, or rear-facing action cam. Drop in a single clip or your whole SD card dump; uploads resume if your wifi blips, and your card can be wiped the moment the upload finishes.
- How does it measure how close a pass was?
- For every tracked vehicle it estimates the side gap to you in feet, using the road surface as a reference and adapting to your camera and mount. Passes are bucketed by clearance into Severe, Near, and Safe, and parked vehicles you rode past and oncoming traffic get filtered out.
- Can it identify the vehicle and read the plate?
- For every non-safe pass it returns the vehicle's make, model, year range, color, and body type, and reads the plate from several candidate frames by consensus so a single blurry frame can't invent characters. When the plate isn't legible from any angle, it says so rather than guessing.
- What do I get to give the police?
- Every non-safe pass is one click from a police-ready incident PDF: an annotated frame with the measured distance, the identified vehicle, the plate, GPS, and a timestamp. The full ride also lands as a timeline with a chapter per pass, and every detected vehicle can be outlined over the footage with its measured clearance.
- What if my camera doesn't record GPS?
- Phone-track sync (coming soon) covers that: drop in a Strava export or any GPX from your phone and it's aligned to the video second-for-second, so every incident is pinned to a map with an exact street address. It rolls out to every account at launch, with no plan change required.