Dash cams · legibility
How far can a dash cam read a plate at night?
Thirty feet, at best. That is the answer across nine cameras measured under one method, and four of the nine could not resolve a number plate after dark at any distance at all. It is a long way from what the boxes imply, and it is the only figure on a dash cam that decides whether the footage is worth anything when you need it.
What actually happens at night
MotorTrend put cones at ten-foot intervals in front of a car, parked a second car at each interval, and recorded the distance at which its plate stopped being readable. They ran it in daylight and again at night, with the same cars and the same plate, on nine cameras. That is a plain test, and as far as we can find it is the only published one that reports the answer as a distance rather than as a photograph.
Daylight is comfortable: 30 to 65 feet, and every camera managed something. At night the range collapses to between nothing and 30 feet — roughly two car lengths. Past that, on every camera tested, a plate is a bright smear.
Nine cameras, one plate, two lighting conditions
Scale runs to 30 ft, the highest measured value here.
use it MotorTrend, Best Dash Cams for 2026: 9 Tested Picks Ranked by Real-World Performance, 3 April 2026. Cones set at 10 ft intervals ahead of a 2026 Subaru Outback, a second car parked at each interval, and the distance recorded at which its number plate became indecipherable. Run once in daylight and once at night, nine cameras, same car, same plate. Setup time was taken on a stopwatch from opening the box to a mounted, updated, correctly aimed camera.
| Product | Day range ft to unreadable use it | Night range ft to unreadable use it | Setup min:sec use it | Recording claimed marked | Where to buy |
|---|---|---|---|---|---|
| Rove R2-4K Dual Pro our pick Starvis 2 sensor | 60 | 30 | 6:29 | 4K, 30 fps | Check on Amazon |
| Garmin X310 | 65 | 25 | 4:22 | 4K, 30 fps | Check on Amazon |
| Garmin Dash Cam Live | 45 | 20 | 2:56 | 1440p, 30 fps | Check on Amazon |
| Garmin X210 | 40 | 20 | 3:26 | 1440p, 30 fps | Check on Amazon |
| Nextbase 622GW | 55 | 20 | 15:05 | 4K, 30 fps | Check on Amazon |
| Cobra SC 250R 4K, and still nothing at night | 50 | none | 10:15 | 4K, 30 fps | Check on Amazon |
| Cobra SC 220C Front camera records at 1080p | 40 | none | 7:26 | 1440p, 30 fps | Check on Amazon |
| Nextbase Piqo 2K | 35 | none | 8:28 | 2K, 30 fps | |
| Nextbase 322GW | 30 | none | — | 1080p, 60 fps |
use it MotorTrend, Best Dash Cams for 2026: 9 Tested Picks Ranked by Real-World Performance, 3 April 2026. Cones set at 10 ft intervals ahead of a 2026 Subaru Outback, a second car parked at each interval, and the distance recorded at which its number plate became indecipherable. Run once in daylight and once at night, nine cameras, same car, same plate. Setup time was taken on a stopwatch from opening the box to a mounted, updated, correctly aimed camera.
marked MotorTrend, Best Dash Cams for 2026 — listed specifications, 3 April 2026. Recording resolution, frame rate and field of view as published by each manufacturer and reprinted in the test's tables.
Resolution does not predict night range
Sort that table by resolution and it does not sort by night range. The clearest case is a pair: the 4K Cobra SC 250R could not read the plate at night at any distance, while the 1440p Garmin X210 managed 20 feet. The Cobra records 2.25 times as many pixels per frame and read the plate zero times.
Two of the four cameras that failed completely after dark were recording at 4K or 2K. Five cameras reached 20 feet or more, and two of those five were 1440p — so 1440p appears at both ends of the result and 4K appears at both ends of the result. Resolution is the number on the box, on the listing and in the search results, and on this evidence it does not sort the cameras by the thing you are buying one for.
That comparison is our arithmetic over MotorTrend's published figures, not a claim they make. Their measurements are the sourced part; the ordering and the conclusion are ours, and you can check both against the table above.
Why the plate is the hard part
A number plate is retroreflective. It is engineered to throw headlight light straight back at its source, which is exactly where your dash cam is sitting. MotorTrend's own explanation of the failures is that the plates were overexposed — too much light, not too little. The white background blows out and takes the characters with it.
That reframes the whole purchase. The camera's job at night is not to gather more light; it is to hold a bright retroreflector and a dark street in the same frame. That is dynamic range and exposure control, and it is why a Starvis 2 sensor with good HDR beats a higher pixel count. The best night result in the test came from the Rove R2, and MotorTrend attributes it to that camera's Sony Starvis 2 sensor rather than to its resolution. The Rove records at 4K — and so does the camera that read nothing at all after dark, the Cobra SC 250R.
The other measured number: setup time
The same test timed installation from opening the box to a mounted, updated, correctly aimed camera. The spread is larger than anything in the image quality: 2:56 for the fastest, 15:05 for the slowest, and one camera was abandoned after twenty minutes of firmware installation without finishing.
A five-fold difference in setup is not a rounding error, and it is entirely absent from every product listing. Two of the three Garmins were under three and a half minutes and the third took 4:22; the Nextbase 622GW took a quarter of an hour.
What nobody has measured
- Rain. Nobody has published a legibility distance through a wet windscreen, which is the condition that matters most and the one we would most like to have.
- Moving plates. The test measured parked cars. A plate on a car moving away from you at speed is a harder problem, and no published figure exists for it.
- The same test twice. One run, one night, one plate, one car. Nobody has repeated it, so we cannot tell you how much of the spread would survive a second run.
Where a page needs a figure that does not exist, this site says so rather than reaching for the manufacturer's. That is the whole method.