00:00:00/Guide

Why does CCTV footage look grainy at night?

CCTV footage looks grainy at night because the camera switches to infrared, drops to black and white, and turns its sensor gain right up to pull an image out of the dark; the grain is that amplification made visible. Longer exposures then smear anything that moves, which is why still objects look speckled and moving ones look like ghosts.


If you've just scrubbed to 3am and found a grey blizzard where the driveway should be, nothing is broken. Nearly every camera does this, including expensive ones, because night mode is a stack of compromises rather than a fault. It's worth knowing which compromise causes which kind of ugly, because some of them cost nothing to fix and others aren't worth fighting.

What the camera does when the light goes

At dusk the camera slides its IR-cut filter aside, switches to black and white, and lights the scene with its own infrared LEDs. Colour disappears because infrared carries none. Even under IR the scene is dim, so the sensor turns its gain right up, and gain amplifies the sensor's own electrical noise along with the picture. That noise is the grain. Turn a cheap stereo up to full and you hear hiss; turn a sensor up to full and you see it.

Why moving things smear

The other lever a camera pulls in the dark is a longer exposure. Holding the shutter open gathers more light, so the fence and the parked car look cleaner, but anything that moves during the exposure is painted across every position it passed through. That's why the wall looks fine while the person walking past it is a blur with legs, and why a car crossing the frame can come out as two headlight streaks and nothing else.

The problems the camera didn't cause

The infrared works like a torch strapped to the camera, and torches bounce. A nearby wall, an eave, a hanging basket or a spider's web catches the IR at close range and overexposes into white glare, so the camera dims the whole picture to compensate and the actual scene sinks into black. Point a camera through a window and you get the extreme version: the glass reflects the IR straight back into the lens, and you record a bright smudge of your own reflection instead of the garden.

Compression then eats what's left. Video encoders spend their bitrate on detail and contrast, and a dark noisy scene reads to an encoder like static, so the dark detail that survived the sensor gets averaged into soft grey blocks. A camera left on a low-bitrate setting makes every problem above noticeably worse.

What actually helps, cheapest first

  1. 01

    Clean the dome

    Dust, fingerprints and webs scatter the IR into a general fog, and a spider building across the lens can white out the whole frame. Wiping the dome is the highest-value maintenance a camera ever gets, and most never get it.

  2. 02

    Stop lighting up your own obstacles

    If the foreground glows and the background is black, the IR is bouncing off something close: a wall, an eave, a bracket. Move the camera off the surface it's illuminating, and never shoot through glass; a camera watching a window should be on the other side of it.

  3. 03

    Add ordinary lighting

    A cheap warm outdoor light does more for night footage than an expensive sensor. With enough real light the camera may stay in colour mode all night, and even a little light means less gain, which means less grain.

  4. 04

    Check the bitrate

    Recorders and apps often leave cameras on a low-bitrate stream to save storage. Find the encoding settings and give the camera more to work with; compression hits dark footage hardest, so this is exactly where a starved setting shows.

  5. 05

    Respect the stated IR range

    A camera whose IR is rated to 20 metres can't light a gate 40 metres away, whatever the box art implies. Position cameras so the thing you care about sits inside the rated range, or light it separately.

Night footage will never look like day

It's worth being honest about the ceiling. Gain and long exposure are physics, not settings you can menu your way out of, and the improved version of your night footage will still be a grainy black and white picture. That's usually fine, because the question night footage exists to answer isn't “was it pretty”. It's “was someone at the gate at 3am”, and a grainy frame answers it.

That's also the question an AI can take off your hands. Svid has a vision model describe frames in plain English as footage arrives (a person, a vehicle, a door standing open), and those descriptions are what you search, so “person at the gate” jumps to the 3am moment without anyone scrubbing for it. For fast events there's a Precise watch level that takes a frame every 1 second at 1080p, which is how a car that crosses the drive in three seconds still lands on several frames. None of this sharpens the picture. It reads what the picture already shows.

Why is my security camera black and white at night?
Because it's lit by its own infrared, and infrared carries no colour information. The camera moves a filter aside at dusk and switches to a mono image on purpose. If colour comes back in daylight, everything is working as designed.
Can software fix grainy night footage?
Not honestly. Denoising can smooth the speckle, but it can't add detail the sensor never captured, and anything that appears to is inventing it. The practical moves are more light at the scene, and asking the footage questions it can still answer rather than trying to beautify it.
Why is part of my night image glowing white and the rest black?
The IR is reflecting off something close to the lens: a wall, an eave, a web, or a window you're shooting through. The bright patch drives the exposure down and the rest of the scene goes dark. Move the camera away from whatever it's lighting up and the exposure settles.
Will a more expensive camera fix the grain?
Partly. A bigger sensor needs less gain, so the grain shrinks, but it doesn't repeal the physics. A modest outdoor light usually improves night footage more per pound than a sensor upgrade, which is an unglamorous answer but a cheap one.

Ask the footage, don't squint at it

Create an account at app.svid.ai, add the camera or upload the ugly 3am clip, and search for the moment you care about. The picture stays grainy. The question still gets answered.

Get started

Related footage