Part IVAutofocus

Subject Detection

October 1, 202620 min readFirmware 2.00

How the Z6III finds faces, eyes, birds and vehicles, which AF-area modes use it, how to switch subjects and eyes, where it fails, and which type to pick for each kind of shooting.

A grey heron stands in the reeds 30 metres away, and you have it at 600 mm on the Nikkor Z 180-600mm f/5.6-6.3 VR. If its eye is about 10 mm across, it spans roughly 34 of the 6,048 pixels across the frame, a little over half of one percent of the picture's width. Holding a focus point on a target that size by hand, while the bird turns its head to watch a fish, is close to hopeless.

The Z6III can try to do it for you. Subject detection means the camera inspects the live picture from the sensor, recognises certain kinds of subject in it, and steers focus onto them, onto an eye where it can find one. It is not on in every focus setup, and it fails in ways you can predict. Page numbers point to the Nikon Z6III Reference Guide, edition 09, which describes firmware 2.00.

01.What Subject Detection Does

Three terms first. Autofocus (AF) is the camera turning the lens's focus for you. A focus point is the small rectangle in the viewfinder marking the part of the scene being focused on. Electronic rangefinding is the help you get when focusing by hand: an indicator in the shooting display that says whether the subject under the focus point is sharp, with the point itself turning green when it is (p. 136).

Subject detection sits on top of both. You pick a class of subject, and the camera gives that class priority when it autofocuses and when it runs electronic rangefinding, marking what it has found with a focus point (p. 127). For living subjects it works down a ladder: the face if it can see one, an eye if it can pick one out, and for animals and birds the whole body when neither is visible (p. 127).

What Nikon Publishes and What It Does Not

Nikon's June 2024 launch release said the Z6III supports nine types of subject detection. One way to reach nine from the Reference Guide's lists is people as one, dogs, cats and birds as three, and cars, motorcycles, trains, airplanes and bicycles as five, since 1+3+5=91 + 3 + 5 = 9. That reconciliation is my reading, not Nikon's breakdown. Firmware 2.00 in August 2025 added a dedicated Birds option, and Nikon's release claims it follows birds in flight against busy backgrounds such as forests and rocky mountains. That is a maker's claim, not a measurement.

Neither the Reference Guide nor these two releases describe how recognition works: not the method, the size of image the detector examines, how often it runs, or how small a subject may be before it is missed.

General Background: How Machines Recognise Pictures

This subsection is about image recognition in general, not Nikon's undisclosed design. A typical recogniser is a program tuned on many example photographs in which people have marked the faces, eyes, birds or cars, and on a new picture it returns boxes, each with a label and a confidence score. It does worse when a scene looks unlike its examples, through an odd pose, a half-hidden face or an unusual colour. And it can only use what reaches it as pixels, so a tiny, washed-out or shadowed subject gives it little to work with. Nikon's cautions read like a list of those two problems.

The Heron's Eye in Pixels

The 10 mm eye is an illustrative figure. A thin lens of focal length ff focused at distance dd scales the scene by

m=fd−f=60030 000−600≈0.0204m = \frac{f}{d - f} = \frac{600}{30\,000 - 600} \approx 0.0204

with both lengths in millimetres, so the eye's image is 10×0.0204≈0.20410 \times 0.0204 \approx 0.204 mm wide. The sensor is 35.9 mm wide (p. 997) and a Large FX picture is 6,048 pixels wide (p. 998), so one pixel is

35.96048≈0.00594 mm\frac{35.9}{6048} \approx 0.00594 \text{ mm}

and the eye covers 0.204/0.00594≈340.204 / 0.00594 \approx 34 pixels. At 60 m, mm falls to about 0.0101 and the eye to about 17 pixels. The thin-lens formula only approximates a real zoom, but it shows the scale: every doubling of distance halves the target. Nikon's first caution for every subject type is a subject, or its face, that fills too much or too little of the frame (pp. 129-130), and it gives no pixel threshold for either, so get closer or zoom in before blaming the camera.

02.The Subject Types

The type you choose lives in one item in each menu. For stills it is this one:

The seven options run from Auto through People, Animal, Birds, Vehicle and Airplanes to Subject detection off, and the photo default is Auto (p. 447). The menu page refers back to the focus section (pp. 127-128), which the table condenses.

Table 1. What each subject type looks for and where the focus point lands

OptionLooks forWhere the focus point goes
AutoPeople, animals and vehiclesWhichever subject the camera picks
PeopleHuman facesThe face, then one eye if eyes are found
AnimalDogs, cats and birdsFace, then eye, else the whole animal
BirdsBirdsFace, then eye, else the whole bird
VehicleCars, motorcycles, trains, airplanes, bicyclesThe vehicle, the front end of a train
AirplanesAircraftBody, nose or cockpit, by size
Subject detection offNothingNo detection

With People, if someone turns away after their face has been found, the focus point moves to keep up with them (p. 127). For aircraft, under Vehicle or Airplanes, the part marked depends on the aircraft's size (pp. 127-128), and Nikon does not say whether that means its real size or how large it appears in the frame. For people the guide describes faces and eyes only, with no separate head or body target, so I will not promise one.

Birds arrived with firmware 2.00, as a new choice in both menus according to Nikon's online manual for that version. The item itself predates 2.00, which is why it carries no firmware badge.

Photo and Video Are Set Separately

Video has its own copy of the item, and the camera keeps a separate subject type for each mode (p. 127), so setting Birds for stills does not set it for video.

It holds two settings. Subject detection chooses the type from the same list, defaulting to Auto (p. 579). AF when subject not detected governs Full-time AF, the video focus mode that focuses continuously without a button press, when no subject of the chosen type is in view (p. 606). With ON, the default (p. 579), autofocus carries on regardless. With OFF, it stops until such a subject appears. The focus modes chapter covers Full-time AF.

Finding It in a Hurry

Both default i menus, photo and video, carry an item called AF-area mode/subj. detection that reaches the AF-area mode and the subject type together (pp. 83-84). The menus and controls chapter covers the i menu. The chosen type also shows as an icon in the monitor and viewfinder displays (pp. 931, 939), so a glance tells you whether you left the camera on Vehicle after the airshow.

03.Which AF-Area Modes Use Detection

An AF-area mode is the rule for how much of the frame the camera may use to focus and who chooses where: a small point you place, a larger box you place, or the camera choosing for itself. The AF-area modes chapter covers each one. Detection only takes part in some of them (p. 128). Two focus-mode abbreviations appear below. AF-S (Single AF) focuses once and locks while the shutter button is held halfway, and AF-C (Continuous AF) keeps refocusing as the subject's distance changes (p. 121).

Table 2. AF-area modes and subject detection during autofocus

AF-area modeDetection during AFNotes
Pinpoint AF, Single-point AFNoSingle-point AF is the default
Dynamic-area AF (S), (M), (L)NoPhoto, AF-C only
Wide-area AF (S), (L)YesA box you place
Wide-area AF (C1), (C2)YesBox size you set
3D-trackingYesPhoto, AF-C only
Subject-tracking AFYesVideo only
Auto-area AFYesCamera picks the point

The first row is the trap. The default AF-area mode is Single-point AF in both menus (pp. 447, 579), and it ignores detection, so a camera fresh from the box has Auto detection switched on and doing nothing. The notes column comes from pp. 124-125.

The second row holds a quieter surprise. The guide suggests Dynamic-area AF (L) for fast birds that are hard to keep in the point (p. 124), but that advice is about area size, and the dynamic-area modes are not among those that use detection (p. 128). To have the camera look for the bird's eye, use a wide-area, tracking or auto-area mode. One more rule applies to the wide-area box: when it holds things at different distances, the closest wins (p. 124). The guide does not say how that rule and detection share the box, so I treat the box as a fence that tells detection where to search.

04.Choosing Between Subjects and Eyes

When several subjects of the chosen type are in view, each gets a grey focus point (p. 128). In Auto-area AF, a pair of icons marks the camera's choice, and pressing left or right on the multi selector moves to another detected subject (p. 128). The multi selector cannot place the point in Auto-area AF (pp. 131, 914), so left and right, or a tap on the monitor (pp. 132-133), are how you steer it. In the wide-area modes you steer by moving the box.

The Eye You Did Not Mean

Switch eyes is a role you can give a button. When the camera has found the eyes of a person or an animal, pressing it changes which eye is used for focus. It is offered for photo controls under f2 (p. 705), for video controls under g2 (p. 750), and for the touch Fn area of the monitor under f4 (p. 730). Roles vary from control to control (p. 704), so choose a button that lists it and that you can reach with the camera at your eye.

Touch is less reliable here. Nikon warns that picking an eye by touch with Touch shutter/touch AF, a photo-only option, may focus the eye on the other side, and suggests the multi selector instead (p. 132).

Why One Eye Matters at f/1.8

Take the Nikkor Z 85mm f/1.8 S wide open, focused at 1.5 m on a head-and-shoulders portrait. Well short of the hyperfocal distance (about 134 m here), total depth of field is close to

DoF≈2Ncs2f2=2×1.8×0.03×15002852≈34 mm\text{DoF} \approx \frac{2 N c s^2}{f^2} = \frac{2 \times 1.8 \times 0.03 \times 1500^2}{85^2} \approx 34 \text{ mm}

where NN is the f-number, ss the focus distance and ff the focal length in millimetres, and c=0.03c = 0.03 mm is the conventional full-frame circle of confusion, the largest blur spot still read as sharp. That convention comes from general optics, not Nikon. The sharp zone runs about 17 mm either side of the focus plane.

Turn the sitter to a three-quarter view. With eyes about 63 mm apart (a typical adult figure, used as an illustration) and the head turned 45∘45^\circ, the far eye sits about 63sin⁡45∘≈4563 \sin 45^\circ \approx 45 mm further away than the near one, well beyond the 17 mm margin. Only one eye can be sharp. Portrait convention, not Nikon advice, says it should usually be the nearer one, and Switch eyes is how you make sure.

05.When Detection Struggles

Nikon prints a caution box for each subject type (pp. 129-130). The reasons given with each are my explanation, not Nikon's.

  • Too small or too large in the frame. Too small leaves too few pixels, as the heron showed. Too large runs the face or vehicle off the edges, so the whole shape is never seen.
  • Too brightly or dimly lit. Glare flattens the detail that separates an eye from a cheek, and shadow buries it in noise.
  • Face or eyes covered. Nikon names glasses, sunglasses and hair for people, fur for animals and feathers for birds. Each hides the very feature being hunted.
  • Face and eyes of similar colour. Listed for animals and birds. An eye is easiest to find when it contrasts with its surroundings, so a dark eye in dark plumage gives little to grip.
  • Too much movement. A turning head or a spinning dancer changes the shape from moment to moment.
  • Vehicles partly hidden or matching their surroundings. Some vehicle shapes and colours may not be detected at all (p. 130).
  • Look-alikes. Borders can appear on things resembling dogs, cats or birds, or on non-vehicles (pp. 129-130). For Birds, Nikon suggests switching to an area mode whose focus points are smaller if false detections are frequent (p. 129), which in practice means a smaller wide-area box.

3D-tracking and Subject-tracking AF add their own list (p. 126): they can lose subjects that blend into the background, change size, colour or brightness, move fast, or are blocked or leave the frame.

With some settings, Nikon says detection performance may drop (p. 130). They are high-speed frame capture + (HSFC+), the release modes C15 to C120 that shoot at 15 to 120 frames per second (p. 156), the HLG photo tone mode, which records high dynamic range (HDR) pictures, and HLG or N-Log, Nikon's flat log recording, in the ProRes 422 HQ 10-bit (MOV) and H.265 10-bit (MOV) video file types. Nikon gives no reason and no size for the drop.

The Birds caution box adds a note on flicker. Under fluorescent, mercury-vapour and similar lamps it is more likely, Photo flicker reduction set to ON reduces it, and Nikon recommends OFF where there is none (p. 129). That matters in aviaries. Flicker reduction itself is covered in the shutter types chapter.

06.A Specific Type or Auto

The guide never says outright when to prefer a named type. It does say that Auto finds humans, animals and vehicles and decides among them itself (p. 128), and that for Birds a narrower search helps against false detections (p. 129). The rest is my reasoning from those two facts.

Every extra class Auto considers is another way to be wrong. On the street a passing bus is a vehicle, and Auto is entitled to prefer it to the person you framed. At a dance show the camera will only ever see people, so searching for vehicles and animals buys nothing. A named type also makes behaviour predictable, which is what lets you time a shot.

Auto finds animals, and Animal includes birds (p. 127), but the guide does not say whether Auto draws on the dedicated Birds detection added in 2.00. I would not assume it does, so if birds are the point of the outing, choose Birds. At an airshow, Airplanes ignores the cars in the car park.

Auto earns its place when you genuinely do not know what will appear, such as a family afternoon where the dog and the children both matter. One aside: auto capture, which fires the shutter by itself when its conditions are met, offers no Birds type (p. 573).

07.Detection While Focusing by Hand

Detection also works in manual focus, the mode where you turn the focus ring and electronic rangefinding guides you (p. 136), but it has its own switch.

The options are All, Wide (L), Wide (S) and MF subject detection off, the default (p. 447). Anything but off turns detection on in manual focus (p. 128). All searches the whole frame, puts a grey point over each subject of the chosen type, marks its first pick with the same pair of icons, and lets you step to another with left or right on the multi selector. Wide (L) and Wide (S) limit the search to the current focus area (p. 128). The type searched for is still the one set in AF/MF subject detection options. Video has its own copy, also off by default (pp. 579, 607):

This helps where autofocus keeps choosing the wrong thing, such as a bird perched behind twigs. Switch to manual focus, set Wide (L), put the area over the bird, and the point can sit on the eye while you turn the ring until it goes green. Two limits apply. The in-focus indicator can light on subjects poorly suited to autofocus when they are not sharp, so magnify the view and check (p. 137). And Nikon rules out touch controls for manual focus (p. 133). The manual focus chapter covers the rest.

These suit the author's kit. Defaults are cited above, and where the guide offers no direct advice the reason is my judgement.

Table 3. Subject detection settings and the reason for each

SettingChoiceWhy
Type for birdsBirdsThe dedicated bird detection
Type for other wildlifeAnimalClosest match, though only dogs, cats and birds are named
Type for stage, portraits, streetPeopleVehicles and animals cannot win
AF-area mode, perched birdsWide-area AF (C1), smallA small box aimed past the twigs
AF-area mode, birds against skyWide-area AF (L), or 3D-tracking once heldA large box is easy to keep on the bird
AF-area mode, birds over treesDynamic-area AF (L), no detectionTracking can lose a bird that resembles the leaves (p. 126)
AF-area mode, portraits and streetAuto-area AFNikon suggests it for portraits and snapshots (p. 125)
AF-area mode, stage danceWide-area AF (C1), tallYou choose the dancer, and stage light can defeat 3D-tracking (p. 126)
Switch eyesTouch Fn for portraits (p. 730)Choose the nearer eye at f/1.8
a11 AF-assist illuminatorOFFNikon's advice for animals and birds
MF subject detection areaWide (L)Help without searching the whole frame
Video type for dancePeopleVideo keeps its own setting
AF when subject not detectedOFF on a gimbal, after a testFocus waits rather than hunts

Birds on the Nikkor Z 180-600mm f/5.6-6.3 VR, in AF-C so focus follows movement. For a perched bird in clutter I use Wide-area AF (C1) set to a small box, aimed at the bird and away from the branches, which is also my version of the smaller-focus-point remedy above (p. 129). Against open sky, Wide-area AF (L) gives a large box that is easy to keep on a moving bird, and once the camera has hold of it, 3D-tracking is the alternative. A bird flying in front of trees is the exception, whatever Nikon's release claims for forests. There I use Dynamic-area AF (L), which Nikon suggests for fast birds (p. 124), and do without detection, since that mode does not use it (p. 128). 3D-tracking is no help there either, because it can lose a subject that resembles its background (p. 126). At long range the eye may be too small to find, and the camera then falls back to the face, or to the whole bird if no face is found (p. 127), which is still usable.

Other wildlife on the same lens. Animal is the nearest match, but a deer or fox may or may not be recognised, and Nikon promises nothing. The same small Wide-area AF (C1) box rather than Auto-area AF means that if the animal goes unrecognised, you are still focusing where you aimed.

Stage dance on the Tamron 70-180mm f/2.8 Di III VC VXD G2. People with Wide-area AF (C1), set to a tall box shaped to one dancer, lets you decide who is sharp in a group. I avoid 3D-tracking on stage, because the light changes colour and brightness from cue to cue, and Nikon lists both among the things that can make tracking fail (p. 126). Auto-area AF can be stepped between dancers with left or right (p. 128), but in a shifting group an aimed box is quicker than a press per change. Expect the lighting cautions to bite when a face hits a hot spot or turns into shadow.

Portraits on the Nikkor Z 85mm f/1.8 S. People with Auto-area AF, the mode the guide suggests for portraits (p. 125), and Switch eyes on Touch Fn (p. 730). With sunglasses on, expect eye detection to struggle (p. 129) and the point to stay on the face.

Street. People, not Auto, with Auto-area AF, so the bus does not win and the camera finds the passer-by while you watch the scene. Nikon names snapshots among the uses of Auto-area AF (p. 125), and if it picks the wrong face, left or right moves it on (p. 128).

Gimbal dance video. Set People in the video menu, with Full-time AF and Wide-area AF (L). For AF when subject not detected my judgement is OFF, so that when a spin loses the dancer, focus holds instead of sliding to the back wall. The cost is that focus will not follow a change of distance while the dancer is undetected, so try both settings at a rehearsal.

09.Beginner Mistakes

Expecting detection in Single-point AF. The camera ships with Auto detection and Single-point AF (p. 447), and Single-point AF ignores detection. Switch area modes before judging the feature.

Assuming video follows the photo setting. Photo and video keep separate subject types (p. 127). Check the video menu, or the display icon, before filming.

Sources and verification

Checked against Z6III firmware 2.00, last verified October 1, 2026.

FeedbackBook mode
z6iiiautofocussubject-detectioneye-detectionbirdswildlifeportraits