Part VIIVideo

Video Autofocus and Stabilisation

October 1, 202620 min readFirmware 2.00

How the Z6III focuses and steadies video: focus and AF-area modes, subject detection, AF speed and tracking, focusing by hand, and both kinds of VR.

A dancer runs from the wings to centre stage in about two seconds. For a photograph, focus has to be right for one instant, the 1/500 s the shutter is open, and a burst that misses can be deleted frame by frame. At 30 frames a second, a video camera has to be right 2×30=602 \times 30 = 60 times in those same two seconds, and the audience sees all 60. If the lens slides off her face to the backdrop and back, nobody has to look for the mistake. It plays.

Video autofocus (AF) therefore has a different brief from photo autofocus. Smooth matters more than fast, and a wrong move is worse than no move. This chapter covers the video focus items, the Custom Settings that shape how focus moves, the aids for focusing by hand, and the two ways the Z6III steadies a moving picture, one of which costs part of the frame. Page numbers in parentheses point to Nikon's Z6III Reference Guide, edition 09, which matches firmware 2.00.

01.Why Video Focus Is a Different Job

A video is a run of still frames shown quickly enough to read as motion. The rate is counted in frames per second (fps) and written with a p, so 30p means about 30 whole frames each second (29.97, to be exact, p. 202). Every focus mistake is stretched across frames. Suppose the lens takes 0.4 s to drift off a face and return, a figure chosen for illustration. At 30p that is 0.4×30=120.4 \times 30 = 12 soft frames in a row, and at 60p it is 0.4×60=240.4 \times 60 = 24. In general, a focus event lasting tt seconds, recorded at rr frames per second, fills NN frames:

N=t×rN = t \times r

That drift has a name. Hunting is the camera, unsure which direction is sharper, driving one way, finding the picture softer and reversing. Nikon lists conditions that make autofocus struggle, among them subjects lacking contrast, subjects smaller than the focus point, and lights that change brightness or flicker (p. 122). In a clip, a hunt is part of the clip.

The speed of a focus change is part of the picture too. A slow, even shift from a dancer's hands to her face reads as a choice, the move camera operators call a focus pull. The same shift done in two frames reads as a jolt. Then there is sound: Nikon warns that autofocus and vibration reduction can be heard in the footage (p. 109), and the lens grows louder as focus speed rises (p. 760). A microphone clipped to a performer is far from that noise, and the camera's built-in one is not, as Audio and Timecode explains.

Last, video focus has to run without a finger on a button, because a camera on a gimbal, a motorised handle that keeps it level while you move, may record for four minutes with both hands on the handle.

02.Video Focus Modes

The video recording menu has a Focus mode item of its own (p. 604), which I read as holding its own value, though Nikon does not say so outright. Page 121 describes the four choices:

  • Single AF (AF-S) focuses on a half-press of the shutter-release button, then locks while you hold it.
  • Continuous AF (AF-C) keeps adjusting while you hold the half-press or the AF-ON button.
  • Full-time AF (AF-F) keeps adjusting with no button pressed, following the subject and any change of framing. A half-press locks focus and turns the focus point from red to green. It exists only in video.
  • Manual focus (MF) leaves focus to you and the focus ring.

The default is Full-time AF (p. 579), and it is my choice for the reason Focus Modes gives: on the gimbal no finger is free for a button. Off the gimbal, holding the half-press lock stops a passer-by from stealing focus.

The others have their shots. Continuous AF suits a shot where you decide when focus moves: hold AF-ON as the dancer walks forward, let go when she stops, and focus stays put while someone crosses in front. Single AF and Manual focus suit a locked-off landscape, where focus should never move once set.

03.AF-Area Modes in Video

The AF-area mode decides how much of the frame autofocus considers. Video has its own item, defaulting to Single-point AF (p. 579), and a shorter list than photo. Pinpoint AF, the dynamic-area modes and 3D-tracking are photo only, and Subject-tracking AF is video only (pp. 123-125). AF-Area Modes covers the modes the two menus share.

Nikon points to wide-area AF for smooth focus during pans, tilts and moving subjects in video, and Wide-area AF (L) covers more than Wide-area AF (S) (p. 124). The custom shapes in Wide-area AF (C1) and Wide-area AF (C2) offer 66 sizes in video with the FX image area (FX is Nikon's full-frame format), from 1 by 1 to 21 by 11 focus points (p. 125). One rule matters on a stage: when the area holds things at different distances, the closest gets priority (p. 124). A microphone stand at the lip of the stage is nearer than the face behind it. Subject detection is the counterweight, giving the chosen kind of subject priority (p. 127). Nikon does not say how the two priorities rank, and I expect a detected face to win.

Subject-tracking AF follows one subject. Place the focus point on it and start with the OK button, AF-ON or a half-press. The OK button ends tracking and returns the point to the centre (p. 125). On the touch screen, Move focus point and Touch AF start tracking on whatever you touch (pp. 132-133). Tracking can lose a subject that resembles its background, changes visibly in size, colour or brightness, moves quickly, or disappears behind something (p. 126). My choice is Wide-area AF (L) for a stage, and Subject-tracking AF when one performer dances alone.

Limiting the Dial Choices

Custom Setting g4 trims the AF-area modes you step through when you hold a control given the Focus mode/AF-area mode role in g2, the video button roles, and turn the sub-command dial (p. 759). Every mode is ticked by default (p. 628), and the default video map puts that role on Fn2 (p. 627). My Fn2 is Power aperture (close) in video, as Custom Controls sets out, so g4 has nothing to act on. I change AF-area mode from slot 5 of the video i menu (the quick menu on the i button), AF-area mode/subj. detection (p. 627).

04.Subject Detection in Video

Video keeps its own subject type, so Birds for stills does not mean Birds for video (p. 127). There are seven choices: People, Animal, Birds, Vehicle and Airplanes for particular subjects, Auto to let the camera pick, which is the default (p. 579), and Subject detection off. Detection works in the wide-area modes, Subject-tracking AF and Auto-area AF, not in Single-point AF (p. 128). With People, the camera frames a face and moves to an eye when it finds one (p. 127). Subject Detection covers each type.

Nikon warns that People can falter when a face is too large or small in the frame or the subject moves excessively (p. 129), and a spinning dancer qualifies. Detection may also weaken when HLG (Hybrid Log Gamma) or N-Log (Nikon's flat logarithmic curve) is recorded in the ProRes 422 HQ or H.265 10-bit file types (p. 130), which N-Log, HLG and Exposure Tools explains.

The second half of the item is AF when subject not detected, which governs Full-time AF (p. 606). With ON, the default (p. 579), autofocus runs whether or not a subject of the chosen type is in view. With OFF, it stops until one appears (p. 606). I use People with OFF. With ON and no face in view, I expect the wide area to fall back on its closest-subject rule and drive to whatever is nearest, often the stage floor. With OFF, autofocus sits idle while no face is found (p. 606), so the lens stays where it was until one returns. The cost is that focus will not follow a change of distance while the face is hidden, so I compare both settings at a rehearsal, as Subject Detection advises.

05.How Fast Focus Moves

AF Speed

Custom Setting g5 sets how quickly focus moves in video. The default is 0 (p. 628). Higher numbers are quicker and louder, and the page names +5 as the setting where lens noise is most noticeable, advising lower values if it distracts (p. 760). The g5 screen in Nikon's online manual shows the whole scale, running in whole steps from −5-5 at the slower end to +5+5 at the faster end.

When to apply decides when the speed holds. Always, the default (p. 628), applies it throughout video mode. Only while recording applies it during recording and lets the camera focus as fast as it can at other times (p. 760). That fits video: fast focus while framing up, the gentler move during the take. I choose Only while recording at 0, and if rehearsal footage showed focus snapping between dancers I would try a lower value.

AF Tracking Sensitivity

Custom Setting g6 runs from 1 to 7, with 4 as the default (pp. 628, 761). At 1 (High) the camera is quick to move on: if the subject slips out of the focus point, it refocuses on whatever else occupies that part of the frame. At 7 (Low) it clings to the subject it started with (p. 761). It is the video cousin of the photo setting a3 in The Focus Custom Settings.

Picture a dancer who crosses in front of the soloist for half a second, 0.5×30=150.5 \times 30 = 15 frames at 30p. A high sensitivity may hand focus to the crossing dancer and back, two moves the audience sees, and a low one is more inclined to wait. I use 5, one step toward Low, whenever several dancers share the stage, whether the camera is on a tripod or the gimbal. For a solo, with nobody to cross in front, I leave the default 4.

Fast AF-ON

The AF-ON and Fast AF-ON roles both start autofocus, and they differ in Continuous AF: AF-ON focuses at the g5 speed, while Fast AF-ON focuses at full speed whatever g5 says (p. 750). Back-Button Focus introduced the role.

It suits a reader filming in Continuous AF with g5 slowed for gentle pulls: AF-ON on the AF-ON button for the slow move, Fast AF-ON on a second control for a subject that must be sharp now. Nikon describes the difference only for Continuous AF, not Full-time AF, so my map leaves the AF-ON button at AF-ON and assigns Fast AF-ON nowhere.

06.Focusing by Hand

Manual focus in video works as for stills: put the focus point on the subject and turn the focus or control ring (p. 136). Magnification is the first aid. The zoom-in button enlarges the view, and during recording it steps through 50%, 100%, 200% and 400%, though 50% is missing at 1920 by 1080 and zoom is unavailable while recording RAW video, the unprocessed sensor data (p. 110). Next is the focus indicator, which Nikon calls electronic rangefinding: a mark showing whether focus lies on the subject under the point, in front of it or behind it, with the point lighting green at focus (pp. 136-137). With subjects autofocus finds hard, Nikon warns, the in-focus mark can appear when focus is not quite there, so magnify and check (p. 137).

Focus peaking draws coloured outlines on the sharpest edges as you turn the ring, and is switched on with a12 Focus peaking > Focus peaking display (p. 137). It may not appear on a subject without clear outlines (p. 137). Slot 11 of my video i menu holds Focus peaking, and Focus Limiter, Saved Positions and Manual Focus covers its settings, its clash with zebra stripes and the focus-ring setting that makes a pull repeatable.

Subject Detection While Focusing by Hand

Detection can keep working in Manual focus, since the chosen subject type gets priority in electronic rangefinding as well as in autofocus (p. 127). The default is MF subject detection off (p. 579). All searches the whole frame, marks each subject with a grey point and lets you step between them with left and right on the multi selector. Wide (L) and Wide (S) search only the current focus area (p. 128). The type is the one set under AF/MF subject detection options. I choose Wide (L) to match my AF area, so a switch to manual focus on stage still puts the point on a face.

Half-Press to Cancel Zoom

Half-press to cancel zoom (MF) added in Firmware 2.00, Custom Setting g18, lets a half-press return a magnified view to the full frame during manual focus in video when it is ON (p. 778). The default is OFF (p. 629). I set it ON, as I set its photo twin d19 in Focus Limiter, Saved Positions and Manual Focus: magnify, focus, half-press, and the whole frame is back for a last look before recording.

07.Two Ways to Steady a Picture

A still is one slice of time, too short for a hand's tremor to blur at a fast shutter speed. Video shows frames one after another, so a tremor too small to blur any single frame still shows as the picture wobbling.

The Z6III has two answers. Mechanical vibration reduction (VR) moves hardware to cancel the shake: the camera shifts its sensor on five axes, and a lens with VR adds its own moving glass (p. 1002). The whole frame is kept. Vibration Reduction explains both for stills. Electronic VR moves nothing: the camera records a smaller window of the sensor and shifts it from frame to frame, so the picture is cropped.

Vibration Reduction for Video

Nikon names one option for this item, Same as photo settings, which makes video use whatever the photo Vibration reduction item currently holds (p. 608). It is also the default (p. 579). Nikon names no other, though the menu list on p. 861 notes that this video item's options vary with the lens. The photo item offers Normal, Nikon's choice for still subjects, Sport, for subjects moving fast and unpredictably, and Off (p. 498). Nikon recommends Sport for panning, and in Normal and Sport the camera leaves the pan itself alone and steadies only the shake across it, so a sideways pan is steadied up and down only (p. 498).

Which lenses offer which? Nikon prints no table, saying only that the options vary with the lens and that VR may be unavailable with some (p. 498). One case is spelled out: an F-mount lens with its own VR switch, fitted through the FTZ II or FTZ adapter, greys the item out, and the lens switch takes over (p. 498). Of my lenses, the Nikkor Z 85mm f/1.8 S has no VR of its own and relies on the sensor shift, while the Nikkor Z 180-600mm f/5.6-6.3 VR adds lens VR. Whether the Tamron 70-180mm f/2.8 Di III VC VXD G2, whose stabiliser Tamron calls VC, offers the same menu choices is documented by neither maker. Link VR to focus point does not apply while recording, when the correction is tuned for the centre of the frame (p. 499).

My video VR stays at Same as photo settings, as Vibration Reduction set it. The catch is that video inherits the last photo choice: after stills on Sport, video gets Sport, which suits pans, and after tripod work with Off, video gets Off.

Electronic VR

Electronic VR is OFF by default (p. 579), and ON makes the video crop smaller (p. 211). Nikon's announcement for the camera puts the result at the view of a lens about 1.25 times longer. Take the Tamron 16-30mm f/2.8 Di III VXD G2 at 16 mm: 16×1.25=2016 \times 1.25 = 20, so the shot looks as if taken at 20 mm. The recorded window is 1/1.25=0.81/1.25 = 0.8 of the full width and height, so it keeps 0.8×0.8=0.640.8 \times 0.8 = 0.64 of the area, and 1−0.64=0.361 - 0.64 = 0.36, just over a third of the frame, becomes margin. On the FX video crop of about 35.9 by 20.2 mm (p. 211), that is roughly 28.7 by 16.2 mm, my arithmetic rather than Nikon's figure.

A grey rectangle, the full FX video frame of about 35.9 by 20.2 mm, contains a smaller white window with a thick blue border, about 0.8 of its width and height, holding a sketch of a dancer. Two dashed blue copies of the window sit slightly offset inside the grey margin.
Figure 1. Electronic VR records a window about 0.8 of the frame's width and moves it within the margin to cancel shake. Proportions follow Nikon's 1.25 times figure, and the offsets are illustrative.

It brings limits too. The highest ISO, the sensitivity setting, falls to ISO 51200 with Electronic VR on (p. 209). The camera holds it at OFF with either RAW file type, ProRes RAW HQ 12-bit (MOV) and N-RAW 12-bit (NEV), at 5376 by 3024, at 100p, 120p, 200p or 240p, with the 1920 by 1080 slow-motion option, with Hi-Res Zoom on, and while auto capture is running (p. 609). Video Formats and Codecs covers those formats. Both stabilisers sit in the default video i menu, in slots 7 and 8 (p. 627), and Nikon does not say how they divide the work when both are on.

I leave Electronic VR OFF on a stage, where a further crop would tighten every framing, and on the gimbal, whose motors already steady the picture. I turn it ON for handheld walking shots with a wide lens at 4K 30p (3840 by 2160 pixels at 29.97 fps, p. 202), where 16 mm becoming 20 mm costs little.

On the DJI RS 5

DJI rates the RS 5's tested payload at 3.0 kg, its motors turn on pan, tilt and roll, and its compatibility page lists the Z6III. I keep the 180-600 off it, because that long, heavy telephoto plus the camera would use most of the payload, a budget HDMI and External Monitors adds up item by item. DJI's RS 5 FAQ describes a camera control button on the gimbal whose half-press starts autofocus and whose tap starts or stops recording. Under Full-time AF, though, the camera's own half-press locks focus (p. 121), and neither maker says what a gimbal half-press does to a Z6III in that mode. I expect it to lock focus too, so try it at home before a performance.

Neither Nikon nor DJI documents how the camera's VR behaves while the gimbal's motors are also correcting, so what follows is inference. The gimbal handles large, slow turns, and sensor shift and lens VR are built for small, quick tremor, so the two should mostly act on different motions. A deliberate move is where they could disagree, and Nikon's note that Normal and Sport leave the motion of a pan alone (p. 498) suggests the camera is designed for that. My video VR stays at Same as photo settings. If a rehearsal clip showed a drift or bounce at the end of each pan, setting the photo item to Off for that session would show whether VR is the cause.

Table 1. Video focus and stabilisation settings and the reason for each

SettingChoiceWhy
Focus mode (video)Full-time AFNo button needed
AF-area mode (video)Wide-area AF (L), or Subject-tracking AF for a soloSmooth focus over a large area
Subject detection (video)PeopleFaces first, then eyes
AF when subject not detectedOFF for dance, ON for walking shotsRests when a dancer is lost, runs where no one is
MF subject detection area (video)Wide (L)A face is still marked in MF
g4 Limit AF-area mode selectionAll ticked, the defaultNo video control uses that route
g5 AF speed0, Only while recordingGentle in a take, quick between takes
g6 AF tracking sensitivity5 with several dancers, 4 for a soloWaits out a dancer crossing
g18 Half-press to cancel zoom (MF)ONQuick return to full frame
Fast AF-ONNot assignedDefined for AF-C, and I film in AF-F
Vibration reduction (video)Same as photo settingsFollows the photo item
Electronic VROFF, ON for handheld wide shotsThe crop and its limits

Stage dance. With the Tamron 70-180, everything as in the table, with g6 at 5 while several dancers share the stage. For a solo I switch to Subject-tracking AF, start it before the music and return g6 to 4.

Gimbal dance. The same, g6 included: 5 when the gimbal follows several dancers, as at a recital, and 4 for a solo. Test what the gimbal's half-press does to focus first.

Landscape and street. Manual focus with magnification and peaking on the tripod. Handheld walking shots use Full-time AF with Electronic VR ON, and AF when subject not detected back at ON, since OFF would leave autofocus idle in a scene with no face (p. 606).

09.Beginner Mistakes

Leaving AF when subject not detected at ON on a stage. With no face in view the lens goes looking, and the nearest thing is often the floor. Set OFF and compare both at a rehearsal.

Turning AF speed to +5 because faster sounds better. Focus jumps instead of gliding, and the lens noise grows (p. 760). Start at 0 with Only while recording.

Forgetting that video VR follows the photo item. With Same as photo settings the last stills session decides. Check the photo item before you film.

Expecting Electronic VR at a high frame rate. It is fixed at OFF from 100p up, and with RAW video and slow motion (p. 609).

Trusting the in-focus mark alone in manual focus. Nikon warns it can appear when focus is not quite there (p. 137). Magnify, or watch the peaking.

Sources and verification

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

FeedbackBook mode
z6iiivideoautofocussubject-detectionmanual-focusvibration-reductiongimbal