Editing Fitness Video: Form Checks, Follow-Alongs and Everything Between

Open your clips and put each one in one of three buckets. Demonstration: a clean rep or three of a movement, shot to be understood. Correction: footage where something is going wrong and the point of the video is to show what. Session: continuous work you intend somebody to train along to.

The bucket determines the pace, and the pace determines everything else. Demonstration runs at real speed with maybe one slow repeat. Correction lives between 0.25x and 0.5x and stops moving entirely at the moments that matter. Session runs at exactly 1x for its whole length, without exception, because the viewer is moving in real time and any compression breaks the contract.

This is worth being strict about. Speeding up a follow-along is not a stylistic choice, it is a broken product, and the viewer finds out about 40 seconds in when they are behind. If a 45 minute session is too long for where you are posting it, the answer is to post a shorter session, not a faster one.

1. 01Label every clip as demonstration, correction or session before editing.
2. 02Never mix a correction and a session in the same video.
3. 03Set the project frame rate to match your highest frame rate footage, so slow motion has frames to spend.
4. 04Cut each bucket as its own project even when the source footage overlaps.

When you are showing somebody what is wrong, you are not showing them a squat. You are showing them one instant inside a squat: the frame where the heel lifts, the frame where the bar drifts forward, the frame where the lower back rounds. At full speed that instant is two frames long and nobody sees it, including people who already know what they are looking for.

So a correction video has a shape. Play it once at full speed, so the viewer sees the movement as it actually looks. Play it again at 0.25x through the section where the fault appears. Hold on the worst frame for about a second. Then show the corrected version, again slowed, with the same camera position, so the two are comparable.

That last part matters more than the slow motion. A comparison only works if everything except the thing you are comparing stays the same. Same angle, same distance, same speed on both reps. Move the camera between the wrong rep and the right rep and you have shown two videos, not one correction.

Slow motion also has a floor. Below about 0.25x, footage shot at 30 frames per second turns into a slideshow, because there are no intermediate frames to show. If you know a movement is going into a breakdown, film it at 60 or 120 frames per second. Vidmoat setSpeed goes down to 0.05x, but no editor invents frames that were never captured. The decision that makes a smooth 0.2x replay possible is made at the camera, not on the timeline.

Every coach has drawn a line on a video. The problem is that the thing you are drawing on is moving. A static arrow pointing at where the knee was is worse than no arrow at all, because it is pointing at empty air by the time the viewer reads it.

This is what motion tracking is for and it is genuinely the difference between a coaching video and a video with graphics on it. trackMotion tracks a region of a clip and sticks another clip to it, so a dot follows the bar path, a highlight ring stays on the trailing knee, a label saying heel stays on the heel through the whole descent. The annotation moves with the body, which is the only way the viewer knows what you are pointing at.

Some practical guidance on picking the region to track. Choose something with contrast and hard edges: a shoe, a plate on the end of the bar, a sleeve cuff, a wristband. Avoid tracking bare skin in a bright gym, and avoid anything that leaves the frame or gets occluded, because a tracker that loses its subject usually snaps somewhere embarrassing. If a limb passes behind the torso, track before and after as two separate attachments rather than fighting one long track through the occlusion.

Restraint is the other half. One annotation per point. If you need three, you need three cuts. A frame with an arrow, a ring, a line and two labels is a frame where the viewer reads nothing.

1. 01Pick the single visual point that proves what you are saying.
2. 02Track a high-contrast object near it, not the skin itself.
3. 03Attach one marker or label, and let it live only for the seconds it is relevant.
4. 04Cut before you add a second annotation, rather than stacking them.

There is no camera position that shows a squat. From the side you can see hip depth, spine angle and bar path, and you cannot see whether the knees are collapsing inward. From the front you can see knee tracking, foot pressure and symmetry, and you cannot see depth at all. Every experienced coach knows this, and most fitness videos are still shot from one position.

The fix costs nothing: film the demo rep twice, once from each angle, in the same set. In the edit, show the movement front-on for one full rep, then cut to the side for the next rep, at the same point in the movement so the cut lands cleanly. A cut placed at the top of the rep, where the body is still, is invisible. A cut placed mid-descent is a jump.

If you have two cameras running at once, createMulticamGroup syncs the angles from the audio itself, using cross-correlation rather than asking you to line up a clap, and switchAngle cuts between them. Even without a clap, a dropped dumbbell or a rack bang usually gives the sync enough to work with.

There is one exception worth knowing. Some movements really do have a single correct angle, and hip hinge patterns are the clearest case. A deadlift, a Romanian deadlift or a good morning is a side-on movement, and a front angle adds nothing except a view of a belt. Do not add a second angle for its own sake.

A follow-along workout is a timing product. The viewer is doing the work, watching intermittently, and relying on the video to tell them where they are. Every editing decision serves that and nothing else.

The rule that fixes most follow-along edits: cuts land on interval boundaries. Where a work period ends and rest begins, where rest ends and the next exercise starts, where a round ends. Never in the middle of a rep. A mid-rep cut makes a viewer who glanced away think they missed something, and glancing away is the normal state of somebody doing burpees.

The efficient way to build this is to lay the clock down first and cut to it. addMarkers puts down a whole grid of markers at once, so a 40 seconds on, 20 seconds off circuit becomes a marker every 20 seconds across the entire timeline before you have made a single cut. Then sliceClip cuts on those exact times. You are no longer eyeballing where each block ends, you are cutting to a structure that is already correct.

Transitions here should be almost invisible. A quick dissolve or a fast cut between exercises, nothing more. A whip pan or a glitch between exercise four and exercise five is a stylistic flourish landing at the exact moment somebody is trying to work out whether to pick up a dumbbell.

1. 01Write out the interval structure before editing: work, rest, rounds.
2. 02Lay markers across the timeline at every boundary in that structure.
3. 03Slice on those marker times, so no cut can land mid-rep.
4. 04Keep transitions to a fast cut or a short dissolve throughout.

A countdown on a follow-along is not a graphic, it is the interface. The viewer glances at it while gasping. It has to be legible from across a room, at an angle, on a phone propped against a water bottle.

That means big, high contrast, and in a position that never moves. Numbers that are readable are worth more than numbers that are pretty. Use tabular figures so the digits do not shuffle sideways as the count changes, keep the timer in the same corner for the entire video, and put the current exercise name next to it rather than somewhere else on screen, because those two pieces of information get read together.

Set counters and round indicators follow the same logic. What round am I in, how many rounds are left, what is this exercise called. Three facts, one consistent place, and no animation beyond the number changing.

addHtmlElement is the right tool here because these are custom, structured, repeated graphics rather than one-off text. You write self-contained HTML and CSS and it becomes a clip on the timeline. Build one timer card and one exercise card, then reuse them across the whole session with the values swapped. That consistency is most of what makes a workout video look professionally produced.

<div style="
    display:flex; align-items:center; gap:24px;
    padding:20px 28px; border-radius:18px;
    background:rgba(12,14,18,0.72);
    font-family:Inter, system-ui, sans-serif; color:#fff;
  ">
    <div style="
      font-size:96px; font-weight:700; line-height:1;
      font-variant-numeric:tabular-nums; letter-spacing:-0.02em;
    ">0:40</div>
    <div>
      <div style="font-size:34px; font-weight:600;">GOBLET SQUAT</div>
      <div style="font-size:22px; opacity:0.7;">ROUND 2 OF 4</div>
    </div>
  </div>

Three practical problems come up in almost every gym shoot, and all three are cheaper to solve while filming than in the edit.

Mirrors first. A commercial gym is a hall of mirrors, and a mirror in frame gives you a second, backwards copy of your subject, other members who did not agree to be filmed, and a reflection of you holding the camera. Shoot at an angle to the glass rather than square on, or use the mirror deliberately as your second angle, which is the one trick a mirror-heavy gym gives you for free: film side-on with the mirror showing the front view, and one camera position yields both angles in the same shot. Cropping in during the edit will remove a mirror at the edge of frame, but it costs resolution, so it is a repair rather than a plan.

Muted autoplay next. Most feeds start playing without sound, so the exercise name has to be on screen. Not narrated, on screen, within the first second or two of the movement appearing. If your video is coached rather than silent, run word-level captions over the whole thing so the instruction survives the mute, and let the music sit under the voice rather than over it. autoDuck rewrites the music volume keyframes wherever the voice plays, so cueing does not fight the track.

Vertical crop last. Gym footage is unusually hard to reframe, because a barbell movement is wide and a jump is tall and both get clipped by a 9:16 frame. applyAutoReframe writes per-clip keyframes so the crop follows the subject rather than sitting fixed, which handles a lunge walking across the frame or a sled push travelling the length of the turf. It will not save a wide barbell shot where the plates need to be visible on both sides. For those, either frame tighter when shooting or accept the horizontal version.

Fitness content is one of the few genres where the correct amount of editing is sometimes almost none. A follow-along should feel like a room you are training in, not a montage. If a viewer notices your transitions during a set, the transitions are wrong.

Save the energy for the places it belongs: a session intro, a promo cut, a highlight of a client hitting a lift. Those want beat-driven cutting, and snapToBeat retimes clips onto the music grid with native tempo detection, so a 15 second promo lands its cuts on the beat instead of near it. Inside the actual workout, that same technique is an irritation.

Music volume deserves one specific decision. In a follow-along, the music is a metronome and a mood, and it should sit at a level where a viewer can hear their own breathing and your cues over it. If you find yourself raising your voice on the recording to compete with the track you chose, the track is too loud in the mix.

One honest note on audio repair. Gym rooms are loud, reverberant and full of clanging, and cleanup is limited: the noise reduction available is a high-pass filter, which takes some low rumble out and does nothing about a dropped plate. If you are coaching to camera, a lapel mic or a phone in your pocket recording a separate track will do more for your video than anything that happens afterwards.