A tracking shot is the one camera move that separates “professional” from “AI clip” on first watch. The camera moves with a subject, not toward it and not away from it, but alongside, behind, or in front. Done well it pulls the viewer into the scene. Done poorly the subject morphs, the background warps, or the camera drifts off the rails.
Below are 10 copy-ready tracking-shot templates, each pinned to a specific lens, light, motion, and palette so the model has nothing to invent. They are written in plain English and work across Sora 2, Veo 3.1, Kling, and Runway Gen-4.
TL;DR
- A tracking shot fails when the prompt lets the model guess. Lock the lens, light, motion direction, speed, distance, and palette every time.
- Keep tracking clips to 5 to 8 seconds. All four major models hold consistency in that window; past it, the subject and background start to drift.
- Pick the model for the move: Sora 2 for stylized side and orbit, Veo 3.1 for natural foot pursuit with synced audio, Kling for longer back-tracking single takes, Runway Gen-4 for character consistency across a series.
- Use the 10 templates below verbatim, then swap the subject and palette to fit your scene.
The 5-element checklist for any tracking shot
Every tracking prompt should lock down all five before you hit generate:
- Lens:
35mm anamorphic,50mm prime,24mm wide,85mmfor compression - Light:
golden hour,overcast soft,night practical,dappled forest light - Motion: name the tracking direction (front / side / back / orbit), the speed (walking pace, jog, slow steady), and the distance to subject
- Palette:
teal and orange,desaturated,warm amber,cool blue and grey - 5–8 second restraint: anything longer breaks consistency on tracking moves
The cardinal rule: one subject, one path, one speed. Tracking shots break the moment you stack a second action or a second camera move.
10 copy-ready prompt templates
1. Front-tracking model walk-up
A model in a navy trench coat walks slowly toward the camera down a tree-lined street. Camera tracks backward at walking pace, maintaining 3 meters of distance. 35mm anamorphic lens, golden hour side light, teal and warm amber palette, shallow depth of field. 6 seconds, constant speed, no zoom.
2. Side-tracking parallel car drive
A vintage red convertible drives along a coastal road at steady speed. Camera tracks parallel from the driver's side at 5 meters distance, matching the car's speed exactly. 50mm prime, late afternoon warm light, ocean teal and warm chrome palette. 7 seconds, locked horizon, no shake.
3. Back-tracking subject leaving
A woman in a long ivory coat walks away from camera through an empty stone plaza. Camera tracks forward behind her at hip height, maintaining 4 meters of distance. 35mm anamorphic, overcast cool light, desaturated grey and warm skin palette. 7 seconds, steady gimbal motion.
4. 360-orbit hero center
A standing figure in dark armor holds a sword pointed down at the center of a misty field. Camera orbits 360 degrees around the subject at chest height, constant radius of 4 meters, smooth steady speed. 35mm, low contrast diffused daylight, cool grey and ember orange palette. 8 seconds, no zoom, subject stays centered.
5. Handheld foot-pursuit alley
A runner in a grey hoodie sprints down a narrow rain-wet alley. Camera follows from behind at running pace with slight handheld breathing motion, 2 meters distance. 24mm wide, practical sodium streetlight, deep teal and warm amber palette. 6 seconds, motion blur on edges, no whip pan.
6. Dolly-tracking long corridor
A figure in a white lab coat walks slowly down a long modern corridor lined with glass panels. Camera dollies forward behind the figure on a smooth rail, maintaining 3 meters of distance. 35mm anamorphic, cool fluorescent overhead light, cyan and pale grey palette. 7 seconds, dead-straight path, no rotation.
7. Aerial drone-tracking road
A black SUV drives along a winding mountain road. Drone tracks the vehicle from behind and slightly above, maintaining 30 meters of distance and matching its speed. 24mm wide, golden hour warm side light, forest green and warm asphalt palette. 8 seconds, smooth altitude, no rotation.
8. Steadicam dance-floor
A dancer in a red dress spins slowly on a softly lit dance floor. Camera moves in a steadicam arc around her at chest height, 3 meters radius, smooth continuous motion. 50mm prime, single warm key with cool magenta back rim, deep black background. 7 seconds, no zoom, no shake.
9. Car-rig side mounted
A man in a denim jacket drives an old pickup truck through open desert. Camera is rigged to the driver's side, framing him in medium profile through the open window. Static relative to the truck, but landscape blurs past in the background. 35mm, late afternoon warm light, sandy beige and warm denim palette. 6 seconds.
10. Gimbal hike forest
A hiker in an olive jacket walks up a forest trail through dappled sunlight. Camera tracks from the side at walking pace, weaving slightly between trees to maintain a clear view of the subject. 35mm, dappled warm forest light, deep green and warm khaki palette. 7 seconds, gimbal-stable, no jitter.
5 common mistakes
- Two camera moves in one prompt: tracking plus zoom plus orbit will pick one and break the others
- No distance specified: model may push too close, blowing out the subject’s face
- Speed mismatch:
fast trackingplusslow walkis contradictory and produces drift - No “constant speed”: model adds ease-in/ease-out artifacts you didn’t want
- Path that contradicts the subject’s path: tracking forward while the subject walks toward camera = collision frame
5 push-further moves
- Lock the lens across a series: same anamorphic 35mm in every shot keeps the cut invisible
- Match palette to the time of day: golden hour wants warm amber; overcast wants desaturated cool
- Add a single ambient cue on Veo:
ambient: footsteps on gravel, distant wind - Chain tracking shots via last-frame extraction: see Image-to-Video Prompts
- Cut between front-tracking and back-tracking of the same subject: instant cinematic rhythm
Sora 2 vs Veo 3.1 vs Kling vs Runway for tracking
Each model has a different sweet spot for follow-subject moves. Versions and limits below are current as of June 2026; clip caps especially change often.
| Model (June 2026) | Native clip length | Synced audio | Best tracking use |
|---|---|---|---|
| Sora 2 / Sora 2 Pro | up to ~12s (Sora 2) / up to ~25s (Pro), 720p–1080p | Yes | Stylized side-tracking and orbit, controlled palette |
| Veo 3.1 | 8s standard, extend by chaining ~7s segments | Yes (dialogue, SFX, ambient) | Natural foot pursuit, walking-pace back-tracking |
| Kling 2.5 Turbo / 3.0 | 5–10s (2.5) / up to 15s (3.0, native 4K) | Yes (3.0) | Longer single-take back-tracking, dense scenes |
| Runway Gen-4 | extendable with shot-level camera control | Add separately | Character consistency across a shot series |
How to read it in practice:
- Sora 2: strongest on stylized side-tracking and orbit moves with a controlled palette. Best for cinematic and surreal looks. Holds clean tracking inside its short clips; keep tracking moves to 8s or less even though Pro can run to 25s.
- Veo 3.1: strongest on natural foot pursuit and walking-pace back-tracking, especially with synced ambient audio (the
ambient:line). Standard clips are 8s; chain extensions of about 7s each to go longer. Best for documentary or realism. - Kling: strongest on longer single-take back-tracking and dense scenes. Kling 3.0 runs up to 15s with native 4K and audio; the 2.5 Turbo tier (5–10s) is the cheaper, faster queue. Tends to weaken on close tracking of unfamiliar faces.
- Runway Gen-4: strongest when you need the same character to read consistently across several tracking shots. Its shot-level camera control lets you set framing and movement per shot.
Practical rule: pick the model first, then the template. Forcing a Sora-style orbit on Kling usually wastes a generation. For the current camera-control feature sets, check the vendor pages directly: Google DeepMind Veo and OpenAI Sora.
Per-mood tuning for tracking
- Romantic / nostalgic: slow back-tracking + golden hour + 35mm anamorphic
- Tense / chase: handheld foot pursuit + practical streetlight + 24mm wide
- Epic / scale: aerial drone-tracking + 24mm + desaturated landscape palette
- Product / showcase: orbit + 50mm + single key light + black background
- Documentary / natural: steadicam side-tracking + overcast soft light + muted palette
FAQ
Q: Why does the camera drift off the subject mid-clip?
A: Add camera maintains constant distance to subject and a number — 3 meters. Most drift comes from the model inventing what “maintain” means.
Q: Which model handles tracking best (June 2026)? A: Sora 2 for stylized side and orbit moves; Kling for longer back-tracking takes (3.0 runs to 15s); Veo 3.1 for natural foot pursuit with synced ambient audio; Runway Gen-4 when one character must stay consistent across a series. Test your prompt on at least two before committing credits.
Q: Can I do a one-shot tracking longer than 8 seconds? A: Kling 3.0 holds single takes up to 15s, and Sora 2 Pro runs to about 25s, but tracking specifically tends to drift past 8s on every model. The reliable path is to cut at 8s and chain clips with last-frame continuation. Veo 3.1 extends in roughly 7s segments for the same reason.
Q: How do I stop the background from warping during the tracking move?
A: Slow the camera down and keep the subject centered. Background warp is almost always caused by fast lateral motion relative to the scene. A 50mm or 85mm lens with a clean constant speed instruction warps far less than a 24mm wide push.
Q: Best aspect ratio for tracking shots?
A: Use 2.39:1 for cinematic side and back tracking, 9:16 for foot pursuit and handheld (vertical/social), and 16:9 for general work.
Q: Why do all four models support tracking but give different results? A: They parse the same plain-English motion cues, but each weights them differently. Veo 3.1 and Kling 3.0 follow explicit camera verbs (“tracks parallel”, “orbits 360 degrees”) closely; Runway Gen-4 leans on its shot-level control panel; Sora 2 favors palette and style coherence. Keep the prompt explicit and the differences shrink.
Related articles
- AI Video Camera Movement Prompts
- Cinematic AI Video Prompts
- Image-to-Video Prompt Examples
- How to Improve Motion Consistency in AI Videos
- Drone Shot Video Prompts
Tags: #Camera movement #tracking-shot #Video generation #Prompt