Machine-translated from the Chinese original; the Chinese version is authoritative. 中文原文
This is a searchable, copy-paste master table of camera move commands: when writing AI video prompts, look up terms by category and copy the English phrasing directly. Each term comes with a “when to use” note. Suitable for creators building their own prompt templates, and for veterans transitioning from live-action to AI to align terminology.
How to use this table
- In prompts, use the English terms (shown in code style in the table); the Chinese column is for communication and lookup—don’t mix languages when issuing commands;
- Use only 1–2 camera move terms per prompt, paired with speed and direction words (see “Usage discipline”);
- Actual response of each term across AI models: unverified (to be filled in through testing later); this table only guarantees accuracy of meaning.
Master table
Push/Pull
| Chinese | English | When to use |
|---|
| Push in | push in / dolly in | Focus attention onto the subject, build emotional intensity, or reveal detail |
| Pull back | pull back / dolly out | Move away from the subject to reveal the full environment; a sense of detachment at scene close |
Pan/Tilt
| Chinese | English | When to use |
|---|
| Pan left/right | pan left / pan right | Camera stays put, pans horizontally: establish lateral space, or track a subject moving across |
| Tilt up/down | tilt up / tilt down | Camera stays put, tilts vertically: tilt up for grandeur, or follow a person from face to hands |
Truck/Track
| Chinese | English | When to use |
|---|
| Lateral move | truck left / truck right | Entire camera moves sideways with parallax between foreground and background: sweep across a scene, move parallel with a subject |
| Tracking shot | tracking shot | Move synchronously with a moving subject at a fixed distance: longer narrative passages |
| Follow shot | follow shot | Follow from behind the subject: strong sense of immersion |
Crane/Pedestal
| Chinese | English | When to use |
|---|
| Crane up | crane up | Rise from eye level to a high angle: closing establishing shot, showing scale |
| Crane down | crane down | Descend from above into the scene: opening entry into the action |
| Pedestal up/down | pedestal up / pedestal down | Vertical shift without changing tilt angle: only changes eye height, smoother than a crane move |
Orbit/Arc
| Chinese | English | When to use |
|---|
| Orbit | orbit | Circle the subject 90°–360°: product showcase, emphasize three-dimensional volume |
| Arc shot | arc shot | Move along an arc around the subject (often ≤90°): more dimensional and less stiff than a straight lateral move |
Special
| Chinese | English | When to use |
|---|
| Whip pan | whip pan | Extremely fast horizontal pan with motion blur: transitions, conveying panic |
| Handheld | handheld | Slight shake: documentary feel, tension |
| Static shot | static shot / locked-off | Camera completely still: let movement within the frame speak for itself; dialogue and establishing shots |
| Aerial shot | aerial shot / drone shot | High-altitude perspective: geographic relationships, massive scale |
| Zoom | zoom in / zoom out | Focal length changes, perspective stays fixed: surveillance feel or retro vibe; not the same as dolly in/out |
| Dolly zoom | dolly zoom | Dolly and zoom applied in opposite directions; subject stays the same size while background expands or compresses: dizziness and cognitive impact |
Usage discipline
- 1–2 camera move terms per prompt: stacking 3 or more causes conflicts; the common result is only one gets executed, or it compromises into meaningless shaking → for compound movement, choose a term whose meaning is already compound (e.g.,
orbit, follow shot), don’t collage terms.
- Pair camera moves with speed words:
slow / fast / smooth / steady → the same push in, slow is a gaze, fast is impact; without a speed word, speed is random.
- Pair rotation and lateral movement with direction words:
orbit with clockwise / counterclockwise; pans and trucks with left / right → without direction, repeated generations vary in direction and can’t be compared iteratively.
- Also specify subject movement: e.g., “camera slowly trucks right, a person walks toward the camera” → when both camera and subject are accounted for, the shot comes alive.
- A typical 4–8 second generation length usually only accommodates one complete camera move; for two movements, split into two prompts, generate separately, then edit.
- Actual response differences across models: unverified (to be filled in through testing later).
Common mistakes
- Mixing Chinese and English where the two aren’t synonyms (Chinese says “拉远” [pull back], English says
zoom out) → dolly and zoom get conflated; whether the camera physically moves is left to chance
- Stuffing four camera moves into one sentence (push in + pan + orbit + crane up) → commands conflict, output is shaky or only one survives
- Writing only the camera move without specifying what the subject is doing → you get a slideshow-style pan, stiff and lifeless
- Misreading
tracking shot as “tracking focus” → it means the camera moves with the subject; it has nothing to do with focus