A tool-agnostic quick reference for writing text-to-video and image-to-video prompts. Covers prompt anatomy, camera, shot and lighting vocabulary, aspect ratios, negative cues and fixes for common problems.
|
Part
|
What to write
|
|
Subject
|
Who or what is on screen. Be concrete: "a ceramic teapot", not "an object"
|
|
Action
|
One visible thing that happens: steam rises, a hand waves, mist drifts
|
|
Setting
|
Where and when: "on a pale wooden table", "quiet park path at dusk"
|
|
Camera
|
One shot size and one camera move: "medium shot, slow push in"
|
|
Lighting
|
Source, direction, quality: "soft window light from the left"
|
|
Style
|
Look and texture: "realistic, natural colours", "warm neutral palette"
|
|
Format
|
Aspect ratio, duration and resolution are set in the tool's controls. The text only describes framing
|
|
Order
|
Subject → action → setting first, then camera, light and style
|
|
Scope
|
One shot per prompt. Several locations or a whole story in one short clip is hard to control
|
|
Term
|
Effect on screen
|
|
Static / locked-off
|
Camera does not move; safest when the scene is busy
|
|
Push in / dolly in
|
Camera moves toward the subject; builds focus
|
|
Pull out / dolly out
|
Camera moves away; reveals surroundings
|
|
Pan left / right
|
Camera turns horizontally from a fixed spot
|
|
Tilt up / down
|
Camera turns vertically from a fixed spot
|
|
Tracking / trucking
|
Camera travels alongside a moving subject
|
|
Orbit / arc
|
Camera circles the subject; good for products
|
|
Crane / pedestal up
|
Camera rises vertically, often to reveal a scene
|
|
Handheld
|
Slight natural shake; documentary feel
|
|
Zoom in / out
|
Lens changes focal length; the camera stays put
|
|
Rule
|
One move per shot. Don't combine conflicting moves, e.g. "locked camera" plus "rapid handheld zoom"
|
|
Shot
|
Frames
|
|
Extreme wide (EWS)
|
Subject small in a big landscape; sets location
|
|
Wide / long (WS)
|
Whole subject plus surroundings
|
|
Medium-wide (MWS)
|
Roughly knees up; body language and setting
|
|
Medium (MS)
|
Waist up; the default for people talking or gesturing
|
|
Medium close-up (MCU)
|
Chest up; face plus some context
|
|
Close-up (CU)
|
Face or a whole product fills the frame
|
|
Extreme close-up (ECU)
|
One detail: an eye, a texture, a logo-free label
|
|
Over-the-shoulder (OTS)
|
Behind one person, looking at another or at an object
|
|
POV
|
What the subject sees
|
|
Low / high angle
|
Camera below or above eye level; adds power or smallness
|
|
Term
|
Look
|
|
Soft light
|
Diffused, gentle shadows (overcast, window with curtain)
|
|
Hard light
|
Crisp, defined shadows (direct sun, bare bulb)
|
|
Golden hour
|
Warm, low-angle light just after sunrise or before sunset
|
|
Blue hour
|
Cool, even twilight just before sunrise or after sunset
|
|
Backlight / rim light
|
Light behind the subject; outlines edges
|
|
Side light
|
Light from one side; shows texture and shape
|
|
Key / fill
|
Main light vs. softer light that lifts the shadows
|
|
High key
|
Bright, low contrast, few shadows
|
|
Low key
|
Dark, high contrast, moody
|
|
Practical light
|
A visible lamp, neon sign or screen in the scene
|
|
Colour temperature
|
"Warm" (orange) vs. "cool" (blue); name both if you mix them
|
|
Word
|
Use it for
|
|
Slowly / gently
|
Calm, controlled motion; usually the most stable choice
|
|
Gradually
|
Change that builds across the clip (light rising, mist thinning)
|
|
Steadily
|
Constant speed with no sudden changes
|
|
Briskly / quickly
|
Energetic action; expect more artefacts
|
|
Drifts / floats
|
Weightless motion: mist, smoke, petals, dust
|
|
Sways / ripples
|
Small repeating motion: grass, fabric, water
|
|
Continuous shot
|
Asks for no cuts within the clip
|
|
Real-time
|
Normal speed; avoids an accidental slow-mo look
|
|
Slow motion
|
Stretched action: splashes, hair, falling objects
|
|
Pacing tip
|
One main action plus at most one small background motion
|
|
Ratio
|
Typical use
|
|
9:16 (vertical)
|
Shorts, Reels, TikTok, Stories. Keep the subject near the centre and leave space for captions
|
|
16:9 (horizontal)
|
YouTube, websites, presentations, cinematic landscapes
|
|
1:1 (square)
|
Feed posts, product tiles, carousels
|
|
Check support
|
Not every model offers every ratio; pick one that the chosen model supports
|
|
Set it in controls
|
Choose ratio, duration and resolution in the tool's settings
|
|
"4K" in text
|
Writing "4K" in a prompt does not change the output resolution you selected
|
|
Duration
|
Short clips favour one simple action; save longer ideas for several shots
|
|
Cue / habit
|
Why
|
|
"No titles, captions or logos"
|
Stops stray on-screen text; add your own text later in an editor
|
|
"No cuts"
|
Keeps one continuous shot
|
|
"No people"
|
Use it for landscapes and products that should stay empty
|
|
"Keep the [subject] fully in frame"
|
Phrasing it positively often works better than "don't crop"
|
|
Negative field
|
If the tool has a separate negative-prompt field, put exclusions there; otherwise add one short "No…" sentence at the end
|
|
Avoid: stacked adjectives
|
"Epic, stunning, ultra, cinematic, 8K" adds noise, not detail
|
|
Avoid: conflicting cues
|
"Static camera" plus "sweeping drone shot"
|
|
Avoid: story in one clip
|
Multiple scenes, outfit changes or locations in a single short generation
|
|
Avoid: real brands or people
|
Use unbranded objects and generic people unless you have the rights
|
|
The image already fixes the subject, composition and style. Spend the prompt on motion and camera.
|
|
Describe what moves and what stays still: "steam rises; the teapot stays fixed in frame".
|
|
Start with a clean, well-lit image with a clear subject. Clutter usually turns into wobble.
|
|
Match the image's aspect ratio to the output ratio to avoid unwanted cropping.
|
|
Keep camera moves small (slow push in, gentle orbit) so the scene doesn't drift away from the source.
|
|
Don't re-describe the image with different details. Conflicting text can pull the result off-model.
|
|
Use reference images only on models that accept them, and only images you have the right to use.
|
BEFORE
a cool cinematic video of a cat in a city, epic, 4K, ultra realistic
AFTER
A ginger cat walks along a rain-wet stone wall in a quiet city
alley at dusk. Neon signs reflect in the puddles below. Medium-wide
shot; the camera tracks slowly alongside at the cat's eye level in
one continuous shot. Soft blue ambient light with a warm glow from
the signs, realistic fur texture. No text, captions or logos.
WHAT CHANGED
+ concrete subject, one action, a named setting and time
+ one shot size and one camera move
+ lighting with a source and colour
+ one exclusion sentence
- removed "epic", "cinematic" and "4K" (set resolution in controls)
Ratio and duration: chosen in the settings, e.g. 9:16 for a Short|
Symptom
|
Change
|
|
Motion is chaotic or busy
|
Simplify to one action; add "slowly"; drop background motion
|
|
Camera ignores the move
|
State one move plainly near the start of the camera sentence; remove other camera words
|
|
Shaky or warping scene
|
Try a static camera first, then add movement back
|
|
Subject drifts out of frame
|
Add "keep the [subject] fully in frame"; use a wider shot size
|
|
Subject changes look mid-clip
|
Shorten the prompt; describe the subject once with fixed details; for image-to-video, use a cleaner source image
|
|
Random text or logos appear
|
Add "No titles, captions or logos"
|
|
Wrong mood
|
Change only the lighting line (e.g. soft → low key) and keep the rest
|
|
Framing wrong for vertical
|
Choose 9:16 in the controls; centre the subject; say "medium shot" rather than "wide"
|
|
Looks low-res despite "4K"
|
Pick a higher resolution in the controls; the text doesn't override it
|
|
Many things wrong at once
|
Change one variable per attempt and note which change helped
|
No comments yet.
Cheatography n. The art of mapping complex subjects into clear, instantly navigable reference guides.