zmime.com
Category · Video

Video prompts

One shot, one movement, one action. Video prompts fail when they try to be a whole scene.

  • Text to video
  • Camera moves
  • B-roll
  • Continuity
  • Image to video
01 — Overview

Why these prompts look the way they do

Text-to-video models generate clips of a few seconds, not scenes. The single biggest improvement you can make is to prompt one shot at a time and describe motion explicitly: what the subject does, and what the camera does. If both are vague, the model invents drift, warping and cuts you did not ask for.

Treat the prompt as a shot card a camera operator could execute: subject, action, camera move, lens, light, duration, audio. Then edit the clips together yourself rather than asking one prompt for a sequence.

02 — Anatomy

The shot-card structure for video

  • Shot type Wide, medium, close-up, over-the-shoulder, macro. Sets scale before anything else.
  • Subject + action One subject performing one continuous action for the clip length.
  • Camera movement Static, slow push in, handheld follow, orbit left, crane down. Include speed.
  • Environment + light Location, time of day, weather, key light source and direction.
  • Lens + grade Focal length, depth of field, film stock or colour grade.
  • Duration + pacing Clip length and whether the action completes inside it.
  • Audio Ambience, foley or dialogue if the model supports sound.
03 — The prompts

5 copy-ready video prompts

Replace everything in [brackets]. Keep the constraint lines — they do more work than the descriptive ones.

Nothing matches that word. Clear the field to see every prompt.

Prompt 01

Single shot with explicit camera movement

Any hero clip where you need controlled motion instead of drift.

  • camera
  • hero shot
Template · plain text 
SHOT: Medium tracking shot, 6 seconds.
SUBJECT: [subject], [wardrobe/material detail], [expression or state].
ACTION: [one continuous action that starts and finishes within 6 seconds].
CAMERA: slow dolly in, roughly 30cm over the shot, ending at chest framing. Steady, no shake, no cuts.
LENS: 40mm, f/2.8, subject in focus, background falls off softly.
ENVIRONMENT: [location] at [time of day], [2 background details only].
LIGHT: [source] from camera-left, soft shadows, consistent brightness for the whole clip.
GRADE: natural colour, mild contrast, fine grain.
AUDIO: [ambience], no music, no dialogue.

Keep one continuous take. No text overlays, no logos, no scene changes.
Prompt 02

Product B-roll loop

Website hero loops and ad cutaways.

  • product
  • b-roll
Template · plain text 
SHOT: Macro orbit, 5 seconds, seamless loop (first and last frame match).
SUBJECT: [product], [material], [colour], resting on [surface].
ACTION: product stays still; only the camera moves.
CAMERA: smooth 45° orbit clockwise at constant speed, locked height, no acceleration.
LENS: 100mm macro, f/4, focus fixed on [key detail].
LIGHT: soft top light with a long specular highlight travelling across the surface as the camera moves, gentle rim light behind.
BACKGROUND: [flat colour or gradient], clean, out of focus.
GRADE: neutral, accurate product colour.

No hands, no text, no reflections of a studio, no additional objects entering frame.
Prompt 03

Image-to-video animation brief

Animating a still you already approved, without redesigning it.

  • image to video
  • animation
Template · plain text 
Animate the attached image. Preserve composition, colour and every design element exactly — do not redraw, restyle or reframe.

MOTION TO ADD
- Primary: [subject or element] [specific small movement, e.g. hair drifting, steam rising].
- Secondary: [background element] [subtle movement].
- Camera: very slow push in, under 5% scale change across 4 seconds.

HOLD CONSTANT
- Faces, logos, text and product geometry: perfectly stable.
- Lighting direction and colour grade: unchanged.
- No new objects, no morphing, no style drift.

DURATION: 4 seconds, loopable if possible.
Prompt 04

Talking-head clip with dialogue

Explainers and social ads where lip sync matters.

  • talking head
  • dialogue
Template · plain text 
SHOT: Medium close-up, 8 seconds, static camera on a tripod.
SUBJECT: [age] [description] presenter, [wardrobe], relaxed posture, natural micro-movements, occasional blink.
DIALOGUE (spoken exactly, conversational pace): "[one or two sentences, max 25 words]"
DELIVERY: [tone], steady eye contact with the lens, small hand gesture on the final phrase.
FRAMING: eyes on the upper third, headroom tight, subject centred.
ENVIRONMENT: [setting] softly out of focus, [one recognisable background object].
LIGHT: soft key from camera-right, subtle fill, background 1 stop darker.
LENS: 50mm, f/2.5.
AUDIO: clean voice, quiet [room ambience], no music.

No captions, no text overlays, no cuts.
Prompt 05

Multi-shot sequence plan

Planning a 30-second video as separate generations that cut together.

  • sequence
  • planning
Template · plain text 
I am making a [duration] video about [topic] for [platform]. Break it into individual shots I will generate one at a time.

For each shot, output a table row with: shot number, duration, shot type, subject action, camera movement, light, and the exact prompt I should paste into the video model.

CONTINUITY RULES TO REPEAT IN EVERY PROMPT
- Character: [fixed description, same wording every time]
- Wardrobe: [fixed]
- Location and time of day: [fixed]
- Lens family and grade: [fixed]

STRUCTURE
1. Establishing wide (3s)
2. Subject introduction (4s)
3-6. Action beats (4s each)
7. Detail insert (2s)
8. Closing wide (4s)

Keep every shot to one action and one camera move. Flag any shot where the action cannot finish inside its duration.
04 — Craft notes

What separates a good prompt from a wasted run

Do this 05

  • Prompt one shot at a time and edit the clips together yourself.
  • State camera movement and its speed, or say "static".
  • Match the action to the clip length so it completes on screen.
  • Repeat character, wardrobe and location verbatim across shots.
  • Generate several takes; video output varies more than images.

Not this 05

  • Asking for a full narrative with multiple cuts in one prompt.
  • Describing two simultaneous camera moves.
  • Expecting readable on-screen text from the model.
  • Long dialogue in a short clip — lip sync degrades fast.
  • Vague motion words like "dynamic" or "epic".
05 — Questions

Asked often, answered plainly

Why do my clips warp or drift halfway through?

Usually too much requested motion for the duration, or conflicting instructions (a moving camera plus a moving subject plus a changing environment). Cut one variable, shorten the action, or extend the clip.

How do I keep the same character across shots?

Reuse the exact same description string in every prompt, keep wardrobe and lighting fixed, and where possible start each shot from a reference frame of the previous one. Small wording changes read as a different person.

Is image-to-video more reliable than text-to-video?

For brand work, yes. You approve the frame first, then ask only for motion, which removes most composition and identity risk. Keep the motion brief small and explicitly forbid restyling.

06 — Sources

Read the primary documentation

Vendors change flags, limits and defaults often. Where this site disagrees with official docs, the docs win.

07 — Tool packs

Same work, phrased for your tool

  1. 01

    Prompt for Sora

    Write Sora prompts like shot cards. One subject, one action, one camera move, one light setup.

    • Video model
     4 prompts
  2. 02

    Prompt for Veo

    Veo is strongest when you approve the frame first and ask only for motion and sound.

    • Video model
     4 prompts
  3. 03

    Prompt for Gemini

    Gemini is at its best when you hand it a lot of material and a precise job to do with it.

    • Multimodal model
     5 prompts