Prompt Structure (one scene)
Order |
Element |
1 |
Subject (who/what, 3–5 fixed traits) |
2 |
Action (one verb per shot) |
3 |
Setting (place + time of day) |
4 |
Camera (shot size + movement) |
5 |
Lighting / mood |
6 |
Style (film stock, animation type) |
7 |
Duration / aspect ratio |
Example: "Woman in yellow raincoat, short black bob, closes a noodle stall; rainy night market; medium shot, slow push-in; neon reflections, warm key light; 35mm film look; 6s, 9:16."
Shot Size
Term |
Use for |
Extreme wide (EWS) |
Establishing location |
Wide (WS) |
Full body + environment |
Medium (MS) |
Waist up, dialogue |
Medium close-up (MCU) |
Chest up, emotion |
Close-up (CU) |
Face, reaction |
Extreme close-up (ECU) |
Eyes, product detail |
Over-the-shoulder (OTS) |
Conversations |
POV |
Viewer as character |
Camera Movement
Term |
Effect |
Static / locked-off |
Calm, readable |
Push-in / dolly in |
Building tension |
Pull-out / dolly out |
Reveal, ending |
Pan left/right |
Follow, survey |
Tilt up/down |
Scale, reveal height |
Orbit / arc |
Hero product shot |
Tracking |
Follow moving subject |
Crane / drone rise |
Epic establishing |
Handheld |
Realism, urgency |
Whip pan |
Energetic transition |
Lighting & Mood Keywords
Keyword |
Result |
Golden hour |
Warm, soft, long shadows |
Blue hour |
Cool, moody dusk |
Hard key light |
Dramatic contrast |
Softbox / diffused |
Clean product look |
Rim light |
Subject separation |
Neon / practicals |
Night city |
Overcast |
Flat, even, natural |
Volumetric fog |
Depth, light beams |
Platform Quick Specs
Platform |
Ratio / length |
YouTube Shorts |
9:16, up to 3 min |
Instagram Reels |
9:16, cover safe in 4:5 crop |
TikTok |
9:16 |
YouTube long-form |
16:9 |
Feed ads |
1:1 or 4:5 |
Lengths change often; check each platform's help page before publishing.
9:16 Safe Zones
- Keep text in the middle third of the frame
- Bottom ~20%: username, caption, music bar
- Right ~15%: like / comment / share buttons
- Top ~10%: status bar, search
- Put faces slightly above centre |
Image-to-Video Tips
- Start image = first frame: compose it as a shot
- Describe motion only, not the image content again
- One camera move per clip
- Match prompt aspect ratio to source image
- Low motion strength for faces, higher for scenery |
|
|
Negative Prompt / Things to Avoid
Avoid |
Why |
Multiple actions per shot |
Model blends them |
Text in the scene |
Usually garbled |
"Beautiful, amazing" |
Adds nothing |
Many characters |
Identity drift |
Fast complex hands |
Artifacts |
Brand logos |
Distortion, legal risk |
Aspect Ratios
Ratio |
Pixels (typical) |
Where |
16:9 |
1920×1080 |
YouTube, web, TV |
9:16 |
1080×1920 |
Shorts, Reels, TikTok |
1:1 |
1080×1080 |
Feed posts, ads |
4:5 |
1080×1350 |
Instagram/Facebook feed |
4:3 |
1440×1080 |
Retro, presentations |
21:9 / 2.39:1 |
2560×1080 |
Cinematic letterbox |
Character Consistency
- Write a character sheet: face, hair, 1 signature item, 1 colour
- Paste the same wording in every scene
- Reuse a reference image where the tool allows
- Avoid changing outfit, age or lighting style mid-story
- Generate scene 1 & 2 first; check the cut |
|
|
Timing Rules of Thumb
- Narration: ~2.5 words per second
- Hook: first 1–2 s
- Scene length: 3–8 s per generated shot
- Short: 15–45 s performs well for most niches
- Story video: 60–180 s = 10–25 scenes |
Workflow Cheat
1. Idea → 6–12 story beats
2. Character sheet
3. One prompt per beat (structure above)
4. Generate 2 scenes, check consistency
5. Generate the rest
6. Voice-over → adjust durations
7. Music bed at −18 to −12 dB under voice
8. Captions, export, SRT
Multi-scene planners such as Everygen Studio's AI Director automate steps 1–7 for videos up to 180 s. To go the other way, from a long video to clips, see Long Video to Shorts. |
|
Created By
Metadata
Comments
No comments yet. Add yours below!
Add a Comment
Related Cheat Sheets