Prompt templates and examples

These templates adapt the techniques in the instructional guides. Replace image, video and audio placeholders with actual studio mentions, then adjust characters, locations and timing for your production.

1. A 30-second two-person emotional scene

Prepare: two character images, a location image, a blocking image and the required voices.

[Assets] @Image1 defines the woman, @Image2 the man, @Image3 the seaside location.
@Image4 defines blocking only: woman on the left, man on the right, same axis.
[Style] Natural cinematic light, gentle sea breeze, restrained camera, no extra people.
[0–12s] Medium shot, slow push. She hesitates, offers the flowers and says softly,
“These are for you.” He pauses instead of accepting immediately.
[12–22s] Fixed close-up of her waiting. Tension becomes uncertainty.
No new dialogue and no sudden crying.
[22–30s] Return to the two-shot. He gently accepts the flowers.
Both relax as the camera slowly pulls back.
[Sound] Clear dialogue, natural waves and clothing sounds. No music or subtitles.

Instructional example: a few shots, pauses and subtle expressions convey a relationship.

2. A storyboard grid with separate identities

@Image1 provides the grid, read left to right and top to bottom.
It defines shot order and composition only, not identity, line-art style or labels.
@Image2 defines the protagonist, @Image3 the visitor, @Image4 the same study at night.
Shot 1: establish the study as the protagonist examines a scroll.
Shot 2: the visitor enters from screen right; the protagonist looks up.
Shot 3: they face each other as the protagonist moves the scroll behind their back.
Shot 4: an anomaly appears in the scroll; show the object changing, then the reaction.
Keep clothes, room layout and lighting continuous. No panel borders or subtitles.

Grid-storyboard example: composition and character identity come from separate references.

3. Turn coarse previs into natural acting

@Video1 defines initial positions, key paths, arrival order, pauses and camera intent only.
Red corresponds to @Image1, blue to @Image2 and yellow to @Image3.
Preserve depth, left/right relationships, important occlusion and final positions.
Do not mechanically copy the models' cadence, rigid turns or limb poses.
Actors shift their gaze before starting, decelerate near their destination,
and settle before speaking. Keep breathing, blinking and subtle weight changes.
Smooth the camera along the original path and render natural human performance.
Remove greybox materials, axes, paths, controllers and interface overlays.

Previs-driven output example: retain blocking intent while generating natural performance.

4. Re-render detailed previs

@Video1 is a structurally complete previs reference.
Preserve subject geometry, action, spatial layout, camera movement and cuts.
@Image1 defines final appearance and materials; @Image2 defines the environment style.
Replace grey materials and the blank background with plausible light, textures and detail.
Do not change geometric relationships or retain production guides.

Detailed-previs example: re-render the visual style while retaining structure and motion.

5. Localize dialogue

Preserve @Video1's relationships, event order and performance intent.
Use each character's image for identity and their audio reference for voice.
Use natural American English and the approved lines below, not a rigid literal translation.
Character A: “I thought you'd forgotten.”
Character B: “Not for a single day.”
A speaks first; B responds after a pause. Synchronize lips to the target language.
Remove original-language dialogue. No narration or subtitles.

Before using a template

Replace placeholders; verify mentions; resolve conflicting styles; allow enough time; do not combine uninterrupted-take instructions with hard cuts; define sound clearly. Example outputs demonstrate methods, not guaranteed results for your references or model version. For detailed asset roles, timelines and checks for two-person, three-person and dynamic previs scenes, see Director scene examples.