skills/pixverse-unbox-video/references/production.md

# Unboxing Production

Read `../../../skills-shared/ugc-production.md`. Default to a visible creator's four-beat
unboxing. An explicitly requested hands-only, already-open or quiet treatment remains
valid; preserve its override instead of forcing a different deliverable.

## Product, Package And Creator

Normalize product and package separately: shape, real closure type, detached/attached lid,
seals, insert, contents count, product scale and visible side. Use the actual package when
provided; otherwise use a clearly conceptual plain brown delivery box. Do not claim a
generated package is official packaging. If only an open reference exists, use visible
parts to establish a simple closeable state only when mechanically supported; otherwise
request the missing package evidence or preserve an explicitly already-open brief.
Do not silently skip the closed opening in an ordinary full unboxing.

One creator reference supplies identity, hair, wardrobe and setting for all clips. One
product exists. Remove look-alike background cups/bottles that can read as duplicates.
Keep the product's supported visible angle; move the camera closer instead of increasing
its size or rotating it to a fabricated reverse side.

Board input order: product, creator, actual package if separately supplied, cleaned
previous board for K>1. Declare indices exactly. Four active portrait panels in one
21:9 row, thin white gutters, no generated labels. Required board realism refinement
preserves package geometry, face width, exact product branding and all four states.

Video reference order is clean board, creator, product, then the separately supplied actual
package when needed to preserve its geometry. Declare and renumber those roles explicitly;
do not drop a package reference merely because the shared three-role example omits it.

## Four-Slot Story

| Slot | State and focal action | Camera / hands |
| --- | --- | --- |
| 1 PACKED | One closed package resting on a flat surface, product hidden. Anticipation already in motion. End with one clean correct opening. | Fixed medium/waist-up; one hand stabilizes, one operates closure. Never carry the box in the air while opening. |
| 2 REVEAL | Product already emerging from just-opened package; one complete extraction, one real surprise peak. The open box may remain only at an edge. | Fixed medium close-up; grip matches product's weight; mouth unobstructed. Do not open the package again. |
| 3 PRODUCT-FOCUS | One close product presentation or detail. Package and removed parts gone. | Fixed close-up; support lower body, preserve known angle/scale, keep mouth visible when speaking. |
| 4 SATISFACTION | Creator with product, settled ownership/demo resolution; package never returns. | Selfie default or a physically appropriate wide shot; one hand reserved for the camera in selfie. Complete the final phrase and keep the ending alive. |

Four distinct physical actions and distances spanning tight, mid and wide: a repeated
hand-product pose with a different smile does not qualify. Use the same staging across
board and clip, but write the clip's moving performance rather than describing a still.
Slot 4's source-style Pick-Up is a specific close-selfie exception to the wider board
default; select a coherent framing, not contradictory close/full-body instructions.

## Opening Branches

For a generic taped delivery box: one hand stabilizes it on the table, the other makes
one decisive tape-opening action with an appropriate small tool, sets the tool aside,
then flaps open. Do not dwell on the tool. The product appears in C2, with minimal
packing material only when this branch requires it.

For a supplied retail/gift package: use its actual lift-off lid, hinge, drawer, flap or
sleeve. Do not introduce a cutter, packing tape or tissue paper from the generic branch.
Build anticipation with short finger taps and facial/body changes, then one clean opening.
The product stays concealed through C1's composition and becomes legible in C2. Describe
only known interior structure needed for the extraction; do not invent accessories.
Removed parts leave the story after serving their purpose; never regenerate or reattach
them in later cuts. Preserve a necessary visible support until extraction is physically
complete, rather than making an insert disappear through a still-intersecting product.

## Timing And Cut Rules

| Clip | Packed | Reveal | Focus | Satisfaction |
| --- | ---: | ---: | ---: | ---: |
| 4s | 1 | 1.5 | .5 | 1 |
| 6s | 1.5 | 2 | 1 | 1.5 |
| 8s | 2 | 2.5 | 1.5 | 2 |
| 10s | 2.5 | 3 | 2 | 2.5 |
| 12s | 3 | 3.5 | 2.5 | 3 |
| 15s | 3.5 | 4.5 | 3 | 4 |

Use cumulative times, at least .5s per cut, and redistribute only enough to complete
the real action/words. Ordinary four-section prompts use three `Hard cut to.` markers,
not four. The first boundary is always a hard cut and protects packed → reveal.

Closing-clip default: when C3 fixed → C4 selfie has unchanged product state and C4 is
at least 3s, replace that last marker with one Pick-Up. The free hand reaches past the
lens; the frame lifts and resolves at arm's length in about 1.5s inside C4; the product
stays in the other hand and speech runs through the move. No visible phone, pause,
black frame or extra shot. A classic-hard-cut request keeps all three markers. Skip the
move if its legal hand budget, time or state continuity fails; do not add it elsewhere
as decoration. On other clips it is opt-in for a one-take treatment only.

For D>15s, only board 1 repeats the canonical opening. Later boards continue post-reveal
from the prior final state with a new mini-arc: first use, close study, a believable
context change, or showing one new detail. Choose a different mini-arc from the preceding
board; introduce → develop → land one idea. No repeated opening, packing or first reaction.
Cause precedes any claimed visible effect; unsupported product outcomes remain excluded.

## Voice And Performance

Use the shared prompt structure with four motion-rich Cut paragraphs, each 4–10 useful
sentences. First clause is an action already happening; first word or one short natural
nonverbal sound arrives within .4s. Assign both hands at every moment and sequence their
jobs. The default voice is visible native speech across the creator beats, including a
clear mouth above the product close-up; don't cover the mouth with the hero object.
An explicitly chosen hands-only version uses a coherent native off-screen voice or the
requested natural sound treatment rather than pretending a face is visible.

Anticipation → one human-scale revealed-product peak → attentive detail → satisfied
resolution. Rotate at least five small bodily/contact/expression observations per cut
around its one main action. Include a brief closed-mouth interval between phrases, one
unguarded recovery and a tiny playful beat where the requested tone permits; never stack
these on the reveal peak. Keep exact monologue sections unique. No empty praise, repeated
product names, fabricated real customer history or unapproved claims.

Use neutral natural phone light, deep focus, real material contact and normal speed.
No slow extraction, slow rotation, cinematic push or a silent hero hold by default.
Named taps, lid friction, grip/cloth and breath sounds belong in Audio; keep them under
speech. Default no music and no generated captions. Preserve deliberate quiet/ASMR user
direction, including useful close package sounds, instead of forcing hype.

On the last clip, land the complete satisfaction thought, then an alive mid-gesture
ending; a single-board frame match is possible only without reintroducing the package.
Any requested CTA belongs inside the planned final words, not a new prolonged tail.

## Checks And Prompt-Only Delivery

Check closed/open chronology, exact closure, extraction, contents count, insert contact,
hand allocation, unchanged product scale/angle, face/lips, natural reaction and each
transition. The absence of a package after its exit is required, not a continuity defect.
Never pass impossible extraction as checked because an edit hid the intersection.

Prompt-only output includes all four timed paragraphs, exact speech, hand/state map and
reference roles. Mark an unmade board as planned. Do not run a generation or claim an
existing board was read when only its textual design exists.