# Top 3 validation shot brief

## Purpose

This note converts the first three priority validations into an executable
internal shot brief.

It is the first-cycle operating spec for:

- `TSV-001` trio-led setup
- `TSV-002` gadget reveal
- `TSV-005` hero or dialogue closeup

Use this brief to lock framing, action, owned-input requirements, and failure
conditions before generation starts.

## Global execution rules

These rules apply to all three shots:

- master framing: `16:9` at `1920x1080`
- compose center-safe so a later `9:16` crop does not cut the hero read
- run both workflow variants:
  - `generic-baseline`
  - `owned-asset-tuned`
- use only approved owned or rights-holder-cleared inputs
- keep one clear idea per shot
- favor planted or lightly moving poses over complex action
- no-invention rule:
  - no extra characters
  - no invented wardrobe
  - no invented gadgets
  - no genericized background that breaks franchise identity

## Deliverable contract for each shot

Every top-3 validation shot should produce:

- `12` still variations for review
- `3` shortlisted stills
- `1` hero still
- `1` short motion preview in the `2-4` second range
- `1` metadata package recording asset rows, source references, prompt recipe,
  seed logic, and notes

If a shot cannot produce a reviewable motion preview without drift, it may still
pass as a still-first validation, but the failure must be recorded explicitly.

## Shot summary

| Shot ID | Lead subject | Campaign job | Primary owned inputs | Main risk |
| --- | --- | --- | --- | --- |
| `TSV-001` | Sam + Clover + Alex | establish the trio as the core marketing unit | `sam_character_pack`; `clover_character_pack`; `alex_character_pack`; `hero_outfit_pack`; `mission_location_pack_a` | face drift, silhouette overlap, invented suit details |
| `TSV-002` | Sam + approved hero gadget | prove object fidelity and readable hand-to-prop attachment | `gadget_pack_core`; `sam_character_pack`; `mission_location_pack_a` | floating gadget logic, hand failure, unreadable prop silhouette |
| `TSV-005` | Clover closeup | prove face lock and expression control at approval distance | `clover_character_pack`; `hero_outfit_pack` | eye drift, broken mouth shapes, over-finished but off-model face |

## `TSV-001` — trio-led setup

### Objective

Prove that the workflow can hold Sam, Clover, and Alex in one coherent branded
frame without identity drift, pose confusion, or costume invention.

### Shot statement

The trio stands in a mission-ready setup frame inside one approved mission
environment. This is a hero grouping, not a fight beat.

The frame should read as:

- one team
- one world
- one moment
- three distinct identities

### Benchmark intent

This shot follows the official trailer's trio-first campaign grammar:

- readable group wides
- medium hero grouping
- clean silhouette separation
- immediate identity read

### Framing and camera

- camera distance: medium-wide to medium group shot
- crop target: roughly mid-thigh to waist-up depending on the chosen owned
  references
- default lineup:
  - Clover camera-left
  - Sam center
  - Alex camera-right
- all three faces must be readable in the same frame
- no full profiles
- no heavy tilt or distorted lens language
- keep enough headroom and side safety for title or crop adaptation

### Action beat

- mission-ready pause, not combat
- slight forward energy is acceptable
- one shared focus direction is acceptable
- avoid jumps, spins, or crossing limbs
- do not give one character a prop that steals the shot

### Environment

- required environment: `mission_location_pack_a`
- the location must read as specific and repeatable, not generic action blur
- background complexity should support the trio, not compete with it

### Wardrobe

- required wardrobe source: `hero_outfit_pack`
- all three must remain in the same approved mission-suit family
- no mixed outfit logic unless the live brief explicitly requires it

### First-glance read

Within the first glance, the frame must communicate:

- Sam, Clover, and Alex are all present
- the trio is on-model
- the shot belongs to the same campaign world as the official trailer

### Hard failure conditions

Fail this shot if any of the following appear in the hero pick:

- one face drifts off-model
- one character duplicates another character's facial logic
- suit details mutate between characters
- overlapping limbs make the trio unreadable
- the environment turns into a generic or inconsistent space
- any extra invented character appears

### Motion preview rule

If motion is attempted, keep it simple:

- subtle camera drift
- body settle
- slight head movement

Do not attempt:

- full run cycles
- acrobatic motion
- hand-heavy interactions

### Fill before run

- source episode IDs:
- source timecodes:
- selected owned reference files:
- prompt recipe version:
- baseline seed set:
- tuned seed set:

## `TSV-002` — gadget reveal

### Objective

Prove that the workflow can show one approved hero gadget clearly, attached to
the right character, with readable hand-object continuity and minimal finishing
load.

### Shot statement

Sam reveals, activates, or inspects one approved gadget in a clean mission
environment. The gadget is the hero prop, but Sam's identity must remain
readable in the same frame.

### Benchmark intent

This shot follows the official marketing system's gadget punctuation language:

- fast prop readability
- clear attachment to the hero
- clean graphic silhouette
- simple environment support

### Framing and camera

- camera distance: medium or medium-close shot
- Sam must remain the only featured character
- the gadget must sit clearly in the foreground or midground
- keep at least one full hand visible in contact with the gadget
- keep Sam's face readable at the same time
- avoid extreme close crops that force the prop or face out of frame

### Action beat

- approved beats:
  - reveal
  - activation
  - inspection
- avoid:
  - combat strike
  - throwing motion
  - transformation sequence
  - multi-step gadget choreography

### Environment

- required environment: `mission_location_pack_a`
- the background should stay simple enough that the gadget silhouette is clean
- no crowding, extra props, or layered tech noise unless it is directly approved

### Wardrobe and prop logic

- required prop source: `gadget_pack_core`
- required character source: `sam_character_pack`
- wardrobe can inherit from the Sam pack if mission-suit coverage is complete
- if Sam pack coverage is incomplete, pull matching support from
  `hero_outfit_pack`
- do not redesign the gadget
- do not add unapproved screens, weapons, or secondary attachments

### First-glance read

Within the first glance, the frame must communicate:

- this is Sam
- this is the approved gadget
- Sam is physically holding, wearing, or activating it correctly

### Hard failure conditions

Fail this shot if any of the following appear in the hero pick:

- the gadget floats or detaches from the hand/body logic
- fingers merge into the prop
- the gadget silhouette becomes unreadable
- the gadget design drifts off approval reference
- Sam's face becomes secondary to prop chaos
- the environment overwhelms the prop read

### Motion preview rule

If motion is attempted, keep it simple:

- activation pulse
- small wrist or hand movement
- light camera push

Do not attempt:

- complex gadget transformation
- full-body running
- multiple interacting props

### Fill before run

- selected gadget ID:
- source episode IDs:
- source timecodes:
- selected owned reference files:
- prompt recipe version:
- baseline seed set:
- tuned seed set:

## `TSV-005` — hero or dialogue closeup

### Objective

Prove that the workflow can hold a clean, reviewable hero face at close range
with stable expression, hair shape, and costume lock.

### Shot statement

Clover holds a simple dialogue or reaction closeup in her approved hero outfit,
with a soft franchise-coded background and no competing elements.

### Benchmark intent

This shot follows the official trailer's face-driven approval logic:

- front or three-quarter hero angle
- simple readable expression
- short duration
- clean finishing potential

### Framing and camera

- camera distance: closeup or loose closeup
- frame target: shoulders to top of head, with enough space for hair silhouette
- default angle: front three-quarter
- keep both eyes readable
- keep the nose, mouth, and jawline stable
- no extreme beauty-angle distortion
- no direct-to-camera address unless the live brief explicitly requires it

### Expression beat

Use one simple expression family only:

- alert confidence
- focused reaction
- light dialogue emphasis

Avoid:

- shouting
- broad comedy
- exaggerated lip shapes
- highly asymmetrical expression

### Environment

- background may be soft or lightly defocused
- it still needs to feel like the same mission world
- no crowd
- no extra face or prop behind Clover

### Wardrobe and face logic

- required inputs: `clover_character_pack`; `hero_outfit_pack`
- keep hair silhouette, eye shape, and costume neckline stable
- do not invent jewelry, makeup styling, or outfit details not present in the
  approved reference pool

### First-glance read

Within the first glance, the frame must communicate:

- this is Clover
- the face is on-model
- the expression is clear and controlled
- the shot is finishable without rebuilding the face

### Hard failure conditions

Fail this shot if any of the following appear in the hero pick:

- eye drift or mismatched eye logic
- broken mouth or teeth shapes
- hair volume or silhouette drift
- plastic or over-smoothed skin that loses likeness
- expression inconsistency between frames in the motion preview
- costume neckline or shoulder details mutate

### Motion preview rule

If motion is attempted, keep it minimal:

- blink
- tiny head settle
- slight eye-line shift

Do not attempt:

- full lip-sync
- big head turns
- hair-heavy motion

### Fill before run

- selected expression family:
- source episode IDs:
- source timecodes:
- selected owned reference files:
- prompt recipe version:
- baseline seed set:
- tuned seed set:

## Immediate use

As soon as owned materials land:

1. fill the `Fill before run` fields for all three shots
2. map the chosen episode references into
   `docs/internal/reference-extraction-requests.csv`
3. extract deterministic stills with `spies-reference-extract`
4. run the baseline and tuned variants against this brief without changing the
   shot definitions mid-pass