中
Chapter 20. Locations, Props, and On-Screen Text

Part IV — Art Development and the Asset Bible

Chapter 20. Locations, Props, and On-Screen Text#

In this chapter
20.1 A location is not a background image20.2 The location bible20.3 Empty plates and coverage maps20.4 Character activity zones20.5 Location states and destructible states20.6 Grade props by narrative function20.7 The prop bible20.8 Manage important text separately, always20.9 Phone and screen UI20.10 Prop handoffs and ownership20.11 Controlling the scale of locations and props20.12 Testing text legibility20.13 SOP for locations and props20.14 Fault tree20.15 Checklist, exercises and deliverables20.16 The location graph and the cost of movement20.17 Two-track production for critical text assets20.18 Physical specifications and stand-ins for propsA note on sources

20.1 A location is not a background image#

A reusable location contains spatial structure, materials, fixed features, light sources, entrances, character activity zones and camera zones. Keep only one image of a banquet hall and the model will reinvent the doors, tables, screen and stage as soon as the angle changes.

The goal of a location asset is to make every shot feel as though it happens inside one space you could walk through.

20.2 The location bible#

location:
  id: LOC_BOARDROOM_DAY
  name: Shenghua boardroom
  narrative_meaning: procedural power and the public record
  geometry:
    room_shape: rectangular
    north: floor-to-ceiling windows
    east: main door and access control
    south: 85-inch meeting display
    center: long table, twelve places
  fixed_anchors:
    - dark grey stone wall
    - company mark position right of the screen, composited in post
    - the chair's seat backs onto the windows
  materials: matte dark wood table, grey stone, low-reflectance glass
  camera_zones: [door-side establishing, table-end master, Lin Yun side reverse, screen insert]
  lighting: cool side light from the north windows, low-warmth top light
  forbidden: [round table, gold decoration, door moved to the window side]

The compass directions are used by the continuity chapters for eyelines and screen axis.

20.3 Empty plates and coverage maps#

For each principal location produce at least: a forward empty plate, a reverse empty plate, both side angles, important details, entrances, the main prop area, and day/night or story-lighting versions. Empty plates contain no characters, which makes them useful for compositing and repair.

A coverage map records which fixed anchors each camera position can see. If a reverse angle shares no elements at all with the master, the audience may conclude the room has changed.

20.4 Character activity zones#

Define where people may stand, sit, enter, be blocked, and display proof. In the boardroom, the door side means exclusion, the center of the long table means formal power, and standing at the screen means controlling the evidence.

Blocking uses these stable zones first, then designs special positions. That way spatial change carries narrative meaning and reduces random background variation.

20.5 Location states and destructible states#

Locations lose power, get rained on, have things broken, tables disturbed, doors locked, screens damaged. As with wardrobe, keep a base location and state overlays.

location_state:
  base: LOC_BANQUET_HALL
  id: LOCSTATE_E006_BLACKOUT
  changes:
    overhead_lights: off
    led_screen: cached_blue_glow
    emergency_lights: on
    table_state: glasses_undisturbed
  starts: E006_S18
  ends: E007_S03

Power being restored does not automatically restore knocked-over glasses or people's positions.

20.6 Grade props by narrative function#

Background props supply lived-in texture. Action props participate in behavior. Proof props change facts. Continuity props carry state across shots or episodes. The higher the tier, the stricter the asset requirements.

Proof props need their own front, back, dimensions, material, hand-scale reference and text layer. The old pen is both continuity and proof, so it needs normal, opened and card-removed states.

20.7 The prop bible#

prop:
  id: PROP_AUTHORIZATION
  owner: CH_LINYUN
  purpose: proves Lin Yun represents the acquirer
  dimensions: three A4 pages in a black folder
  visual_anchor: fund mark position top left, blue signature area bottom right
  text_layer: GRAPHIC_AUTHORIZATION_V03
  states: [closed, open_title, signature_page]
  hand_scale_reference: REF_PROP_AUTH_HAND_01
  first_appearance: E001_S01_edge
  hero_insert: E001_S09
  rights: fictional institution and signature

20.8 Manage important text separately, always#

Contracts, phones and signage rendered by a generative model are unreliable. Produce important text as a graphic asset and composite it through screen replacement, planar tracking or a separate insert.

A text asset stores the factual truth, the visual layout, font licensing, version, language, first appearance and every reference. When an amount or a date changes, the system can find the episode, the ads, the subtitles and the later dialogue.

graphic_asset:
  id: GRAPHIC_LEDGER_PAYMENT_01
  factual_fields:
    account_holder: Lin Zhenghai
    payment_date: 2026-07-03
    amount: 3000000
  display_format: enterprise payment ledger UI
  languages: [zh-CN, en]
  safe_area: central 70% of 9:16
  references: [E001_S18, E002_S02, AD_PROOF_A]

20.9 Phone and screen UI#

Phone UI has to be legible without impersonating a real product. Build a fictional but consistent operating system, typeface, notification style, clock and battery state. Time on the phone must advance between shots and never run backwards.

Design screen content as a static image first, then decide whether it scrolls, zooms or is tapped. Generation keeps a clean screen plane and its reflections; the text is composited afterwards. Leave space and tracking marks for a finger tap.

20.10 Prop handoffs and ownership#

Record who holds it, where it rests, and when it passes to someone else. Props do not appear automatically when a shot needs them. The authorization moves from Lin Yun's folder to the center of the table, and is then closed by Zhou Lan; each step changes ownership and visible state.

prop_event:
  prop: PROP_AUTHORIZATION
  from: CH_LINYUN_right_hand
  to: TABLE_CENTER
  at: E001_S08
  state_after: open_title

20.11 Controlling the scale of locations and props#

Locations and proof props are up-front investments and should be reused across episodes. A new location must state which scenes and camera positions it carries. A proof object must state its planting, its use, and its value on re-watch.

Merge any lavish location that appears once and carries no emotional or commercial value. Ordinary background props need not all be registered — only assets affecting continuity, action, text or rights enter the registry.

20.12 Testing text legibility#

Test the key fields at the target phone size. The audience does not need to read a whole contract; they need the title, a name, a date, an amount or a result. Shot and subtitle must not compete for the same area.

Display time must be sufficient. Legibility depends on type size, contrast, movement, depth of field and the context the audience already has — not on resolution alone.

20.13 SOP for locations and props#

First, count the principal locations and high-value props from the script. Second, build each location's plan, fixed anchors, empty plates, camera zones and lighting. Third, define activity zones and narrative meaning. Fourth, build state overlays for changes. Fifth, grade props as background, action, proof or continuity. Sixth, produce multiple states and hand-scale references for key props. Seventh, convert all important text into separate graphic assets. Eighth, record handoffs, ownership and references. Ninth, test legibility on a phone.

20.14 Fault tree#

Symptom: changing camera position looks like changing rooms. There is one background and no plan, reverse plate or shared anchors.

Symptom: doors, windows and tables wander. The prompt describes atmosphere and no fixed geometry. Reference the location bible and the approved empty plates.

Symptom: the key document cannot be read. You relied on generated text. Move to a separate graphic asset and an insert shot.

Symptom: props appear from nowhere. Ownership and handoff events are missing. Add prop events and propagate the state.

Symptom: the screen looks great and the story fields are wrong. The graphic asset is not connected to the fact ledger. Generate the content fields from the series and evidence ledgers.

20.15 Checklist, exercises and deliverables#

Check that locations have geometry and orientation; that empty plates cover forward and reverse; that lighting has a source; that state changes have a start and end; that key props have dimensions and states; that text is separately editable; that phone time and UI are consistent; that handoffs are recorded; and that it is legible on a phone.

Exercise one: draw the boardroom plan and four camera zones. Exercise two: split a DNA report into a physical prop and a text layer. Exercise three: trace the ownership of a ring across five shots. Exercise four: test the minimum legible size for names, amounts and dates on a phone screenshot.

Deliverables for this chapter: the location bible, floor plans, the empty plate pack, camera zones, location states, the prop bible, prop events, the graphic asset registry, the phone UI kit, and the legibility report.

20.16 The location graph and the cost of movement#

A list of locations tells the team what exists. A location graph also states how they connect, how characters reach them, how much story time the movement takes, and whether they can see or hear each other. The corridor outside the office, the meeting room, the archive and the underground car park are not four isolated backgrounds; they are one space of action.

The graph prevents spatial teleportation. If a character is in a 28th-floor meeting room and appears at a suburban warehouse ten seconds later, there must be an elision or a transition. If banquet music is audible on a call, the source of that signal needs explaining. The graph also serves production: which similar locations can share structural assets, which shots must establish new space, and which movements can be compressed with a sound bridge and an insert.

Each location records visual anchors, audio anchors, text risk, destructible elements and its principal camera zones. When locations reuse walls or furniture, add a differentiating layer so the audience does not mistake two story locations for one room.

20.17 Two-track production for critical text assets#

Evidence documents, phone messages and screen data split into story truth and visual presentation. Story truth holds names, amounts, dates, relationships and versions. Visual presentation decides typeface, layout, scrolling, occlusion and the readable area. Changing an amount edits the truth source and re-renders; it never means hand-editing five images five times.

During image generation, preserve a clean screen, paper or sign plane and composite the text in post. When a character has to hold it, lock the paper size, folds, orientation and finger positions first, then place the content — a content change should not force regeneration of the whole hand. If gibberish the model produced will ultimately be covered, check that none of it shows at the edges.

Text acceptance covers at minimum: factual correctness, complete glyphs, correct orientation, display duration, legibility at phone size, subtitles not overlapping, and clarity after platform compression. When important evidence appears, the audience must have time to finish reading it — "you can see it if you pause" is not a passing standard.

20.18 Physical specifications and stand-ins for props#

A hero prop records dimensions, material, sense of weight, how it opens, front and back, damage states, and its scale relative to a hand. If the old pen's length, clip direction and metal color change every shot, the audience will doubt it is the same pen. If the authorization changes paper size, the hand action cannot be continuous either.

Complex props can have stand-in versions: closed, open, a high-detail close-up version, a simplified wide version, and a destructible version. Stand-ins share the same prop ID and core identifying features while declaring which shot sizes they suit. A wide does not need engraved lettering rendered; a close-up must not use a simplified model missing the critical evidence.

When a prop changes irreversibly, create a state branch. After the pen is opened and the card removed, the intact version leaves the present timeline; flashbacks may still reference the historical version. The asset library holds both the physical version and its valid story interval, so that a repair artist does not reach for the "cleaner" older image by mistake.

A note on sources#

Location libraries, prop libraries and compositing text in post are central to stable AI microdrama production. This chapter connects them to space, state and evidentiary fact, so that having reference images does not substitute for having a trackable world.