正文
获取原文件references/stage-contract.md
本阶段只拥有 剧集/<EP>/分镜.md 中的 SHOT-...、镜头职责、来源、时长、起止边界、声音、视觉依据和
冻结关键帧。它继承剧本事实与视觉设定，不改写剧情或资产身份。
覆盖比较、场次视觉计划和调度推演都留在上下文；只有最终镜头与关键决定进入 分镜.md。来源使用
场景 ID。已存在的图片提示词条目可用 IMG-... 可见标题引用；它不是已生成图片的证明，也不是分镜的前置门禁。没有条目时直接引用《视觉设定.md》，不建立 coverage、shots、keyframes、audition 或接受记录。
本阶段规则
SHT
| ID | Class | Knowledge |
|---|---|---|
| SHT-01 | structural_invariant | Every production-relevant scene action, line, sound and on-screen text is carried by a shot or explicitly omitted with a creative reason. Each shot's 来源 names scene ids that exist in 剧本.md, and every scene is either claimed by a shot or recorded with its reason on the storyboard's 未拍场次 line. |
| SHT-02 | reviewed_invariant | Each shot has a dramatic/viewing purpose and preserves its source meaning. |
| SHT-03 | craft_default | Keep a short shot focused on the smallest action/reaction unit that carries its purpose; combine or split it according to performance, information, and continuity rather than a fixed count. |
| SHT-04 | craft_default | Change framing/camera because attention, pressure, alignment, reveal, or rhythm changes. Each shot states its camera behavior, where it starts and stops, and why, chosen from the shot's purpose: follow movement that is itself the point, move to take in a reveal, push in as pressure lands on someone, and lock off only when none of these applies—a lock-off is stated with its reason too, never left blank or written bare. A locked camera is a choice, not the episode-wide default; emphasis devices—a short push-in, a freeze frame, a white flash—are reserved for emotional explosions and major reversals, while following, reveal, and pressure moves are not emphasis devices. |
| SHT-05 | structural_invariant | A keyframe projects one shot boundary and the shot's declared visual basis. Any IMG-... used is an existing image-prompt heading, not a claim that generated media exists; when none is suitable, the shot names the relevant textual visual-setting entries without inventing an ID. |
| SHT-06 | reviewed_invariant | A keyframe is one freezeable instant, not an ordered action chain. |
| SHT-07 | taste_option | Lens vocabulary, tempo, and locked/handheld/formal style follow visual direction. |
| SHT-08 | reviewed_invariant | Each authoritative source action is realized once; repeated coverage adds reaction/detail/recontextualization rather than replaying it. The one exception is a declared impact repetition at a peak: the same key action shown from two or three angles, each repeat naming the primary shot, every repeat running from the same before-state to the same after-state so the state changes once, and the next shot starting from the after-state. |
| SHT-09 | reviewed_invariant | Exact Location/View orientation and visible anchors match the camera side used by the shot. |
| SHT-10 | reviewed_invariant | Rendered keyframe prose contains only facts from the boundary that keyframe declares. A start frame carries no state first created by the shot's motion or end; an end frame carries no state already spent before it. Neither frame borrows the other's facts. |
| SHT-11 | craft_default | When information changes another person's power, relationship, knowledge, or choice, preserve that reception visibly; shot count, framing, and duration follow the consequence and project profile. |
| SHT-12 | reviewed_invariant | Each audience-visibility fact binds its exact source, carrier, permission, trigger, and protection method; framing neither reveals that fact early nor hides the carrier this shot must communicate. |
| SHT-13 | reviewed_invariant | Multi-character blocking projects sourced, directed relationships into compatible positions, gaze, distance, and action lines for the current boundary. |
| SHT-14 | reviewed_invariant | A contested moving object preserves ownership, trajectory, direction, time/round state, and end location across cuts unless an authorized ellipsis says otherwise. |
| SHT-15 | reviewed_invariant | When the creator has declared delivery-surface overlay regions with their permanence and source, what a shot must be read for—face and gaze, readable evidence text, the decisive hand action—does not sit only inside those regions, and shots bind the declared version. An undeclared surface leaves the rule inactive: record it as unresolved and do not restage against a guessed region. |
| SHT-16 | structural_invariant | The episode duration is the arithmetic sum of visible shot durations; an unresolved duration stays visibly unresolved, and a target delta is reported rather than used as a universal quality gate. |
| SHT-17 | structural_invariant | A keyframe declares which boundary it freezes and binds that shot's matching boundary field. An end keyframe is a projection of end_boundary, never a second end-state authority, and per-shot keyframe count stays open: one start frame by default, an end frame only when the delivery workflow consumes it. Handing over a start/end pair delegates the motion between them to interpolation, so an action the shot exists for cannot rest on that gap alone. |
| SHT-18 | craft_default | For a scene where directing choice materially changes audience knowledge, alignment, spatial pressure, performance ownership, or the landing, an accepted scene visual plan may bridge project direction and shots; it binds exact screenplay blocks, direction/profile, Location/View and relevant asset states, ordinary scenes skip it, and it never owns screenplay facts or shot boundaries. |
| SHT-19 | reviewed_invariant | When a coverage audition is used, its approaches genuinely differ by knowledge timing, alignment, performance space, strongest image, landing, losses, or production fit; it uses no fixed option, grid, framing, or shot-count formula, and the selected approach is stated before formal shots are written. |
| SHT-20 | reviewed_invariant | Shot revision identity follows directing responsibility rather than array position or text similarity: reorder preserves IDs, insertion creates one, split/merge retires replaced IDs and creates successors, and active coverage plus downstream refs are reconciled before delivery. |
| SHT-21 | reviewed_invariant | A keyframe's copyable text carries only what will be visible; IDs, workflow notes, file paths and craft commentary stay outside the prompt. Two prompts that differ only by identifiers are a template, not two frames. |
| SHT-22 | structural_invariant | Each shot carries a required 视觉依据 field, written after the frozen keyframe, listing every visible character, place, or prop whose identity, Look, or geography must remain recognizable; each item resolves to an existing 视觉设定.md entry. Whatever the keyframe text calls by name — an entry name or its declared 画面代称 — appears there. Small/background subjects remain in scope when recognition matters; offscreen voices and unidentifiable fragments do not. |
| SHT-23 | structural_invariant | An omitted reference choice is not consent to text-to-video. After reference discovery, each shot either binds all verified real images as REF-..., declares creator-supplied pictures as PLAN-... slots that locate an existing IMG-... board or SHOT-... keyframe, preserves partial bindings while naming the still-missing start/identity/geography/prop images, or records the creator's explicit text-to-video choice. A prompt entry, plausible filename, or text description never counts as a real image, and a PLAN-... slot never claims one: it states that the picture is supplied outside the project at generation time, which is why it cannot be sent to production. |
| SHT-24 | reviewed_invariant | A visual basis describes this frame, not the episode's cast list. Each declared 控制 scope stays within what this frame can carry, and a binding is not carried forward merely because the subject remains in the scene. |
| SHT-25 | structural_invariant | Every real reference image a shot binds declares one 用途 from the closed set 身份/造型状态/地理/构图/尺度/效果/起始帧/结束帧/风格, so the shot can say which start frame and which identity/geography/prop pictures a downstream job sends. At most one 起始帧 and one 结束帧 per shot, and 结束帧 requires 起始帧. |
| SHT-26 | structural_invariant | A shot's 来源 is its claim on the screenplay: whatever it quotes, this shot is the one that films it. Every 视觉设定 entry the quote calls by name is therefore either in this shot's 视觉依据 or declared 画外 there. Quoting an action performed by someone the frame never shows takes that action off every list — no other shot claims it and nothing reports it missing, so it simply never gets filmed, and the gap only surfaces when a person watches the finished film. The check matches entry names rather than 画面代称, because a 来源 quote comes from 剧本.md; an entry that opted out with 画面代称：无 stays out of this check too. |
| SHT-27 | reviewed_invariant | When understanding a beat requires seeing its effect on a character, provide a readable reaction or consequence. A separate shot, staging within a shared shot, sound, or an object may carry it. Judge legibility against the intended reveal; do not require a shot count or exclude group shots by default. |
| SHT-28 | craft_default | Shot durations come from the scene's sound-event timeline together with the accepted target_avg_shot_seconds: the timeline sets scene length, the target sets roughly how many cuts it holds, and the average is a scene-level target rather than a per-shot quota. Durations are never folded from screenplay word or paragraph counts. Under an accepted rhythm profile, close-range shots, as the profile field defines them, make up at least close_shot_share_min of the episode's shots. A shot shorter than the model's native minimum is designed at its narrative length, and the storyboard proposes to the creator a multi-shot container (多镜容器) of adjacent same-scene shots, declared in the delivery profile and noting that the target dialect may require an accepted keyframe for every cut; until the creator declares one, or where the dialect cannot carry it, the shot runs to the native minimum and holds its end state. A missing or proposed profile leaves the numeric checks inactive. |
| SHT-29 | craft_default | Dialogue coverage starts from one line per shot, a listener or reaction shot where a line lands, shot/reverse-shot for exchanges, an establishing shot that carries people and action, medium framings for the talk, and close-ups only where a line must land; shots with neither line nor action are cut, and inserts (a phone screen, a hand, an object) are used as cheap rhythm tools. A long line is split across speaker and listener shots rather than sped up. |
CON
| ID | Class | Knowledge |
|---|---|---|
| CON-01 | structural_invariant | Linked end and next start states match or have an explicit owner revision. |
| CON-02 | reviewed_invariant | Knowledge, injury, ownership, weather, light, or physical state does not teleport/regress without story cause. |
| CON-03 | craft_default | Track downstream-relevant deltas, not the whole 设定集 in every shot. |
| CON-04 | structural_invariant | A continuity change states before, after, cause/source scene, effective range, and affected visible IDs. |
| CON-05 | taste_option | Declared montage, ellipsis, dream, or subjective imagery may intentionally break ordinary continuity. |
| CON-06 | structural_invariant | A continuity change names every existing downstream document it affects; future work is described, not pre-created. |
| CON-07 | structural_invariant | A continuity lock declared in 视觉设定.md fixes one verbatim surface plus its shot and image scope; every in-scope frozen keyframe, motion prompt, and named image prompt carries that surface unchanged. |
规则分级由高到低：structural_invariant（结构缺陷，阻断）、
reviewed_invariant（需证据判断）、craft_default（常用做法，可覆盖）、
taste_option（创作者选择，不作缺陷）。创作者已接受的事实优先于本表。