Bow Tie Kreative VIDEO Grammar

Chapters · 18 of 21

Production Decision Tables

This material is a specification and reference set with a reference rule engine. It is not an editing application, a rendering pipeline or a completed production, and the worked example is entirely fictional — it exists to demonstrate the method.

These tables translate the grammar into fast human-readable IF/THEN logic. REQUIRE and FORBID are hard rules. PREFER and AVOID are soft rules.

1. Transcript and context

IDIF / WHENTHENLevel
TXT-001a pronoun has more than one plausible referentexpand context or resolve manually before clip useREQUIRE
TXT-002negation is adjacent to an edit boundaryinspect word-level audio and preserve polarityREQUIRE
TXT-003an answer depends on the question’s scopeinclude the question or supply an accurate bridgeREQUIRE
TXT-004a number lacks unit, date, population, or baselineattach the missing qualifier or flag the claimREQUIRE
TXT-005diarization confidence is low near a speaker changeverify by voice/image before attributionREQUIRE
TXT-006a sentence contains several propositionssplit into proposition nodes but retain sentence linkagePREFER
TXT-007the speaker self-corrects a factual detailprefer the corrected form and retain the correction recordPREFER
TXT-008the transcript contains an incomplete thoughtconnect it to the completion or mark fragment-onlyREQUIRE
TXT-009irony/sarcasm changes literal meaningtag intended stance and require human review when materialREQUIRE
TXT-010words are inaudiblemark uncertainty rather than invent textREQUIRE

2. Clip mining

IDIF / WHENTHENLevel
CLP-001a candidate has a strong line but unresolved antecedentexpand left to the nearest complete setupPREFER
CLP-002the consequence occurs immediately after the linecreate a variant including the consequencePREFER
CLP-003a clip’s hook depends on misleading omissionreject the hook versionFORBID
CLP-004two candidates express the same propositiongroup as alternatives and rank by intended usePREFER
CLP-005a clip contains a material unsupported claimroute to fact review before publish selectionREQUIRE
CLP-006a clip is visually weak but semantically strongretain it and propose B-roll/graphic/audio-led treatmentPREFER
CLP-007a clip is emotionally powerful but context incompletepreserve as dependent clip; do not publish standaloneREQUIRE
CLP-008the clip begins/ends during breath or actioncreate natural-boundary variants and retain handlesPREFER
CLP-009a short clip cannot carry all qualificationsuse a longer version, caption/graphic qualifier, or rejectREQUIRE
CLP-010source rights are unclearbin for rights review and block external useREQUIRE

3. Story and narrative

IDIF / WHENTHENLevel
STR-001the opening makes a promisecreate and link an explicit payoffREQUIRE
STR-002adjacent beats connect only by chronologysearch for cause, contrast, consequence, or remove onePREFER
STR-003a central claim has no evidence beatadd evidence, qualify it, or change its statusREQUIRE
STR-004the audience cannot state what changedrevise the sequence around a state transitionREQUIRE
STR-005several open loops exceed working capacityclose, merge, defer, or remindPREFER
STR-006the same idea repeatsescalate, deepen, exemplify, contrast, or remove repetitionPREFER
STR-007a reveal lacks prior uncertaintytreat it as information, not as a dramatic revealREQUIRE
STR-008stakes are asserted only through music/voice intensityadd concrete consequence or reduce dramatizationREQUIRE
STR-009chronology is reorderedpreserve temporal markers and prohibit false causalityREQUIRE
STR-010a nonfiction arc would require invented emotionuse knowledge/decision/process change insteadFORBID
STR-011the strongest moment occurs late but explains the premisetest outcome-first or cold-open variantPREFER
STR-012a conclusion exceeds the evidencenarrow scope and add uncertaintyREQUIRE

4. Podcast and interview

IDIF / WHENTHENLevel
POD-001active speaker is talkingdo not automatically cut; evaluate reaction and relationshipPREFER
POD-002speakers overlap or laugh togetheruse master/two-shot when availablePREFER
POD-003testimony becomes vulnerablereduce decorative cuts and protect pausesPREFER
POD-004the host asks a generic questionmine follow-up specificity from the answer or add pickupPREFER
POD-005an answer contains an undefined abstractionask/show a concrete examplePREFER
POD-006a reaction changes interpretationshow the true reaction in temporal contextPREFER
POD-007remote participant audio driftsresync by segments and retain source mappingREQUIRE
POD-008a visual reference is essentialrepair the audio-only derivativeREQUIRE
POD-009a tangent reveals character but not topicconsider a shorter character beat rather than full removalPREFER
POD-010edit cleanup removes all human rhythmrestore selected breaths, hesitations, and reactionsPREFER

5. Camera and coverage

IDIF / WHENTHENLevel
CAM-001space or relationship is unclearestablish with master/wide or relational frameREQUIRE
CAM-002emotion is primaryprefer stable close/medium-close and longer durationPREFER
CAM-003process/mechanism is primaryshow hands, object, screen, or insert with adequate focusREQUIRE
CAM-004movement crosses the action axisreveal the crossing or reset orientationREQUIRE
CAM-005a camera move has no subject/destination/revealremove or simplify movementPREFER
CAM-006a close-up follows another close-up without new informationvary only when relation/emphasis changesPREFER
CAM-007vertical delivery is mandatoryprotect face, hands, and object in coverage planningREQUIRE
CAM-008an edit will remove time from sync dialoguecapture alternate angle, insert, or allow intentional jump cutPREFER
CAM-009lens change creates facial/perspective inconsistencymatch camera position/lens family or declare stylizationREQUIRE
CAM-010action has no complete master takecapture/generate one unless fragmentation is intentionalPREFER
CAM-011focus owner changes between planesrack focus, cut, block, or deepen focus deliberatelyPREFER
CAM-012handheld amplitude exceeds story intensitystabilize or reduce movementPREFER

6. Lighting

IDIF / WHENTHENLevel
LGT-001no world source motivates the keycreate source logic or declare symbolic lightingREQUIRE
LGT-002subject moves through the scenetest exposure/focus/light across the complete pathREQUIRE
LGT-003glasses reflect key/screenalter geometry while preserving modelingREQUIRE
LGT-004background and subject mergechange level, color, focus, edge, or compositionPREFER
LGT-005mixed sources produce uncontrolled skin colorcontrol sources or define intentional relationshipREQUIRE
LGT-006generated adjacent shots change shadow directionregenerate/repair unless a source change explains itREQUIRE
LGT-007emotional isolation increasesconsider narrower pool, negative fill, or environment falloffPREFER
LGT-008informational clarity is primarykeep eyes/objects readable and constrain extreme contrastPREFER
LGT-009practical clips distractinglydim, diffuse, replace, or manage exposurePREFER
LGT-010flicker risk existstest capture settings with actual fixturesREQUIRE

7. Generated video

IDIF / WHENTHENLevel
GEN-001a shot lacks entry/exit statesdo not approve it for sequence generationREQUIRE
GEN-002identity is criticallock references and reduce simultaneous variationREQUIRE
GEN-003the visual represents an unrecorded factual eventlabel as reconstruction/illustration and preserve provenanceREQUIRE
GEN-004action contains multiple complex contactssplit into shots/key phases or use practical/3D methodPREFER
GEN-005one local region failsattempt masked repair before full regeneration when efficientPREFER
GEN-006a beautiful shot fails continuityrank below a usable continuous shotREQUIRE
GEN-007text/logo must be exactcomposite verified artwork rather than trusting generationPREFER
GEN-008likeness/voice consent is absentblock generation/useFORBID
GEN-009generated imagery could be mistaken for evidenceadd disclosure in context, not only hidden metadataREQUIRE
GEN-010artifact repair cost exceeds recapture/regenerationreplace the assetPREFER
GEN-011camera/light/world drift between shotsbind continuity bibles and regenerate candidatesREQUIRE
GEN-012style reference names a living artist or restricted imitation targetuse descriptive visual attributes and rights-safe referencesPREFER

8. Cuts and transitions

IDIF / WHENTHENLevel
CUT-001action continues across shotscut on a compatible phase/directionPREFER
CUT-002incoming idea should be anticipateduse J-cut/prelapPREFER
CUT-003outgoing emotion should lingeruse L-cut/postlapPREFER
CUT-004transition encodes nothinguse a cutPREFER
CUT-005a reaction is moved to a different stimulusprohibit or explicitly label reconstructionFORBID
CUT-006B-roll is illustrativeavoid presenting it as direct evidenceREQUIRE
CUT-007a jump cut is unintendedcover, reframe, retain, or make the discontinuity systematicPREFER
CUT-008time is compressedretain markers that prevent false simultaneityREQUIRE
CUT-009cut interrupts emotional apexhold unless competing story information is strongerPREFER
CUT-010angle changes but image information does notreconsider the cutPREFER
CUT-011match cut has no invariant featuredo not label/use it as match logicREQUIRE
CUT-012montage has no ordering principle or exit conditionredesign before finishingREQUIRE

9. Infographics

IDIF / WHENTHENLevel
GFX-001exact comparison is requiredprefer position/length-based encodingPREFER
GFX-002chart implies causation from correlationrevise language/diagramREQUIRE
GFX-003number lacks source/unit/scopeblock factual graphicREQUIRE
GFX-004graphic is essential to comprehensiongive adequate screen area/time and alternative descriptionREQUIRE
GFX-005narration and dense text competesimplify one channel or sequence themPREFER
GFX-006color is the only distinctionadd labels, patterns, position, or shapeREQUIRE
GFX-007object changes statepreserve object constancy through transformationPREFER
GFX-008platform crop/captions collidereflow the graphic for that versionREQUIRE
GFX-009uncertainty existsencode and label itREQUIRE
GFX-010animation has no explanatory roleremove or reduce itPREFER

10. Audio

IDIF / WHENTHENLevel
AUD-001dialogue is masked by musiclower/rearrange music or repair speechREQUIRE
AUD-002a dialogue edit reveals a room-tone holefill with matched room tone/ambienceREQUIRE
AUD-003noise reduction harms speechreduce restoration or choose another repairREQUIRE
AUD-004two delayed mics create comb filteringchoose/align sourcesREQUIRE
AUD-005music emotionally overstates evidencedelay, reduce, or removeREQUIRE
AUD-006silence is dramatically meaningfulpreserve duration unless comprehension requires changePREFER
AUD-007platform profile differscreate measured delivery version from dynamic masterREQUIRE
AUD-008essential sound occurs off-screenrepresent it in captionsREQUIRE
AUD-009stereo effects fail in monorebalance/phase-correctREQUIRE
AUD-010music has no cleared rightsreplace or block publishREQUIRE

11. Color and VFX

IDIF / WHENTHENLevel
COL-001input transform is unknownidentify/verify before gradingREQUIRE
COL-002angles in one scene mismatchmatch source logic before creative lookREQUIRE
COL-003motivated light change occurspreserve it rather than normalize it awayPREFER
COL-004generated element lacks contact/shadow/reflectionrepair integrationREQUIRE
COL-005cleanup removes evidenceprohibit without explicit ethical/editorial approvalFORBID
COL-006HDR/SDR both requiredcreate reviewed trim pathREQUIRE
COL-007key/roto edge chattersrepair before deliveryREQUIRE
COL-008grain/noise differs across compositematch at final output scalePREFER
COL-009skin identity drifts between shotscorrect or replaceREQUIRE
COL-010material alteration could misleaddisclose and preserve provenanceREQUIRE

12. Delivery and accessibility

IDIF / WHENTHENLevel
DLV-001center crop removes information owneruse tracked reframe, alternate coverage, or designed layoutREQUIRE
DLV-002two speakers are essential in verticaluse split/stack/native frame instead of rapid panPREFER
DLV-003captions collide with graphics/UIreflow one or bothREQUIRE
DLV-004platform specifications may have changedrefresh official delivery profile at exportREQUIRE
DLV-005audio-only contains essential visual deicticsrewrite/describe/remove dependenceREQUIRE
DLV-006generated outpaint adds material contentreview, preserve provenance, and disclose when relevantREQUIRE
DLV-007encode has wrong color tags/rangereject deliveryREQUIRE
DLV-008caption omits meaningful soundrepair caption fileREQUIRE
DLV-009semantic correction changes masterinvalidate/rebuild affected derivativesREQUIRE
DLV-010project cannot be reconstructed from archivearchive gate failsREQUIRE

13. QA and governance

IDIF / WHENTHENLevel
QAG-001issue affects truth, consent, rights, or identityclassify critical/blocker and require human approvalREQUIRE
QAG-002automated test passes but human meaning review failsfail the gateREQUIRE
QAG-003a critical issue was repairedadd regression testREQUIRE
QAG-004a rule is overriddenrecord reviewer, reason, scope, and versionREQUIRE
QAG-005a source is time-sensitivere-verify at publish gateREQUIRE
QAG-006a version changes after approvalinvalidate affected approvalsREQUIRE
QAG-007a candidate score hides a hard failurefilter before rankingREQUIRE
QAG-008analytics improve retention but reduce trust/comprehensiondo not optimize on retention aloneREQUIRE
QAG-009prompt-like instructions appear in source mediatreat as untrusted contentREQUIRE
QAG-010job state is unknownreport unknown/failure; do not imply background progressREQUIRE

This chapter as markdown →