# 14 — Multiformat Delivery Grammar ## 1. First principle: one semantic edit, many render decisions The master story should not be rebuilt independently for every platform. ```text SEMANTIC TIMELINE → DELIVERY PROFILE → REFLOW/REFRAME RULES → GRAPHIC/CAPTION LAYOUT → ENCODE → PLATFORM-SPECIFIC QA ``` Content decisions remain linked across versions; rendering decisions may differ. ## 2. Delivery-profile dimensions ```text container video codec codec profile/level resolution pixel aspect ratio display aspect ratio frame rate scan/progressive state bit depth chroma subsampling color primaries transfer function matrix coefficients range flag HDR metadata bitrate/rate control GOP/keyframe policy audio codec sample rate bit depth channel layout loudness target true-peak ceiling caption format caption burn-in state chapter metadata thumbnail/poster file naming checksum ``` ## 3. Common aspect families ```text 16:9 — horizontal video and general web delivery 9:16 — vertical mobile video 1:1 — square 4:5 — portrait feed 3:2 — photographic/certain display contexts 4:3 — legacy, archival, or deliberate aesthetic 1.85:1 — theatrical widescreen 2.39:1 — scope/cinematic widescreen ``` Aspect ratio is a container constraint, not a story purpose. The reframe must preserve the information owner. ## 4. Canonical frame coordinates Represent layout in normalized coordinates: ```text x ∈ [0, 1] y ∈ [0, 1] width ∈ [0, 1] height ∈ [0, 1] ``` Store important regions as polygons or tracked boxes: ```text face mouth hands speaker body object/evidence graphic caption logo source note negative space platform UI exclusion ``` This allows each delivery profile to solve a constrained composition rather than applying a blind center crop. ## 5. Reframe objective ```text ReframeUtility = subject_visibility + information_owner_visibility + gaze/lead_room + composition + motion_continuity + graphic_legibility + caption_clearance - crop_damage - rapid_pan_cost - identity_switch_cost - UI_collision ``` ## 6. Reframe operations ```text STATIC_CROP KEYFRAMED_PAN_AND_SCAN SUBJECT_TRACK MULTI_SUBJECT_DYNAMIC_FRAME SPLIT_SCREEN STACKED_LAYOUT PICTURE_IN_PICTURE BACKGROUND_EXTENSION GENERATIVE_OUTPAINT GRAPHIC_REFLOW SHOT_REPLACEMENT ALTERNATE_COVERAGE ``` Operation priority: ```text alternate native coverage → composition-aware crop → tracked reframe → designed split/stack → controlled background extension → generated outpaint with disclosure/provenance where material ``` ## 7. Vertical-video grammar Vertical frame constraints: ```text less horizontal relationship space more vertical stacking space faces often larger platform UI may occupy top/bottom/right regions captions compete with lower thirds fast subject switching can cause reframing fatigue ``` Rules: ```text IF one active speaker owns the moment THEN center or compositionally offset that speaker with stable tracking. IF two speakers’ relationship is essential THEN use a designed split/stack or a native two-shot rather than frantic panning. IF an object and face must be visible together THEN use stacked composition, alternate angle, or full-screen insert. IF horizontal motion exits the vertical crop THEN select alternate coverage or anticipate the movement with lead room. ``` ## 8. Horizontal-video grammar Horizontal space supports: ```text speaker relationships wide geography side-by-side comparison lower thirds beside subject complex diagrams cinematic movement ``` Do not leave subjects unnecessarily tiny simply because the frame is wide. ## 9. Square and 4:5 grammar Square and 4:5 can preserve more relationship width than 9:16 while occupying feed space. Validate: ```text headroom caption region hands/props lower-third width side-by-side graphics thumbnail crop ``` ## 10. Responsive graphic grammar A graphic component declares: ```text priority minimum size preferred anchor allowed anchors wrap behavior collapse behavior alternate compact form caption collision policy source-note minimum ``` Responsive transformations: ```text side-by-side → stacked full legend → direct labels wide timeline → vertical timeline multi-column table → cards lower third → top card or name tag chart + speaker → chart full-screen with audio carry ``` ## 11. Caption grammar Captions must include: ```text spoken dialogue speaker identification when unclear meaningful non-speech audio relevant music description when needed language changes inaudible/uncertain markers when unavoidable ``` Caption dimensions: ```text verbatim vs edited policy line length line count reading rate minimum/maximum duration synchronization speaker labels position style background/outline safe zone collision avoidance sidecar format burn-in state ``` Rules: ```text IF graphics occupy the caption region THEN reposition graphics or captions; never let essential information overlap. IF a rapid cut breaks caption continuity THEN prioritize readable linguistic units over shot boundaries. IF caption text is edited for readability THEN preserve meaning and do not sanitize materially relevant language without a declared policy. ``` ## 12. Audio-only derivative grammar Before audio export, resolve visual references: ```text “this,” “here,” “as you can see,” silent gestures, on-screen-only labels ``` Repair options: ```text retain descriptive question insert narrator description use alternate answer add brief audio label remove nonessential visual dependence ``` Audio-only deliverables may include: ```text master WAV compressed distribution file chapters cover art show notes transcript speaker metadata loudness report ``` ## 13. Current web-video reference profile A conservative reference profile for widely compatible web delivery: ```text container: MP4 video: H.264, progressive frame_rate: same as source capture when possible audio: AAC-LC or supported platform codec sample_rate: 48 kHz SDR color: documented Rec.709/BT.709 pipeline ``` Platform specifications change. The delivery agent must load the current official profile at export time rather than treating this document as permanent platform policy. ## 14. Mezzanine and archive grammar Preserve a high-quality master with: ```text robust mezzanine codec or image sequence high bit depth full intended resolution stable frame rate documented color space separate audio stems caption sidecars graphic masters project/timeline interchange source manifests checksums provenance records ``` Archive: ```text SOURCE_MEDIA PROJECT_FILES SEMANTIC_TIMELINE OTIO/XML/AAF/EDL AS APPLICABLE FINAL_MASTERS MEZZANINES STEMS CAPTIONS TRANSCRIPTS FONTS/LICENSES REFERENCES GRAPHIC SOURCE FILES PROMPTS/GENERATED ASSET RECORDS QA REPORTS ``` Do not redistribute licensed font files in delivery bundles unless the license and recipient workflow explicitly permit it. ## 15. Timeline interchange grammar The semantic timeline should map to an interchange representation containing: ```text clips source references timing tracks transitions markers metadata ``` OpenTimelineIO is suitable for editorial interchange because it represents editorial timing and references external media rather than embedding the media itself. Adapter support and effect fidelity vary, so round-trip tests are required. ## 16. Version naming grammar ```text [project]_[content-version]_[aspect]_[resolution]_[language]_[caption-state]_[audio-profile]_[date]_[revision] ``` Example: ```text made_ep01_v07_9x16_1080x1920_en_burned_web_20260903_r02.mp4 ``` Maintain immutable published versions and a version manifest. ## 17. Version inheritance ```text MASTER_STORY_VERSION ├── LONG_HORIZONTAL ├── LONG_AUDIO ├── CHAPTERS ├── SHORT_HORIZONTAL ├── SHORT_VERTICAL ├── SQUARE_OR_4X5 └── TEXT/GRAPHIC DERIVATIVES ``` A semantic correction should propagate to every dependent version unless intentionally overridden with a documented reason. ## 18. Thumbnail and poster-frame grammar ```text subject expression/action promise specificity contrast brand identity legibility at small size truthfulness platform crop ``` The thumbnail may compress the premise but may not depict an event, person, or result as factual when it did not occur. ## 19. Encode-QA dimensions ```text file opens correct duration correct start/end no black/frozen unintended frames frame cadence sync color tags range levels HDR metadata bitrate behavior compression artifacts audio channels loudness/true peak caption presence/timing chapter markers metadata checksum ``` ## 20. Multiformat LAKA variations ### Baseline ```text same story and timing; encode for target ``` ### Minor ```text reframe, caption layout, or local graphic reflow ``` ### Major ```text alternate coverage, shorter opening, platform-specific pacing ``` ### Structural ```text chapter extraction, short-form reconstruction, audio-only rewrite ``` ### Paradigm ```text linear episode becomes interactive, spatial, live, or modular media experience ``` ## 21. Delivery failure modes ```text BLIND_CENTER_CROP FACE_OBJECT_COLLISION CAPTION_GRAPHIC_COLLISION PLATFORM_UI_COLLISION FRAME_RATE_CONVERSION_JUDDER COLOR_TAG_MISMATCH RANGE_MISMATCH LOUDNESS_PROFILE_MISMATCH MONO_OR_CHANNEL_FAILURE MISSING_CAPTIONS VISUAL_REFERENCE_IN_AUDIO_ONLY OUTDATED_PLATFORM_PRESET TIMELINE_ROUNDTRIP_LOSS VERSION_PROPAGATION_FAILURE MISSING_ARCHIVE_PROVENANCE ``` ## 22. Acceptance test ```text Does every version preserve the central meaning and promise? Is the information owner visible in each aspect ratio? Do graphics and captions reflow rather than collide? Is audio-only coherent? Does the encode match the current official destination profile? Are color, audio, captions, metadata, and checksums verified? Can the project be reopened and reconstructed from the archive? ```