Paper Collage Video v0.19.0

Paper Collage Video 0.19.0

Paper Collage Video 0.19.0 turns visual style, whole-film motion direction,
three-dimensional path locomotion, persistent travelling worlds, and provider
recovery into one evidence-bound Remotion production contract.

Executable visual and motion direction

• Project and Storyboard v12 use a dynamic Style Catalog and executable Style
  Profile v2. Surface treatment, transition recipes, visual SFX limits, and
  required asset checks are frozen into the same profile used by generation,
  rendering, quality review, and fingerprint invalidation.
• A compiler-owned whole-film Motion Contract replaces descriptive motion
  prose. It binds performance roles, pacing, camera language, transitions,
  ambient motion, final holds, exceptions, events, and proof moments.
• The existing style and fictional-voice review now includes a
  motion-language-card.json and records the exact approved motion
  fingerprints without adding another routine human gate.
• Comic-inspired output remains ordinary shot-based video. Dialogue,
  narration, and exposition stay in audio plus subtitles; short editable
  visual SFX are allowed only for a bound discrete impact or motion sound.

Three-dimensional path locomotion

• Registered limited-animation subjects can follow one cubic Bézier
  path-locomotion route through screen space and optical depth.
• Runtime travel uses physical-pixel arc length, tangent auto-orientation,
  angle unwrapping, smoothing, bounded turn rate, projection scale, dynamic
  depth ordering, and velocity-driven planar/toward/away state loops across
  16:9, 9:16, and 1:1.
• Optional camera.follow consumes the same resolved path and stays clamped
  inside one declared coherent world surface. Duplicated camera or scale
  keyframes and background swaps are rejected.
• The spatial contract proves screen and depth travel, direction sectors,
  heading error, turn continuity, gait cadence, view-state binding, camera
  binding, world binding, and viewport coverage with start/turn/end evidence.

Persistent worlds, encounters, and semantic depth

• Looping worlds now have provider-free topology proof before generation,
  source- and render-scale seam evidence, truthful visible surface roles,
  sparse foreground scatter, and camera-compensated world-motion proof.
• Encounter contracts bind one narration cue and one world-anchored target to
  ordered enter, approach, answer, and exit phases. Visibility toggles cannot
  teleport a participant through the lifecycle.
• Semantic depth slices, registered scene families, trajectory contracts, and
  adjacent-scene continuity proofs preserve identity, grounding, occlusion,
  relative order, and causal travel across finite and repeating worlds.

Provider budget, recovery, and asset integrity

• Provider-native results are recorded unchanged before deterministic canvas
  normalization, source-rectangle extraction, registration, keying, masking,
  or other local derivatives.
• Registered state and layer sheets support observed chroma-key planes,
  bounded baked-checkerboard alpha recovery, full-sheet masked repair, and
  deterministic horizontal mirroring while preserving accepted cells and one
  shared registration canvas.
• A human-approved image-attempt cap remains smaller than or equal to the
  production profile ceiling. Any later increase is an explicit append-only
  old/new/usage audit event; rejected, abandoned, and unused provider results
  still count when quota was consumed.
• Provider reservation, recording, and recovery enforce one canonical
  provider/model identity before provenance is written.

Runtime identity and lifecycle hardening

• runtime-build.json recursively fingerprints the actual local ESM dependency
  closure, schemas, fixtures, Remotion entrypoints, and generated package
  surfaces across source and packaged workspaces.
• The unreachable legacy publish-approval stage is removed. complete
  remains the local production terminal state; approve-publish is an
  optional post-completion audit event for one explicitly authorized
  destination, action, and scope.
• The Remotion build chain pins transitive fast-uri to patched version
  3.1.5, and source/package tests enforce that minimum against
  CVE-2026-18446.
• The latest release supports only the current project, storyboard, asset,
  proof, and production contracts. Old project files are intentionally not
  migrated through compatibility adapters; rebuild an older production as a
  fresh current-contract project when it needs revision.

Release validation

• Source workspace: 307/307 tests passed.
• Fresh packaged workspace: 289/289 tests passed; doctor reported READY,
  TypeScript passed, and starter-demo validated with 0 errors / 0 warnings.
• The fresh packaged starter-demo rendered a real 36-frame H.264/AAC preview,
  passed 7/7 quality checks, and reached the local complete state.
• TypeScript, Remotion bundle, schema-v12 validation, and the moderate-level npm
  audit passed; the audit reported zero vulnerabilities.
• Registered-family and alpha-band proofs passed. The path-locomotion proof
  rendered and verified 16:9, 9:16, and 1:1 previews plus a 16:9 final artifact,
  with zero provider calls.
• Source, packaged template, and fresh workspace share runtime fingerprint
  1ab31994036d26e7bc17c1fa0bcffd3a13fa6629f7effdb099d391ceb0c99594.

Provider and publication boundary

No image, voice, or video provider call is required to build or validate this
release. The release contains reusable source, deterministic fixtures, and the
packaged Codex Plugin. It does not publish user projects, credentials, private
prompts, or generated production media, and it carries no new media attachment.

Install

codex plugin marketplace add cyberlesterr/paper-collage-video
codex plugin add paper-collage-video@paper-collage-video

Start a new Codex task after installation so the versioned Skill snapshot and
packaged runtime come from the same release.

Requirements and licensing

• Node.js 20 or newer
• FFmpeg and ffprobe
• Python 3.11 or newer
• macOS and Ubuntu are validated; Windows support remains best effort

The software is MIT-licensed. Bundled fixture and style-catalog media use the
repository-demo-only terms in ASSET_LICENSES.md. Remotion and other
dependencies retain their own licenses as described in
THIRD_PARTY_NOTICES.md.

Paper Collage Video v0.16.0

Paper Collage Video 0.16.0

Paper Collage Video 0.16.0 turns the repository into a stricter, evidence-bound
production system for editable Remotion paper-collage films. It closes the
highest-priority failures found during real story production without weakening
human approval, provider budget, rights, preview, or publication gates.

Production planning and budget control

• New projects first choose aspect ratio, one of three bundled visual directions,
  and a parallax preference, then compare complete draft, balanced, and
  full-depth production scenarios.
• Each scenario states story scope, motion language, depth plan, state families,
  provider recommendation, expected image calls, proposed approved cap, local
  derivatives, avoided calls, and final-film tradeoffs.
• A production profile is only a planning ceiling. Provider attempts are
  enforced against the smaller exact cap approved by the human.
• Story-critical semantic actions must resolve to actual registered states,
  local-motion targets, or layer source packages before they count as covered.

Layer-complete assets and registered animation

• Project and Storyboard schema v10 plan complete rear, subject, and front source
  packages before any layer-aware provider request.
• Registered depth stacks preserve one coordinate system, stable depth order,
  complete hidden content, and explicit reveal envelopes for 16:9, 9:16, and
  1:1.
• Identity-bound state sheets carry one active identity reference plus per-state
  facing and anchor declarations. Deterministic processing retains a shared
  canvas and produces fingerprinted anchor overlays for quality review.
• Provider-native mixed-surface sheets support opaque reference/rear cells and
  alpha or observed chroma-key subject/front cells without rewriting the
  provider root.
• Quota-consuming calls remain distinct from deterministic crops, keying,
  registration, masks, looping derivatives, and other local work.

Persistent worlds and causal motion

• looping-environment provides far, mid, walkable-ground, and near strips with
  seamless source/render proof and real visible-surface validation.
• Scenes can bind one tracked subject and multiple participants using explicit
  screen- or world-space anchoring and near-layer occlusion relationships.
• Ground displacement is shared by world-anchored subjects, so markers and
  participants stay attached to the travelling world.
• Signed trajectory contracts prove forward travel, relative order, overtaking,
  state changes, offscreen exits, finish order, and final stops instead of
  relying on prose or unsigned distance.

Editorial, timing, and revision contracts

• Editorial v9 binds narration words, phrases, sentences, emphasis, SFX phases,
  music beats, and manual cues to actual local media timing.
• Editable typography, annotations, counters, charts, tables, maps, diagrams,
  responsive directing, and semantic paper transitions remain React/SVG
  primitives rather than baked raster UI.
• motionPolicy=locked-static protects deliberate still scenes from profile
  motion floors.
• A human-authorized semantic revision can change listed scenes while preserving
  the approved provider cap and invalidating all dependent style, proof, preview,
  and final evidence.

Evidence-bound quality and delivery

• Composition, asset, subtitle, transition, relationship, and semantic-contract
  review surfaces have independent fingerprints and lifecycle rules.
• Quality contact sheets bind the exact scaffold, targets, evidence files, and
  report hash; stale or cross-surface approvals are rejected.
• State identity, anchor drift, facing, alpha bands, chroma-key residue, depth
  reconstruction, visible world surfaces, actor grounding, occlusion, and signed
  world direction are deterministic quality checks.
• project:assets-ready creates one seal over current assets, audio, subtitles,
  timeline, quality, and proof state. Preview and final rendering reject a stale
  seal.
• Subtitle delivery is verified from encoded frames independently of the
  subtitle-free composition review surface.

Dependency maintenance

• remotion and @remotion/cli are pinned together at 4.0.499.
• PostCSS is updated to the compatible 8.5.23 patch.
• fast-uri remains on the patched 3.1.4 line used by the current dependency
  graph.
• These unified updates supersede the older automated dependency PRs that
  proposed partial or stale versions.

Compatibility

• 0.16.0 supports only the latest project, storyboard, asset, evidence, and
  production contracts.
• Old projects are intentionally not migrated or loaded through compatibility
  adapters. Rebuild an older production as a fresh latest-contract project when
  it needs revision.
• The output capability remains: editable layered collage stories, recurring
  identities, registered limited animation, persistent travelling worlds,
  functional diagrams, actual-audio editing, responsive typography, subtitles,
  preview review, and locally accepted final delivery.

Release validation

• Source workspace: 204/204 tests passed.
• Fresh packaged workspace: 189/189 tests passed.
• TypeScript, doctor, Remotion bundle, schema-v10, and npm moderate-level audit
  passed; the audit reported zero vulnerabilities.
• Registered-family and alpha-band proofs passed with zero provider calls.
• The looping-world proof rendered and verified 16:9, 9:16, and 1:1 previews
  plus a 16:9 final artifact under Remotion 4.0.499.
• A separately installed packaged workspace reached doctor READY and rendered
  the sealed starter preview with current validation, quality, audio, subtitle,
  contact-sheet, and encoded-frame evidence.

Provider and publication boundary

No image, voice, or video provider call is required to build or validate this
release. The release contains reusable source, deterministic fixtures, and the
packaged Codex Plugin. It does not publish user projects, credentials, private
prompts, or generated production media, and this release intentionally carries
no media attachment.

Install

codex plugin marketplace add cyberlesterr/paper-collage-video
codex plugin add paper-collage-video@paper-collage-video

Start a new Codex task after installation so the versioned Skill snapshot and
packaged runtime come from the same release.

Requirements and licensing

• Node.js 20 or newer
• FFmpeg and ffprobe
• Python 3.11 or newer
• macOS and Ubuntu are validated; Windows support remains best effort

The software is MIT-licensed. Bundled fixture and style-catalog media use the
repository-demo-only terms in ASSET_LICENSES.md. Remotion and other
dependencies retain their own licenses as described in
THIRD_PARTY_NOTICES.md.

Paper Collage Video v0.8.0

Paper Collage Video v0.8.0

This release makes semantic correctness and composition relationships explicit
parts of the editable Remotion production contract instead of leaving them to
image-generation prompts.

Highlights

• Installable Codex Plugin that bootstraps an independent Remotion workspace.
• Rhythmic storyboards with authored scene blueprints, beats, cues, keyframes,
  and proof times.
• Composition Contract v4 with registered supported-subject and
  registered-environment groups, shared canvases, source masters, relationship
  proofs, and stale-review invalidation.
• Reusable identity, topology, mechanism, and diagram contracts. Coexisting
  characters require structural differences; working objects require complete
  parts, connections, load paths, motion constraints, and reference evidence.
• Schema-v3 image requests with semantic bindings that combine every applicable
  contract instead of checking only one dominant risk.
• An append-only image-generation attempt ledger with atomic budget reservation,
  truthful rejected/abandoned accounting, exact reuse, and final-canvas size
  validation.
• Deterministic rejection of procedural SVG filters on diagram text, arrows,
  icons, and borders.
• Three default human decisions: concept/provider/budget, style/fictional voice,
  and rendered preview. Local final delivery remains separate from publication.
• A lightweight two-second fixture used only for installation smoke tests;
  production projects and media are not bundled into the Plugin package.

Compatibility

• New projects use project schema v4; v3 and earlier projects are intentionally
  not migrated or silently downgraded.
• Ledger-enabled projects require schema-v3 image requests.
• Start a new Codex task after upgrading so the v0.8.0 Skill snapshot is loaded.

This release contains reusable source, the packaged Plugin, and technical test
fixtures only. It does not publish production project media or add a new manual
Release attachment. The older v0.5.0 repository demo remains available under
the limited terms in ASSET_LICENSES.md.

Install

Give Codex the repository URL, or run:

codex plugin marketplace add cyberlesterr/paper-collage-video
codex plugin add paper-collage-video@paper-collage-video

Start a new Codex task and invoke $make-paper-collage-video. The Plugin will
create a writable workspace outside its installation cache and run environment
diagnostics before production begins.

Requirements and Support

• Node.js 20 or newer
• FFmpeg and ffprobe
• Python 3.11 or newer for image processing and proof generation
• macOS and Ubuntu are validated; Windows support is currently best effort

Image and voice generation require capabilities or provider accounts selected
by the user. The project does not publish generated media automatically.

Licensing

The software is MIT-licensed. Bundled showcase and fixture media is excluded
from MIT and is limited to repository demonstration, testing, and evaluation;
see ASSET_LICENSES.md. Remotion and other dependencies retain their own
licenses, including Remotion's Company License requirements in some commercial
use cases; see THIRD_PARTY_NOTICES.md.

Paper Collage Video v0.6.0

Paper Collage Video v0.5.0

Second public preview of the configuration-driven paper-collage video pipeline
and Codex Plugin.

Highlights

• Installable Codex Plugin that bootstraps an independent Remotion workspace.
• A clean v2 project protocol with seconds-only timeline authoring and no v1
  migration or compatibility branches.
• Mandatory hash-bound technical and semantic image quality review.
• Environment depth layers, camera keyframes, per-character narrative motion,
  scene transitions, action audio, and project-local fonts.
• Adaptive story planning when duration, scene count, both, or neither is given.
• Configurable text, image, and fictional-voice providers with provenance
  manifests and resumable work items.
• Capability selection plus four explicit human approvals for concept,
  style/voice, preview, and final publication; automatic stages remain
  resumable.
• Project validation, narration synchronization, subtitle alignment, LUFS and
  true-peak reporting, CPU-aware rendering, and scene-aware contact sheets.
• A lightweight two-second fixture used only for installation smoke tests;
  production projects and media are not bundled into the Plugin package.

Complete Repository Demo

The Release has one media attachment only:
Tie Chu Mo Zhen — 77.7-second 1080p final film.
It demonstrates the v2 quality gate, six-scene timeline, depth and narrative
motion, subtitles, fictional narration, action sound effects, and delivery
checks. Preview renders, motion proofs, contact sheets, and source assets are
not attached.

The demo is repository-demo-only material rather than an MIT-licensed media
asset. See ASSET_LICENSES.md before redistributing or reusing it.

Install

Give Codex the repository URL, or run:

codex plugin marketplace add cyberlesterr/paper-collage-video
codex plugin add paper-collage-video@paper-collage-video

Start a new Codex task and invoke $make-paper-collage-video. The Plugin will
create a writable workspace outside its installation cache and run environment
diagnostics before production begins.

Requirements and Support

• Node.js 20 or newer
• FFmpeg and ffprobe
• Python 3.11 or newer for chroma-key character-sheet processing
• macOS and Ubuntu are validated; Windows support is currently best effort

Image and voice generation require capabilities or provider accounts selected
by the user. The project does not publish media automatically.

Licensing

The software is MIT-licensed. Bundled showcase and fixture media is excluded
from MIT and is limited to repository demonstration, testing, and evaluation;
see ASSET_LICENSES.md. Remotion and other dependencies retain their own
licenses, including Remotion's Company License requirements in some commercial
use cases; see THIRD_PARTY_NOTICES.md.

Paper Collage Video v0.5.0

Paper Collage Video v0.5.0

Second public preview of the configuration-driven paper-collage video pipeline
and Codex Plugin.

Highlights

• Installable Codex Plugin that bootstraps an independent Remotion workspace.
• A clean v2 project protocol with seconds-only timeline authoring and no v1
  migration or compatibility branches.
• Mandatory hash-bound technical and semantic image quality review.
• Environment depth layers, camera keyframes, per-character narrative motion,
  scene transitions, action audio, and project-local fonts.
• Adaptive story planning when duration, scene count, both, or neither is given.
• Configurable text, image, and fictional-voice providers with provenance
  manifests and resumable work items.
• Capability selection plus four explicit human approvals for concept,
  style/voice, preview, and final publication; automatic stages remain
  resumable.
• Project validation, narration synchronization, subtitle alignment, LUFS and
  true-peak reporting, CPU-aware rendering, and scene-aware contact sheets.
• A lightweight two-second fixture used only for installation smoke tests;
  production projects and media are not bundled into the Plugin package.

Complete Repository Demo

The Release has one media attachment only:
Tie Chu Mo Zhen — 77.7-second 1080p final film.
It demonstrates the v2 quality gate, six-scene timeline, depth and narrative
motion, subtitles, fictional narration, action sound effects, and delivery
checks. Preview renders, motion proofs, contact sheets, and source assets are
not attached.

The demo is repository-demo-only material rather than an MIT-licensed media
asset. See ASSET_LICENSES.md before redistributing or reusing it.

Install

Give Codex the repository URL, or run:

codex plugin marketplace add cyberlesterr/paper-collage-video
codex plugin add paper-collage-video@paper-collage-video

Start a new Codex task and invoke $make-paper-collage-video. The Plugin will
create a writable workspace outside its installation cache and run environment
diagnostics before production begins.

Requirements and Support

• Node.js 20 or newer
• FFmpeg and ffprobe
• Python 3.11 or newer for chroma-key character-sheet processing
• macOS and Ubuntu are validated; Windows support is currently best effort

Image and voice generation require capabilities or provider accounts selected
by the user. The project does not publish media automatically.

Licensing

The software is MIT-licensed. Bundled showcase and fixture media is excluded
from MIT and is limited to repository demonstration, testing, and evaluation;
see ASSET_LICENSES.md. Remotion and other dependencies retain their own
licenses, including Remotion's Company License requirements in some commercial
use cases; see THIRD_PARTY_NOTICES.md.

Paper Collage Video v0.4.0

Paper Collage Video v0.4.0

First public preview release of the configuration-driven paper-collage video
pipeline and Codex Plugin.

Highlights

• Installable Codex Plugin that bootstraps an independent Remotion workspace.
• Adaptive story planning when duration, scene count, both, or neither is given.
• Configurable text, image, and fictional-voice providers with provenance
  manifests and resumable work items.
• Capability selection plus four explicit human approvals for concept,
  style/voice, preview, and final publication; automatic stages remain
  resumable.
• Project validation, narration synchronization, CPU-aware preview rendering,
  technical reports, and scene-aware contact sheets.
• One complete four-scene showcase, tie-chu-mo-zhen, plus a lightweight
  two-second fixture used only for installation smoke tests.

Install

Give Codex the repository URL, or run:

codex plugin marketplace add cyberlesterr/paper-collage-video
codex plugin add paper-collage-video@paper-collage-video

Start a new Codex task and invoke $make-paper-collage-video. The Plugin will
create a writable workspace outside its installation cache and run environment
diagnostics before production begins.

Requirements and Support

• Node.js 20 or newer
• FFmpeg and ffprobe
• Python 3.11 or newer for chroma-key character-sheet processing
• macOS and Ubuntu are validated; Windows support is currently best effort

Image and voice generation require capabilities or provider accounts selected
by the user. The project does not publish media automatically.

Licensing

The software is MIT-licensed. Bundled showcase and fixture media is excluded
from MIT and is limited to repository demonstration, testing, and evaluation;
see ASSET_LICENSES.md. Remotion and other dependencies retain their own
licenses, including Remotion's Company License requirements in some commercial
use cases; see THIRD_PARTY_NOTICES.md.