Repository navigation

v2 - #10

Draft

mihar-22 wants to merge 119 commits into

mainfrom

t3code/improve-caption-rendering

Commits on Sep 2, 2026
- chore: upgrade toolchain to rollup 4, vitest 3, typescript 5.9, prettier 3
  
  mihar-22
  and
  claude
  committed
  677c875
- fix(vtt)!: escape rendered cue HTML and harden tokenizer nesting
  
  mihar-22
  and
  claude
  committed
  ab55dea
- fix(vtt): report reversed cue ranges, keep decimal percentages, parse long hours
  
  mihar-22
  and
  claude
  committed
  67f6aa1
- feat(vtt): add HLS X-TIMESTAMP-MAP helpers
  
  mihar-22
  and
  claude
  committed
  2c705a5
- feat(srt): strip coordinates, map font colours, apply numpad alignment tags
  
  mihar-22
  and
  claude
  committed
  ee70060
- feat(ssa)!: rewrite parser with PlayRes scaling, override tags, and embedded fonts
  
  mihar-22
  and
  claude
  committed
  d55c63c
- feat: add TTML/IMSC1, SCC (CEA-608), LRC, and SBV parsers
  
  mihar-22
  and
  claude
  committed
  f3e5c2f
- feat(renderer): sorted cue index, identity diffing, frame-accurate sync, faster collisions
  
  mihar-22
  and
  claude
  committed
  c998c04
- feat(styles): balanced wrapping, container units, text-stroke outlines, edge-style presets
  
  mihar-22
  and
  claude
  committed
  21e240e
- docs: document new formats, renderer APIs, styling, and security model
  
  mihar-22
  and
  claude
  committed
  7db9948
- fix(renderer): apply cue position horizontally and size fonts from the cue box
  
  mihar-22
  and
  claude
  committed
  6ff0657
- test: add Vitest 4 browser mode with Playwright and real-layout positioning tests
  
  mihar-22
  and
  claude
  committed
  9140499
- chore: drop misplaced pure annotations
  
  mihar-22
  and
  claude
  committed
  c0a9899
- build: replace rollup with tsdown
  
  mihar-22
  and
  claude
  committed
  0b060e2
- fix(renderer): transform-aware collision boxes, fixed cues, reading-order stacking
  
  mihar-22
  and
  claude
  committed
  6cc4878
- feat(ssa): size cue boxes to their text and pin positioned cues
  
  mihar-22
  and
  claude
  committed
  e447158
- test(browser): cover transforms, fixed cues, and SSA layout in Chromium
  
  mihar-22
  and
  claude
  committed
  88e269b
- docs: regenerate README screenshots from sandbox scenarios
  
  mihar-22
  and
  claude
  committed
  7e8967f
- feat(vtt): support STYLE blocks with scoped, sanitized ::cue styling
  
  mihar-22
  and
  claude
  committed
  85869ae
- fix(vtt)!: spec-conformant file parsing with explicit real-world tolerances
  
  mihar-22
  and
  claude
  committed
  fe52207
- feat(scc): decode CC2 via the channel option
  
  mihar-22
  and
  claude
  committed
  f026f88
- feat(ttml): cell units, font sizes, vertical writing modes, sibling end inference
  
  mihar-22
  and
  claude
  committed
  a27e262
- fix(renderer): spec-conformant line placement for vertical, negative, percentage, and RTL cues
  
  mihar-22
  and
  claude
  committed
  26acd00
- feat(renderer): event-driven sync mode driven by native TextTrack cue boundaries
  
  mihar-22
  and
  claude
  committed
  98ce53d
- fix(ssa): honour \1c colour tags and keep toggled formatting properly nested
  
  mihar-22
  and
  claude
  committed
  b5d486d
- docs: document STYLE blocks, strict vs tolerant parsing, sync modes, and development workflow
  
  mihar-22
  and
  claude
  committed
  a3d9de1
- test(browser): screenshot-based visual regression for representative scenarios
  
  mihar-22
  and
  claude
  committed
  c700e99
- chore: ignore vitest attachments
  
  mihar-22
  and
  claude
  committed
  55f871d
- feat(cea): stream-oriented CEA-608 decoder with field 2 (CC3/CC4) and cc_data parsing
  
  mihar-22
  and
  claude
  committed
  34ed96f
- feat(cea): CEA-708 (DTVCC) decoder
  
  mihar-22
  and
  claude
  committed
  e22e32a
- docs: document CEA-608/708 stream decoding and visual regression testing
  
  mihar-22
  and
  claude
  committed
  4ed8285
- refactor(renderer): split positioning into measure, layout, and write phases
  
  mihar-22
  and
  claude
  committed
  ddc9b8b
- feat(cue)!: structured layout and text style model on VTTCue with JSON round-tripping
  
  mihar-22
  and
  claude
  committed
  91b5a52
- feat(element): <media-captions> custom element
  
  mihar-22
  and
  claude
  committed
  9cf1f9c
- feat(track): CueTrack with incremental sorted index, mutable end times, eviction, and events
  
  mihar-22
  and
  claude
  committed
  e8c39b3
- feat(renderer): render from a CueTrack with live updates and retention-based eviction
  
  mihar-22
  and
  claude
  committed
  dd362ee
- feat(renderer): build cue DOM directly, dispatch enter/exit events, optional screen reader announcer
  
  mihar-22
  and
  claude
  committed
  a9f94b2
- feat(styles)!: cascade layer, registered custom properties, and per-format parser entries
  
  mihar-22
  and
  claude
  committed
  9613e63
- feat(cea): live mode emitting open-ended cues that are updated in place
  
  mihar-22
  and
  claude
  committed
  b7b3b71
- docs: layout model, CueTrack, custom element, cue events, announcer, layers, parser entries
  
  mihar-22
  and
  claude
  committed
  5f41c74
- test(wpt): vendor the web-platform-tests WebVTT parsing suites
  
  mihar-22
  and
  claude
  committed
  671bd0e
- fix(vtt)!: resolve every parser bug surfaced by the WPT suite
  
  mihar-22
  and
  claude
  committed
  8ddfc26
- docs: describe the WPT conformance harness
  
  mihar-22
  and
  claude
  committed
  d186c31
- feat(cue): layout height, background images, animations, class names, and an entity registry hook
  
  mihar-22
  and
  claude
  committed
  77062e1
- ci: test, type-check, and build on pull requests; manual workflow to record Linux visual baselines
  
  mihar-22
  and
  claude
  committed
  c21d1b1
- fix(cue): support open-ended (Infinity) end times on the native VTTCue
  
  mihar-22
  and
  claude
  committed
  917998a
- docs(sandbox): layout model, live track, and custom element scenarios with screenshots
  
  mihar-22
  and
  claude
  committed
  453f30e
- feat(entities): optional full HTML named character reference table
  
  mihar-22
  and
  claude
  committed
  6baa190
- docs: entities entry, open-ended cues, sandbox scenarios, and CI
  
  mihar-22
  and
  claude
  committed
  c3c7d6b
- fix(cue): keep Infinity end times intact on the fallback cue class
  
  mihar-22
  and
  claude
  committed
  a346195
- feat(cea): CEA-708 pen sizes, edges, window fill and borders, print and scroll directions, display effects, word wrap
  
  mihar-22
  and
  claude
  committed
  67bf627
- feat(ttml): set animations, image cues, wall-clock time base, drop-frame SMPTE times
  
  mihar-22
  and
  claude
  committed
  1bb296a
- docs: TTML animations, images, clock time base; CEA-708 fidelity
  
  mihar-22
  and
  claude
  committed
  b36c973
- style: format workflows and add .prettierignore for vendored and generated files
  
  mihar-22
  and
  claude
  committed
  c98cc90
- style: format remaining files
  
  mihar-22
  and
  claude
  committed
  55085d8
Commits on Sep 3, 2026
- build: adopt Vite+ (oxfmt, oxlint, vitest, tsdown) as the toolchain
  
  mihar-22
  and
  claude
  committed
  38e6c56
Commits on Sep 7, 2026
- feat(cue): per-span styles, inline drawings, clip paths, and an animation model
  
  mihar-22
  and
  claude
  committed
  24cf263
- feat(renderer): media-synced animations, transform-aware collisions, stacking/lineStep/safeArea options
  
  mihar-22
  and
  claude
  committed
  a50b1f2
- docs: renderer options, span styles, drawings, and animation model; CSS hooks for small caps and forced cues
  
  mihar-22
  and
  claude
  committed
  4274edb
- feat(cea): 708 pen backgrounds, font tags, horizontal scroll, wipe directions, ticker windows
  
  mihar-22
  and
  claude
  committed
  838850d
- feat(ssa): typesetting fidelity: per-span tags, fades, move, \t, karaoke sweeps, drawings, clips, effects
  
  mihar-22
  and
  claude
  committed
  ee8b606
- docs: SSA/ASS tag coverage after the typesetting work
  
  mihar-22
  and
  claude
  committed
  26a4fe3
- feat(ttml): container animations, span styles, dropPAL, forced display, padding, and IMSC fixes
  
  mihar-22
  and
  claude
  committed
  213f4b0
- test(imsc): vendor 62 W3C IMSC test documents and check parsing conformance
  
  mihar-22
  and
  claude
  committed
  e93a4e7
- docs: TTML/IMSC coverage and test suites
  
  mihar-22
  and
  claude
  committed
  a5e8005
- feat(track): dedupe option for cues repeated across HLS/DASH segments
  
  mihar-22
  and
  claude
  committed
  12192d9
- perf(tokenizer): cache tokens per cue, invalidated on text or span changes
  
  mihar-22
  and
  claude
  committed
  5c29485
- feat(renderer): reduced-motion support and user preset attributes
  
  mihar-22
  and
  claude
  committed
  98232a1
- ci: size budgets, publint and attw package checks, coverage, API docs, release workflow
  
  mihar-22
  and
  claude
  committed
  1204074
- docs: reduced motion, caption presets, track dedupe, and the expanded CI/tooling scripts
  
  mihar-22
  and
  claude
  committed
  e1073fb
- refactor(renderer): split ordering, animation sync, announcer, and style injection into modules
  
  mihar-22
  and
  claude
  committed
  fa0b88e
- feat: add SAMI and MicroDVD parsers
  
  mihar-22
  and
  claude
  committed
  f8be5ea
- feat(mp4): demux fMP4 wvtt and stpp subtitle tracks into cues
  
  mihar-22
  and
  claude
  committed
  af43fe5
- test(bench): parse, tokenize, track, layout, and CEA decode benchmarks with a recorded baseline
  
  mihar-22
  and
  claude
  committed
  ddb95ff
- perf: O(1) CueTrack membership and 708 commits only after a completed packet
  
  mihar-22
  and
  claude
  committed
  918cee1
- test(corpus): real-world style fixtures for every format with reviewable JSON snapshots
  
  mihar-22
  and
  claude
  committed
  4cbeaa7
- fix(parsers)!: harden against hostile input found by fuzzing
  
  mihar-22
  and
  claude
  committed
  f0f35c0
- chore: exclude corpus fixtures and snapshots from formatting and linting
  
  mihar-22
  and
  claude
  committed
  98fe852
- chore: enable noImplicitAny across the codebase
  
  mihar-22
  and
  claude
  committed
  30e2e0c
- feat(playground): interactive visual playground for every parser and renderer feature
  
  mihar-22
  and
  claude
  committed
  8231415
- fix(vtt)!: drop the non-spec settings-on-a-separate-line tolerance
  
  mihar-22
  and
  claude
  committed
  e3ca69a
- fix(ssa): lazy tag emission and screen-fixed \clip rectangles on layout-positioned cues
  
  mihar-22
  and
  claude
  committed
  ed315f7
- fix(styles): do not paint-contain cue boxes so rotated, scaled, and animated text is not clipped
  
  mihar-22
  and
  claude
  committed
  a0ca8f0
- fix(ssa): rotate and scale around the alignment anchor or \org
  
  mihar-22
  and
  claude
  committed
  5f47863
- fix(playground): make the SSA \clip sample visibly slice the glyphs
  
  mihar-22
  and
  claude
  committed
  548a680
- fix(vtt): assigning a region no longer throws in Firefox and WebKit
  
  mihar-22
  and
  claude
  committed
  11782f6
- test(browser): run the layout suites in Firefox and WebKit too
  
  mihar-22
  and
  claude
  committed
  4b40269
- feat(renderer): composable core with pluggable features and a media-captions/renderer entry
  
  mihar-22
  and
  claude
  committed
  179f1ee
- build: size budgets from tree-shaken usage probes
  
  mihar-22
  and
  claude
  committed
  081f242
- feat(playground): renderer feature presets
  
  mihar-22
  and
  claude
  committed
  202e35b
- fix(styles): fit image cues to their box instead of tiling
  
  mihar-22
  and
  claude
  committed
  2c0ab18
- feat(canvas): headless measurement and a canvas writer (media-captions/canvas)
  
  mihar-22
  and
  claude
  committed
  0ba6707
- docs: canvas exploration findings and a typed style model proposal
  
  mihar-22
  and
  claude
  committed
  adfc2e2
- feat(playground): canvas view
  
  mihar-22
  and
  claude
  committed
  3689a80
- feat(model)!: typed cue style model, writers serialise it
  
  mihar-22
  and
  claude
  committed
  100fab4
- docs: record the typed style model outcome
  
  mihar-22
  and
  claude
  committed
  d8b9cd9
- fix(renderer): keep vertical cues in place when the overlay is re-laid out
  
  mihar-22
  and
  claude
  committed
  c347499
- feat(canvas): vertical writing modes
  
  mihar-22
  and
  claude
  committed
  910519c
- feat(playground): picture-in-picture and fullscreen through a composited captureStream
  
  mihar-22
  and
  claude
  committed
  522f12f
- docs: migration guide from 1.x, PiP compositing, and vertical text notes
  
  mihar-22
  and
  claude
  committed
  bc04599
- ci: declare the pnpm version so pnpm/action-setup can install it
  
  mihar-22
  and
  claude
  committed
  cb90180
- ci: record the pnpm overrides in the lockfile
  
  mihar-22
  and
  claude
  committed
  d8e7fbd
- chore: format package.json
  
  mihar-22
  and
  claude
  committed
  73d0898
- test(fuzz): loosen the parse budget on CI runners
  
  mihar-22
  and
  claude
  committed
  ec9df1b
- ci: run the docs script explicitly and document the new entries
  
  mihar-22
  and
  claude
  committed
  36965e9
- feat(canvas): ruby annotations and orthographic 3D rotations
  
  mihar-22
  and
  claude
  committed
  46dfe50
- fix(vtt): ignore mismatched end tags as the spec does
  
  mihar-22
  and
  claude
  committed
  793a1c9
- feat(parse): add lenient option and run the WPT suite spec-exact
  
  mihar-22
  and
  claude
  committed
  f1d88ea
- ci: record linux visual baselines as an artifact
  
  mihar-22
  and
  claude
  committed
  637ab62
- test(browser): add linux visual baselines and run the suite in CI
  
  mihar-22
  and
  claude
  committed
  5a4e797
- docs: lenient parsing, spec end tags, canvas ruby and 3D, CI visual suite
  
  mihar-22
  and
  claude
  committed
  0da8071
- fix(canvas): let ruby annotations overflow the line box like browsers
  
  mihar-22
  and
  claude
  committed
  af00e53
- test(browser): ruby box agreement is engine-aware
  
  mihar-22
  and
  claude
  committed
  eb613e7
- fix(playground): drive the canvas view on every time change
  
  mihar-22
  and
  claude
  committed
  c601b68
- fix(canvas): composite cue opacity as a group, like CSS opacity
  
  mihar-22
  and
  claude
  committed
  2437af2
Commits on Sep 9, 2026
- fix(renderer): compute region boxes from their anchors
  
  mihar-22
  and
  claude
  committed
  f47c25c
- feat(canvas): match the DOM on region scroll, inline sizes, and line breaks
  
  mihar-22
  and
  claude
  committed
  93a9ce2
- test(browser): DOM/canvas parity sweep over every playground sample
  
  mihar-22
  and
  claude
  committed
  8a984b7
- test(browser): compare the alignment edge in the parity sweep
  
  mihar-22
  and
  claude
  committed
  659bb05

Commits

Commits on Sep 2, 2026

Commits on Sep 3, 2026

Commits on Sep 7, 2026

Commits on Sep 9, 2026