feat(paper2video): group wrapped-line cues + card-aware ink-tighten spotlight

[LingAo Xiao (ArmandXiao)]

Summary

The video spotlight previously drew the shape's raw declared box (PPTX off/ext or a semantic estimate), so it either dropped a wrapped title's second line or spotlighted a lot of empty leading/padding. This makes the spotlight track the intended logical unit — tightly for text, fully for cards — via two composed stages:

1. Wrapped-line grouping (generate_visual_cues.py) — before emitting a cue, union same-group single-line title/tagline fragments into the whole wrapped run, so a cue that matched one line spotlights the whole title/tagline. Multi-line paragraph blocks are never merged, preserving the script's per-chunk progression. Guarded by VIDEO_CUE_GROUP_WRAPPED_LINES + geometry knobs (WRAP_X_EPS, WRAP_MIN_ASPECT, WRAP_GAP_FRAC).
2. Card-aware ink-tighten (render_video.py) — shrink each spotlight box to the painted ink (glyphs) for transparent text so it hugs the content, but keep a filled card/panel whole via robust detection: page bg = the dominant colour over a coarse grid of the source slide (robust to a corner accent bar, unlike sampling the four corners), and the box is a card when most of its border ring hugs one fill colour (close_frac, which survives a thin accent bar on one edge) that differs from the page bg. Feather/dim + card thresholds are env-tunable (same defaults); best-effort — degrades to the declared box without Pillow/numpy or on low confidence.

Architecture: grouping decides WHAT unit to highlight (structure domain); ink-tighten decides HOW tightly + preserves filled cards (pixel domain); per-chunk progression is untouched.

Files changed

• ResearchStudio-Reel/skills/paper2video/scripts/render_video.py (+112): new _ink_tighten_box() (letterbox-aware, card detection), VisualCue.no_ink_tighten override, env-tunable feather/dim + card thresholds, wired into encode_segment's spotlight loop.
• ResearchStudio-Reel/skills/paper2video/scripts/generate_visual_cues.py (+91): new union_wrapped_line_cues() + helpers, called once per slide before appending cues.

Validation (new vs upstream-main render, identical deck+audio+cues)

• Loose/text deck (2607.05394) — a 2-line tagline split across boxes + loose labels + filled takeaway card: 23/23 spotlight frames change, avg 5.08% pixels. Tagline now covers both lines tightly, filled cards stay fully lit (right padding no longer dimmed), loose labels hug their glyphs.
• Card-heavy deck (2607.04438, 10-min, 38 cues) — already-tight boxes: 37/38 frames byte-identical. Card-aware correctly preserves all 8 filled cards (no "hole in the card" regression), 1 minor text tighten. Confirms this is a backward-compatible superset — it only changes a box when there is something loose to hug or a card to preserve.

Known follow-up (not in this PR)

VIDEO_SPOTLIGHT_FEATHER_PX alone is dominated by the ratio term (min(1920,1080) * 0.052 = 56), so setting only FEATHER_PX does not change the feather. Making FEATHER_PX authoritative (and picking a tighter default) is a small follow-up.

[LingAo Xiao (ArmandXiao)]

Superseded by #43, which preserves this spotlight work while moving the generic PPTX-to-video runtime into the independently maintained top-level pptx2video package. Closing this PR in favor of the consolidated standalone extraction.