Video Tool · Caption Tool · Operate Surface · 2026 · Solo — caption segmentation algorithm, word-break interaction, wpm analysis, timing strip renderer, WebVTT export, Operate surface design, Vercel production deployment
Breathline
Browser-local caption rhythm bench: set cue breaks on a word rail, inspect per-cue timing and wpm, and export a WebVTT file — all computed locally from clip duration and break positions.
The problem
Video editors turning spoken audio into subtitles need to break one continuous passage into readable cues: short enough to read before the next cut, spaced to match the voice's natural rhythm. Subtitle software is heavyweight; manual timing is slow. Breathline solves this for the single-cue case: paste the transcript, tap before any word to set a break, and get instant wpm and duration feedback per cue — plus a downloadable WebVTT file ready for video players.
Architecture
Key decisions
Word-rail interaction over text selection
Tapping a word on the rail to toggle a cue break is faster and more reliable than text-range selection, especially on touch devices where selection handling is inconsistent.
Proportional timing by word count
Distributing cue time proportionally by word count (rather than character count or equal splits) produces timing that aligns with natural speech rhythm while remaining instantly computable without audio analysis.
Source availability caveat
This case study was added via transparent Vercel-only recovery: the product is live and functional at foundry-2026-09-05-breathline.vercel.app, but the source repository was not recoverable during reconciliation. Architecture is inferred from live behavior. Implementation notes available on request.