Replace the slower seven-shot presentation with a separate 14-shot Remotion cut at the same 23.816667-second picture duration. Open with “Why is your web terminal still slow?”, introduce Wraith through an edge-to-face reveal, then use opposing browser angles, diagonal reveals, a screen dive and faster product cuts. Preserve the earlier composition/export and supplied BIA soundtrack.
Add matched xterm comparisons backed by 36 fresh serial runs across dense, dashboard and scroll/full-repaint workloads. The compatible path supports conservative throughput bounds of 4.4×, 1.3× and 2.4× against xterm WebGL. Dense DOM p50 time bounds are 22.5 ms versus 1.5 ms for Wraith, supporting the 15× CPU-time shot. Ratios round down slowest Wraith / fastest xterm; GPU completion, main-thread parse/draw/layout time and display FPS are distinguished. Retain raw runs, runtime/runner hashes and a verified hardware Metal adapter probe. The separate native 4,500+ shot uses its own workload and is not a comparison numerator.
Actual xterm/Wraith capture uses identical bytes, grid, font and edited playback cadence; input and visible-text hashes match for all eight warmed fixture states. No artificial xterm slowdown. Source music, captures and exports remain local.
Validation: strict TypeScript and formatting; 71 key frames, all fourteen motion strips and encoded contact/proof/reveal images reviewed. All 1,429 frames decode at 1920×1080/60 fps; no browser errors, unintended silence or black gaps detected. Stereo 48 kHz AAC measures -14.04 LUFS / -4.92 dBTP. Picture duration stays 23.816667 seconds; AAC/container duration is 23.829 seconds. Music cue residuals are at most 8.923 ms; listening review is available in Studio and is not claimed by the measured checks. Export and source hashes are retained in video/claims/hype-flex-review.json.
Local output: video/wraith-hype-flex.mp4. Rebuild and preview: video/remotion/FLEX.md, composition WraithFlex. No terminal runtime changes.
Closes #24. Stacked on #23. No merge, release or public video publishing.