Splice

Overview

Splice starts from a simple provocation: this is a video-heavy studio site, but there is no stock footage anywhere in it. Every moving pixel — the hero reel, the section-transition glitch bursts, the film-grain overlay, the eight reel-card stills, the SMPTE colour bars — comes out of ffmpeg’s own procedural generators (Perlin noise, the Mandelbrot fractal renderer, a colour gradient, a standard test pattern), cut together and graded like a real edit rather than downloaded from anywhere. Nothing is filmed, photographed, or licensed, because nothing needs to be.

The signature pieces treat that footage as real signal, not decoration. A vectorscope and a waveform monitor sample the actual playing frame on every animation tick — genuine broadcast engineering, computed live via canvas rather than a looping graphic. The scrub timeline under the hero reads a real per-frame luma dataset baked at build time from the footage itself, exactly the way a real NLE caches a clip’s waveform instead of recomputing it on every drag.

The content model — a Reel post type carrying client, year, duration and studio role, tagged across six disciplines (edit, colour, VFX, motion graphics, broadcast design, live visuals) — generalises across the whole post-production trade: an edit house, a colour boutique, a VFX shop, a motion studio, a broadcast-design outfit or a live-visuals crew can all run the same eight reels with different weighting. A dev-server limitation surfaced for the first time in this catalogue too — the PHP built-in server does not honour HTTP Range requests, which breaks video seeking — so the local router now serves genuine 206 Partial Content responses for the bundled clips, documented as a dev-only shim since production web servers handle this natively.