Video Automation Service

Machine-cut video, assembled from real media

VAS takes a script and a voiceover and returns a cut, graded 1080p film. The edit is decided per film rather than templated: every scene’s treatment is planned from what is being said, then validated against a closed vocabulary before a frame is rendered.

The videos below were produced by VAS. The system is under active development, and the cut, timing and sourcing behaviour change between builds — treat these as work in progress rather than a finished showreel.

In development Opens a viewer account — the working dashboard, without settings, administration or the service log. No sign-in.

Input: A written script or a topic, with an uploaded voiceover or a synthesised one.
Edit: Scene and shot treatment planned per film. A deterministic validator gates every plan, so an invalid edit is a refused plan and never a broken video.
Picture: Remotion. 1920×1080, 30 fps, H.264 high 4.1, yuv420p, faststart.
Sound: AAC 192 kbit/s. Mixed to −16 LUFS and −1.5 dBTP, measured on the delivered file rather than assumed.
Media: Pexels, Pixabay, Coverr, Openverse and Wikimedia Commons. Every asset carries its source and licence.
Rendering: Workers claim jobs over HTTP and hold no inbound port. Idle render machines power themselves down and are woken when work arrives.

Signed in as …

Submit·Library·Cutting·Settings·Music·Administration·Logs·

Library

Every rendered video, grouped by directory.