Built to judge video by the numbers.
CutScore exists because most of what makes a video feel amateur is measurable — and most creators are left guessing. We turn the craft into objective readings against published standards, then hand back a clear score and concrete fixes.
Built by Thomas Linck, founder.
A score is only useful if it's honest. So everything that can be measured is measured deterministically — re-run the same file and the numbers don't move. AI is used only for the genuinely subjective: overall aesthetics and the written synthesis. Your footage isn't fed to a model to manufacture a number.
CutScore is pre-publish quality control for every kind of video. It is not a growth tool: no tags, keywords or thumbnails for discovery. It judges whether the video is well made, not how to get it found — a different, complementary job to tools like vidIQ or TubeBuddy.
We don't invent the targets.
Where a published standard exists, CutScore grades against it — so the targets mean something outside our tool. The core references:
EBU R128 / ITU-R BS.1770
The broadcast loudness standard behind LUFS, loudness range and true peak — the basis for the −14 LUFS / −1 dBTP targets.
WCAG 2.x contrast
The accessibility guideline for text legibility — the source of the ≥ 4.5:1 contrast bar for captions and on-screen text.
BBC subtitle guidelines
Reference for caption timing, line length and reading speed — informing our dwell-time and readability checks.
Open-source measurement is done with ffmpeg, OpenCV and PySceneDetect. Standards are cited as references; CutScore is not affiliated with these bodies.
The engine is real. The product is coming.
CutScore already produces full coaching reports today. The hosted dashboard and API are in private development — join the waitlist for early access.