Embed · Caption Mangle Scanner

A tiny embeddable widget for L&D bloggers

If you write about accessible training video, caption vendor selection, or WCAG remediation, you can drop the GlossCap Caption Mangle Scanner on your post. It's two lines of HTML, ~6 KB, and it shows your readers interactively which of the jargon in their own training library the big auto-captioners mangle. Every embed is also a link back to GlossCap.

Live preview

Here's the widget running on a light-background host page, as it would appear on a typical blog post. Click a pill or type a term.

Try this: scan your own onboarding vocabulary

The chart below is live. Click on a term to see what YouTube's auto-caption model would actually produce, or paste a word from your own company glossary.

Two-line install

Paste this on any page where you want the widget to appear. It renders where you place the target div:

<div id="agent-11-embed"></div>
<script async src="https://glosscap.com/embed.js"></script>

The widget inlines its own styles and has no dependencies. Safe next to any CSS framework. If the target div is missing, the script is a no-op.

What readers actually do with it

The scanner auto-loads with Kubernetes → "cooper Netty's" as its first example so the first frame is already useful. From there, readers can:

Works on light and dark host pages

The widget is self-contained and themed for a light-neutral background, which is where most embeds land (blog posts, comparison articles, L&D Notion pages exported as HTML). It also renders cleanly on a dark host — the widget keeps its own white card so the interior is always readable.

Light host

(only one scanner per page — scroll up to try the live one.)

Dark host

The widget keeps its own light card inside a dark wrapper so copy stays legible regardless of the host theme. Preview the rendering by embedding on a dark page; the inlined CSS is below.

Spec

File/embed.js
Size~6 KB gzipped. No external stylesheet, no font load, no tracking pixel.
DependenciesNone. Vanilla ES5. Works in every browser GlossCap's main site works in.
Accessibilityrole="region", aria-live="polite" on the result panel, form input has an aria-label. Keyboard-navigable pills. No colour-only meaning — the mangle row is also struck through.
Mangle dictionary30 hand-curated real auto-caption errors: Kubernetes, kubectl, Terraform, Prometheus, Grafana, PyTorch, LangChain, SAML (+SAML assertion), OAuth, Okta, Docebo, TalentLMS, Kaltura, Absorb LMS, tirzepatide, semaglutide, cholecystectomy, anastomosis, bronchiolitis, SOX, FINRA, HIPAA, GDPR, Postgres, PostgreSQL, Redis, CI/CD, SDK, API.
AttributionEvery render includes a "Powered by GlossCap" link. This is how embeds pay for themselves — the widget is the free version, the backlink is the price.
Unknown termsFor anything not in the dictionary, the widget returns a graceful "GlossCap reads this from your Notion / Confluence / Google Docs glossary" response, still demonstrating the pitch.
PrivacyAll lookups run client-side. The widget does not make network calls, does not set cookies, and does not send any pasted term back to GlossCap.
LicenseFree to embed on any site. Attribution (the bottom line) must remain visible. If you want an un-branded version for an enterprise intranet, email hello@glosscap.com.

Good places to embed it

Related