Hey Product Hunt I'm Todd. I build apps live on Twitch, and for two years I've had the same low-grade fear running in the background: one careless cat .env, one wrong window shared, and an API key is on the internet forever.
Every tool I found to "blur my screen" was a browser extension which is useless, because you don't leak secrets in a Chrome tab. You leak them in your terminal, your editor, a .env file, a database GUI. So I built Censr: a Mac app that watches whatever you're sharing and blacks out API keys, passwords, and PII before they reach your stream or screen-share. It shows up as a virtual camera in OBS, Zoom, Meet & Discord, runs 100% on-device, and has a panic hotkey ( B) for the "oh no" moments.
The one thing I won't claim: that it's perfect. No detector is. It's defense in depth a safety net for the slip-ups you can't catch yourself not a "100% safe" promise. I'd rather be honest about that than oversell it.
We launch July 1. Two questions for this crowd while I've still got time to act on the answers:
How does it handle dynamic content like password manager autofill that pops up mid-screen-share? Curious whether the detection keeps up with stuff that wasnt visible when you first set it up.
@balcicek59166 Great question and it gets at the core design. Censr doesn't do a one-time "setup scan"; it reads your screen live, a few times a second, so it's always detecting what's on screen right now, not a snapshot from when you started.
So an autofill popup mid-stream is exactly the case it's built for the moment it appears, it's read on the next scan and covered. Two things keep that from being "a beat too late": on a sudden pop-up, the feed holds the last already-scanned frame and only advances once the new content is read + redacted (no uncovered flash), and the output runs on a ~1s delay buffer so detection applies before it reaches viewers.
Two honest notes: if the manager shows the password as dots there's nothing to leak; if it shows a real credential, that's what gets covered. And it's OCR-based: which is what lets it catch dynamic content so an exotic/unreadable format is the edge case, which is why a one-key panic blackout is always there.
The latest build specifically hardened this: better small-text detection for popups on a full display, and that "hold until scanned" logic for sudden reveals.