4 items
Rewrites clickbait headlines with on-device AI and swaps clickbait thumbnails for real video frames. No accounts, no tracking. Clickbait, neutralized — with the AI running on your own machine. Clickbait Neutralizer rewrites manipulative, curiosity-gap headlines into plain, factual ones as you browse, and can swap clickbait video thumbnails for a real frame from the video. A small AI model runs inside your browser, on your own hardware — the headlines you read are never sent to any server. What it does • Rewrites clickbait headlines into calmer, factual versions, in place. • Optionally replaces exaggerated video thumbnails with a real frame from the same video. • De-hypes and de-clutters titles across the sites you visit. • Falls back to instant rule-based cleanup when the AI can't help or can't run. Honest about where it's at This is an early release (v0.9.1). It runs a very small language model on your own device, so the rewrites are noticeably rougher than a big cloud model's — expect some awkward or off rewrites, and a second or two of delay. It's free, and you can even edit the AI's instructions yourself from the options page (with one-click reset) to tune it. If your hardware can't run the model, headline cleanup still works through the built-in rules. Private by design • No accounts, no sign-in, no API keys. • No analytics, no telemetry, no tracking, and no backend server of ours. • The headlines you read and the pages you visit are processed on your device and are never transmitted to us or anyone else. • The one exception: the optional thumbnail feature. To show a real video frame, your browser loads that image from YouTube's own public image servers — the same place it already gets YouTube thumbnails. No headline text or browsing history is involved. Full details are in the privacy policy. Requirements Works best on a machine with a GPU that supports WebGPU (most laptops from the last few years). If the on-device model can't run, headline cleanup still works via the built-in rules. Free, private, and honest about its limits.
rating_count is the Chrome Web Store ratings count, not a written-review count.