AUTO MODE: Video sites → instant central video play. Full Passkey/2FA compatibility on ANY site + auto-skip on grok.x.ai
A script for fast browsing, with a focus on downloading videos from popular video-hosting sites (YouTube, VK, OK, Dzen, Rutube)
How it works now:
Determining the site type Immediately upon loading, the script checks the URL using the regular expression "VIDEO_HOSTS_REGEX"—which covers YouTube, VK, OK, Rutube, Dzen, and Mail.ru. Depending on the result, one of two optimization modes is enabled.
Video hosting mode IntersectionObserver monitors all "video" and "iframe" elements on the page. As soon as the central video occupies more than 65% of the visible area, it receives "preload=auto" and "autoplay=1" — the browser begins to buffer that specific video, rather than wasting resources on the others. All other videos receive preload="none" and autoplay="0".
Standard Site Mode All images and frames are marked with loading="lazy" and decoding="async"—the browser loads only what is in the visible area plus a small buffer (rootMargin: 400px). During idle time (requestIdleCallback), the script prefetches up to 12 links in the visible part of the page—subsequent page transitions are noticeably faster.
Tracker blocking MutationObserver intercepts every new "script", "img", and "iframe" added to the DOM. If the src contains a domain from the list (~30 items: Google Analytics, GTM, Yandex.Metrica, Hotjar, Facebook Pixel, Mail.ru counter, etc.)—the element is immediately removed. This saves bandwidth and reduces the number of HTTP requests.
Font optimization Global CSS with "font-display: swap !important" is injected—the page does not wait for the custom font to load; the text is displayed immediately in the system font, and the font is swapped out after loading. "link rel="preconnect"" tags are also added to major video hosting sites—DNS and TCP connections are established in advance.
Animations with prefers-reduced-motion If the user has reduced motion mode (accessibility) enabled, all animations and transitions are forcibly reduced to 0.01ms—an additional resource saving on low-end devices.
