How to extract performer names and titles from adult video links automatically

What the metadata is and where it lives

• Title: the HTML title tag, og:title, twitter:title, the JSON-LD VideoObject name, and the page's h1. They rarely agree: the title tag usually carries the site name as a suffix and sometimes a quality label.
• Thumbnail: og:image, twitter:image, the JSON-LD thumbnailUrl, the itemprop thumbnailUrl meta, or the poster attribute of the video element.
• Duration: the JSON-LD duration (an ISO 8601 value such as PT12M30S), og:video:duration in seconds, or an itemprop duration meta.
• Performers: a block of performer links on the page (each site names it differently), the video:actor or og:video:actor meta tags, the JSON-LD actor list, and on some sites only the title. This is the field with the most noise: uploader names, recommended-performer widgets and category words all look like names.
• Categories and tags: the keywords meta, video:tag meta tags, the tag strip under the player, and JSON-LD keywords.

Doing it by hand

Open the video page, view the source and search for og:title, ld+json and the word actor. That gives you a title, a thumbnail URL, often a duration, and with luck a cast list, for one video. It works for a handful and falls apart at a hundred, and the raw values need cleaning anyway: titles end in the site name and in labels like HD or 1080p, performer names come back with follower counts glued on, category words show up in the cast, and an age-gate or a bot check returns a page whose title is Just a moment or Age verification, which is not the video's title.

How PornLibrary extracts it

You paste a link and the server fetches the page with browser-like headers, retrying up to three times with a timeout that starts at ten seconds. It then reads, in order: the JSON-LD VideoObject, the Open Graph, Twitter and itemprop meta tags, the title tag and h1, and finally a site-specific extractor for each of the 29 supported sites. The result is one entry with title, preview, duration, performers and tags, which you can edit before saving. The automatic extraction page has the short version; this is the long one.

Titles are cleaned: the site name and quality suffixes are stripped, placeholder titles (Video title, Just a moment, Access denied, page not found) are rejected, and if nothing usable survives the title is inferred from the URL slug, so an age-wall never becomes the saved title. Performer names are split from lists, normalized for look-alike characters and whitespace, stripped of follower-count suffixes, and run through a taxonomy that separates real names from category labels, bare numbers and widget captions. On sites where the extractor is scoped to the video's own cast block (EPORNER, HQPorner and Pornhub among them) that block replaces the noisier union of meta tags, which is how a page with dozens of names in its sidebar saves with the people actually in the video. The same name cleanup runs again when a video is saved, so names typed by hand or sent by the extension get the same treatment.

Duration comes from the JSON-LD first, then og:video:duration, then the itemprop meta. The preview is loaded through PornLibrary's own image proxy from the source site, which is what makes it display reliably; the video file is never fetched.

When a site blocks, or a link dies

Some sites answer 403 to any server, whatever the headers: SpankBang fingerprints the TLS connection itself, and PORNTREX blocks server requests as well. PornLibrary fails fast with a clear error and the Chrome extension is the route: it runs in your own tab, where the site sees a normal visitor, and reads the title and preview from the page directly. The extension adds a save button on 28 sites and is part of lifetime access. If a fetched page turns out to be an age-gate or a 404, the title guard rejects it and the entry falls back to a title inferred from the URL, which you can correct before saving. After saving, with lifetime access, a weekly dead-link check flags entries whose link answers 404, 410 or 451 on two consecutive runs; the metadata stays, so you still know which video it was.

Limits worth knowing

• Site-specific parsing exists for 29 sites; the full list, with what each one yields, is on the supported sites page. On any other URL the generic reading of Open Graph and JSON-LD still runs, but results vary, and the link saves regardless so you can fill the fields by hand.
• Performer detection depends on the site listing the cast. Community tubes without per-video performer links leave the cast to whatever the title says, which may be nobody.
• A site that changes its markup can stop yielding a field until the extractor is updated. The title and preview usually survive through Open Graph; the cast is the fragile part.
• It extracts metadata about the video; it does not download, transcode or store the video itself.