YouTube
Videos, Shorts, and audio through the extraction service. Metadata works without it.
A provider appears here only when it has a typed registry entry, a matching strategy, and a real resolver path. "Extraction service" means media streams depend on a configured extraction provider — a self-hosted yt-dlp companion, the public Render API, or a public media extractor — not that every link is guaranteed.
Videos, Shorts, and audio through the extraction service. Metadata works without it.
Public videos and photo posts. Private, region-locked, or login-gated posts are not accessible.
Public Reels, posts, and carousel items. Instagram frequently requires authentication.
Public videos and Reels. Private, friends-only, and login-gated content is not supported.
Public Pins. Video Pins use the extraction service; image Pins may resolve from public metadata.
Public posts with video or GIF media. Protected and age-restricted posts are not accessible.
Public posts and Reddit-hosted video. NSFW and quarantined communities may be unavailable.
Public and unlisted videos. Password-protected videos require a browser session Fetchora does not use.
Public videos through the extraction service.
Public share links through Cobalt. Password-protected or workspace-restricted videos are not accessible.
Public OK.ru videos through Cobalt. Region restrictions and login-gated videos apply.
Public portal submissions through Cobalt. Adult-gated or supporter-only content requires a session Fetchora does not use.
Public Rutube videos through Cobalt. Some content is region-locked.
Public Streamable videos through Cobalt. Expired or private clips are not recoverable.
Public posts. A native fallback extracts video_url from the public page when available; Meta login walls can still prevent extraction.
Public Spotlight and story pages where the extraction engine supports them. Expired Snaps cannot be recovered.
Public clips and VODs. Live streams are not supported; subscriber-only VODs require authentication.
Public post video where the extraction engine supports it. The installed linkedin-scraper package is Node-only, so the Worker uses public metadata as a fallback. Many LinkedIn pages require sign-in.
Public posts on tumblr.com. A native fallback reads the post JSON for media URLs. Custom-domain Tumblr blogs are handled as generic web pages.
Public posts with media on bsky.app. Custom PDS domains are treated as generic web pages.
Detected by the /@user/status-id path on a public instance. The Worker queries the instance's public status API through masto; instance-specific blocks and auth still apply.
Public video and audio where the extraction engine supports the current VK layout. Region restrictions may apply.
Public short videos where the extraction engine supports Likee. The installed likee-api-wrapper package is Node-only, so the Worker uses public metadata as a fallback.
Public videos. The installed rumble-core package is Node-only, so the Worker uses public page metadata and direct media probes; Rumble's delivery and DRM policies can change without notice.
Clips and VODs through the public Kick API and the extraction engine. The installed kick-api package is Node-only; live streams are not supported.
Public videos where access is available from the extraction host. Some content is region-locked.
Public tracks through the extraction service. Go+ tracks and download-disabled tracks may not be available.
Publicly streamable tracks. The Worker reads the page's data-tralbum preview URLs; the installed bandcamp-fetch package is Node-only. Purchased or private downloads are not accessed.
Public shows where the extraction engine supports them. The installed mixcloud-fetch package is Node-only, so the Worker uses public metadata as a fallback. Mixcloud's streaming terms may restrict downloads.
Public photos and videos. The installed flickr-sdk package needs a Node crypto API, so the Worker uses public page metadata and Flickr-hosted image URLs. Respect the photographer's license and privacy settings.
Public images, GIFs, and videos. The installed imgur package is Node-only, so the Worker probes i.imgur.com direct media candidates and public page metadata.
Public GIFs and short videos via page metadata, Giphy-hosted media URLs, or the extraction service. The @giphy/js-fetch-api package is used when GIPHY_API_KEY is configured.
Public GIFs and short videos via page metadata, Tenor-hosted media URLs, or the extraction service. The installed tenorjs package is Node-only, so the Worker uses public metadata as a fallback.
HTTP(S) URLs that serve MP4, WebM, MOV, MP3, M4A, WAV, OGG, images, or HLS/DASH manifests directly.
Any public page with Open Graph media. Results are limited to media the page publicly advertises.
Native resolver means Fetchora can validate and stream the URL itself. Extraction service means a full format list depends on the optional yt-dlp companion and live platform access. Metadata preview means public oEmbed or Open Graph information may be visible without a downloadable stream. No label promises private, login-gated, DRM-protected, or region-blocked media.
Platform HTML and APIs change frequently. An entry in this directory describes an integration path, not permanent availability. Broken sources should be reported through the configured contact channel.