Video link not downloading? Help and common questions

On our server in a directory outside the public web root. It is not reachable through an open URL; delivery happens only after your click with a job-bound token as an attachment.

No longer retrievable after 30 minutes at the latest; it is deleted by the next cleanup run, which happens every minute – also after an error or a cancellation. There is no permanent media library and no recovery.

The job remembers the source for processing, but logs contain no complete source URLs and no access tokens. Prepared files and preview images disappear with the job.

Where your browser saves downloads – usually the “Downloads” folder. The service cannot see or change your download folder. On a phone the file ends up in the download folder or the files app, depending on the system.

If your browser is set to “Ask where to save each file”, a dialog appears. That is a browser setting, not a MediaLink feature.

Only while the job has not expired. After 30 minutes you have to check the link again and prepare a new file.

Exactly the formats the source reports, with one recommendation. Nothing is upscaled, no “original quality” is invented and no bitrate is estimated.

Some sources report no size or resolution. The service then shows “unknown” instead of an estimated number that could be wrong.

Usually not. Separate video and audio tracks are merged with FFmpeg without re-encoding the picture. No conversions are offered.

No. Watermarks, logos and labels stay. Removing them is not a feature of the service.

A carousel shows every image and video it contains in original order as its own element. Each element is prepared and downloaded individually; there is no bulk download and no ZIP. Images are saved unchanged in the highest version the source reports – without upscaling and without additional re-encoding. Unknown dimensions are shown as unknown.

The address is incomplete, not an HTTPS link, or contains credentials or a non-standard port. Paste the complete address of a single post.

The host is not on the allowlist. Supported are links from Instagram, TikTok, YouTube, Facebook, X (Twitter), Reddit, Pinterest, Vimeo, Dailymotion and Twitch clips.

Profiles, boards, playlists, livestreams, stories and whole accounts are excluded. Allowed are single videos, shorts, reels, Twitch clips, Instagram photos and carousels, and single Pinterest image pins. For carousels each element is prepared and downloaded individually.

The source requires an account. The service deliberately uses no third-party accounts, cookies or sessions. Choose publicly accessible content.

Stories are ephemeral and out of scope for Release 1.0; their downloadability is not proven in the service. The service uses no third-party accounts or session cookies and bypasses no access control. A story link is therefore rejected with an understandable message before a job or a download is created. A later release can check this; until then there is no proven story retrieval for Instagram and Facebook.

A rate or queue limit is active. Wait a moment and try again – in the multi-link list that affects only the one line, the others keep running.

Each job allows 200 MiB and 10 minutes. Choose shorter or smaller content; the limits are not bypassed.

The 30 minutes are over and the file can no longer be retrieved. Check the link again to prepare a new file.

The service is currently not reachable. Try again in a moment.

At most 200 MiB result file, 10 minutes of content duration and 10 minutes of runtime. Intermediate files may not exceed 250 MiB.

By default 2 parallel processing jobs and a queue with 20 slots. Per anonymous session, at most 3 active jobs are intended.

By default 10 checks and 6 preparations per 10 minutes and session. These values are configurable; they exist so the service and the sources are not overloaded.

Up to 5 individual URLs per list. They are checked one after another, not in parallel. More links are counted and ignored, not processed secretly.

At most 20 elements per post and 200 MiB for the whole collection including intermediate files. A single image may be at most 32 MiB and 40,000,000 pixels; videos in carousels stay limited to 10 minutes per element. The limits apply while loading and block further elements instead of silently exceeding them.

The service should not enable bulk downloads and should not overload a source. Limits are part of the security model, not an ad or paywall.

No. Public visibility is not permission. Download only your own content or content you have explicit permission for. Copyright and platform terms apply as well.

No. The service uses no third-party accounts or cookies, no proxy rotation and no DRM removal. Protected content is rejected, not unlocked.

Brand, domain, legal notice, privacy policy, hosting and ad providers have not yet been finally reviewed legally. Technically, not every platform is equally well tested; the status shown for each platform reflects the current state.

Confirmed in person: two Dailymotion videos and two Pinterest image pins. For the Dailymotion videos, metadata, a real download through the multi-link mode in the browser, ffprobe, a full decode and a present, non-silent audio track are proven (mean volume −26.1 dB and −26.7 dB), and the operator confirmed on 2026-11-02 that both files play with picture and sound. For the Pinterest image pins the downloaded file is byte-identical to the original variant loaded again and was compared visually; the image download was confirmed in person on 2026-11-02. Partly proven: one single TikTok video (metadata, a real download, ffprobe and the picture playback by a person confirmed, about eight seconds; sound check open) and one Pinterest video pin (ffprobe and a full decode; personal playback check open). These confirmations cover only the tested links, not every post on a platform. For Instagram, YouTube, Facebook, X (Twitter), Reddit, Vimeo and Twitch, own or explicitly permitted test media are missing; there the status stays “untested”.

Metadata retrieval, a real download, an ffprobe check and playback of picture and sound by a person. Only then may the status for exactly that platform and content type move to “proven locally”. Details are in docs/test-matrix.md.