What this tool cannot do

Deliberate refusals

Some of these limits are technical and some are choices. The choices are listed first, because a tool that could do something and declines to is making a statement, not reporting a bug.

This tool does not download videos or audio, does not decode signature ciphers, does not call YouTube’s internal application endpoints, does not scrape the watch page, and does not sign in as anybody. Each of those would be a step toward circumventing an access control, and none of them is necessary to look at a public preview image.

Captions are the clearest case. Thumbnails are static files on a host that allows cross-origin reads; captions are not, and the difference is deliberate. The timedtext endpoints return zero bytes without a proof-of-origin token from a real player session, and the watch page that carries the signed caption URL sends no cross-origin header, so a web page cannot read it either. Both walls could be climbed from a server by forging a player session. That is where this tool stops, which is why the Captions panel asks you for the file instead.

Limits imposed by what YouTube publishes

Only eight thumbnail files exist per video, and the largest is 1280x720. There is no hidden 4K thumbnail to unlock.

The oEmbed record carries the title, the channel, the channel URL, the default thumbnail and the player dimensions. Everything else people expect — views, likes, description, tags, duration, publish date, captions — is simply not in it.

The WebP versions of the poster images exist, but the path that serves them sends no cross-origin header, so a page can link to them and cannot read them. That is a decision on Google’s side, and no amount of client-side cleverness changes it.

Private, deleted and age-restricted videos publish neither thumbnails nor an oEmbed record to a signed-out visitor.

Limits that come from being a page in your browser

The requests are made by your browser, so anything that blocks Google hosts — a content blocker, a corporate proxy, an offline connection — stops this tool exactly as it would stop any other page.

Nothing is stored between visits. There is no history, no saved list and no account, so a fetched result disappears when you close the tab.

Bulk work is not supported: one video at a time, on purpose. A batch runner pointed at Google’s static hosts is a rate-limiting problem waiting to happen.

A limit with a shelf life

The thumbnail address pattern is a long-standing convention rather than a documented, versioned API. It has been stable for many years, but Google is under no obligation to keep it that way.

If it changes, this tool will report the files as unavailable. It will not guess at replacement addresses or fall back to scraping.

Questions

Limitations

• WebP thumbnails can be linked but not downloaded here: Google serves the /vi_webp/ path without a cross-origin header, so no web page can read those bytes.
• No video or audio downloads of any kind.
• Captions cannot be fetched from YouTube automatically: the endpoints require a player token and the watch page blocks cross-origin reads. The panel reads what you paste or choose — SRT, WebVTT, SBV, TTML, timedtext XML or JSON3, or transcript-panel text.
• Timings taken from the transcript panel are per-line start times, so end times are implied from the next line and are accurate to the second, not the millisecond.
• No view count, like count, description, tags, duration or publish date: oEmbed does not publish them and this tool does not scrape.
• Nothing for private, deleted or age-restricted videos.
• No batch processing and no stored history.
• The thumbnail URL pattern is an undocumented convention that could change without notice.
• Requests are made by your browser, so a blocker or proxy that refuses Google hosts will stop it.