Using Tubekeep

Sort the queue by length, by how long it has sat there, or by channel rather than the order it went in. Most of the backlog is not one problem: a hundred five-minute videos and twenty hour-long ones are different afternoons.

When the answer is a whole pile rather than a handful, narrow the grid first. The filter row asks seven questions: watched, in progress, under twenty minutes, over it, taken away by YouTube, also in another of your lists, in none of them — each button saying how many it would leave.

Seeing the pile is the point: unticking a filter costs nothing, and nothing has happened to your account yet. Then select by clicking, by shift-clicking a range, by dragging a box across the grid, or by taking everything left in one press.

The row also takes a word. A box at its head matches titles and channel names — the same rule the library-wide search uses — and combines with whichever button is pressed, so “watched, and about Rust” is one question with a grid for an answer.

The same row sits above every playlist, where it answers different questions: what is in here twice, and what has been sitting in it longest.

Then remove the selection, or move it into a playlist — one that exists, or one you name in the picker. Either way it becomes a batch, and the batch is paced: the count moves while it works, and it tells you what it did when it stops.

YouTube hides unavailable videos from a list and tells you only how many: a playlist of 400 plays 380, and the page says some are hidden without saying which. Its own menu has the way back in — «Show unavailable videos» — and that is what Tubekeep asks for, list by list.

The Unavailable pane is what comes back, grouped by which of your lists each one is sitting in, saying which are deleted, which have only gone private, and which still exist but will not play for you. A private or blocked video can come back; a deleted one cannot.

What each row can tell you depends on what YouTube kept. A blocked or private video usually keeps its title, channel and length — enough to go looking for another copy, which the pane searches for in one press. A deleted one comes back as a placeholder with no name at all, and the pane says so.

The pane is only as fresh as the last read, which is why it says when that was and offers to scan again. Clearing them is a normal batch with one difference: it cannot be undone, because nothing can add a deleted video back to a playlist. The button says so before you press it.

A video in four lists is four rows here, because removing it is four separate writes to four separate lists. Select across as many as you like and move or copy the lot; a move always adds to the destination before removing from the source, so an interruption leaves a duplicate rather than a gap.

Your own lists and the ones you saved from other people are drawn as two groups, and the difference is not cosmetic: YouTube gives out no per-item handle for a list that is not yours, so saved lists are read-only here. You can search, sort and copy out of one, and remove it from your library — you cannot edit what is inside.

Liked videos has its own pane for the same reason Watch Later does: it is one of YouTube's own lists, not one of yours. You can copy out of it and un-like in bulk. The list itself is account furniture — it cannot be renamed, hidden or deleted, by Tubekeep or by you.

Folders are yours: name them, nest them, then select a batch of channels and file the lot into one with the "File into…" picker. Filing happens immediately and asks nothing, because it changes nothing outside this browser — deleting a folder unfiles its channels and never unsubscribes from anything.

A channel can only be in one folder, which is what keeps the tree readable. If you want a channel in two places, the search is the better tool: it answers across the whole index regardless of where things are filed.

Batches run at the pace YouTube accepts, a pause between requests. If the site starts refusing, the run stops and says so rather than hammering it — that is the normal end of a round, not an error. It starts itself again a minute later, and there is a button to skip the wait; nothing unsent is lost.

Nothing is left half-done: each item either happened or did not, and the library only records the ones YouTube confirmed. So a batch that stopped halfway leaves your account and this index agreeing with each other.

Undo runs the batch backwards as another paced batch, and reaches back thirty days — a finished batch stays reversible that long and is then forgotten on its own, whole. Cancelling is different: cancel drops what has not been sent yet and leaves what has, which is what makes undo possible afterwards.

The free allowance is 500 batch items for the life of the install. Items, not writes: a move is two writes and one item, and undoing a batch is not a second batch. A selection of one is never counted — what YouTube lets you do one at a time is free here without limit.

Every confirm says how many are left, and a batch bigger than what is left is refused whole with your selection standing — so you can narrow it from the filter row's own counts, or lift the ceiling with Pro.

The count is kept in your browser beside the library and travels inside your backup file, so moving to a new machine keeps it.

Filing subscriptions into folders is the one thing here that calls a language model, so it is the one thing that needs you signed in to Deft Foundry. Everything else works signed out.

It files as it goes, and marks what it made: undoing an AI run removes only what the run filed and the folders it created, leaving every folder you built by hand — and every channel you filed yourself — exactly where they were.

A free account has a lifetime allowance of 200. It is cumulative for the life of the account and never refreshes — nothing expires and nothing is taken back. What it counts is everything a run asks about: the channels of the first pass plus the topics of the second, so grouping forty topics under parents is forty off the same 200.

A subscription list comfortably under 200 gets both passes free. One of several hundred sees a third of itself filed before the wall and stays flat, which is enough to see the tree take shape and decide about the rest. Pro lifts the allowance, with a fair-use ceiling of 20,000 a month.

The count is kept on our server rather than in your browser, because a count kept in your browser is cleared along with everything else in it — and because this is the one feature that spends money on our side.

What crosses the wire is names — channel names, @handles, a short channel description, a few titles of videos you saved from each channel, and the folder names you already have — never an id, so the answer comes back by position and the mapping stays in your browser.

Next to the button that starts a run is the language the new names are written in: English, 中文, Español, Português, 日本語 or Deutsch. It is not the interface language — it is what your folders will be called, so it follows what you watch.

Changing it mid-run takes effect on the next batch, deliberately: half a run in one language and half in another builds two names for one topic.

The whole index exports as one JSON file and imports on another machine or after a reinstall. Thumbnails are not in it — they are re-fetchable, and including them would multiply the file size for bytes Google already has. Neither are queued or finished writes: the file is the library, not the log.

An import is a merge, and the copy in front of you wins. It adds what this browser does not have and fills what it left empty — a folder you renamed since stays renamed, and nothing is ever removed. To get back exactly what is in a file, clear the local library first, then import.

Most of the file is a convenience: videos and playlists come back from a harvest more current than any restore. The folder tree over your subscriptions is the part that does not, because YouTube has no folders — that tree exists in this browser and in this file, nowhere else.

Importing changes nothing on YouTube. It writes the local copy only — no playlist is recreated, no video re-added, nothing re-subscribed. A file holding a list you have since deleted on YouTube brings back the local rows, not the list.

Changes you make in this browser show up straight away. Changes made on your phone or another computer are mostly picked up on their own: opening YouTube triggers a light re-check, at most every half hour, that compares each list’s stated count and re-reads only the lists that moved.

A pair of edits that leaves a count unchanged slips past that comparison, which is what Refresh is for — the exact pass, walking each list again from the top, because YouTube appends to the end of a list and anything new is on the last page.

Refresh and the light re-check both cover your playlists and Watch Later. Subscriptions are the one thing both leave alone: re-reading them costs more than half of everything a pass downloads, and they state no count that could be compared cheaply. If you subscribed or unsubscribed somewhere else, use Full re-walk on the Data pane — it reads everything, subscriptions included.

There is no cloud copy and no sync. That is the same decision as everything else here: nothing of yours is on our servers, so nothing of yours is ours to lose.