Building a topic bank so your faceless channel never repeats itself
Channels rarely die from one bad video. They die slowly, from running out of things to say and reaching, without quite noticing, for a topic already covered three months back under a slightly different title. A topic bank is the fix: a standing reserve of validated ideas kept well ahead of upload day, checked against everything the channel has already published rather than against what you happen to remember.
Why running dry happens gradually
The early videos on any channel are easy, because the obvious ideas in a niche have not been touched yet. The difficulty rises with every upload, not because the niche runs out of material, but because finding the next idea that is not a rewrite of an earlier one takes real searching. Under deadline pressure that search gets cut short, and a topic that feels new because the title is worded differently slips through. The channel has not actually run out of ideas; it has run out of ideas it can find quickly under pressure, which is a production problem with a production fix.
Generate more than you need, on purpose
The standing fix is a reserve, not a scramble: keep a month or two of upload slots filled with vetted ideas at all times, refilled in batches rather than the night before a deadline. Generating a batch means producing more candidates than you need and keeping the strongest half, since over-generating costs almost nothing and a filtered batch is a chosen list instead of whatever arrived first. This is the same principle behind batch-producing a month in a weekend, applied one stage earlier, at idea generation rather than video production.
Wording hides duplicates from a human eye
A spreadsheet of past titles feels like protection against repeats and is not, because the same idea has many phrasings. "How ancient Rome moved water" and "the engineering behind Roman aqueducts" are the same video wearing two titles, and catching that by scanning a growing list means holding every past topic in your head while writing every new one, which gets harder exactly as the archive that matters most gets larger. This is the point where a channel needs a check that compares meaning, not just words, since a wording-only check passes the exact duplicates that actually happen.
What a real duplicate check catches, and what it misses
A check that compares the meaning of a new idea against everything a channel has covered catches reworded repeats reliably: different words, same underlying topic, flagged before it gets scripted. It does not catch everything. Two ideas that are conceptually adjacent while sharing almost no vocabulary, an aqueduct video and a separate video on ancient plumbing more broadly, can still both look novel to an automated check the same way they would to a human skimming quickly. Treat automated duplicate detection as a strong first pass that removes the obvious repeats, not a substitute for knowing your own back catalog on the ideas that matter most to get right.
A reserve compounds with everything else
A standing topic bank is what makes the rest of the production system work without strain. A cadence is only sustainable if there is always a validated idea ready when the slot opens, running several channels multiplies the need for a reserve per channel rather than one shared list, and a channel with real depth in its topic bank is the one built on the niches deep enough to sustain a hundred scripts in the first place. The reserve is not a nice-to-have; it is what keeps every other system from running on fumes.
Where Thothium fits
Thothium keeps a per-channel record of every topic it has generated, and when expanding an umbrella idea into a batch, it over-generates candidates and filters new ones against that full history by meaning, not just recent memory, so the same reworded repeat does not resurface months later. It is a strong first pass exactly as described above, not a substitute for your own judgment on the close calls. It is in free alpha, and the form below gets you a key.
Frequently asked questions
How many video ideas should a channel keep in reserve?
A running buffer of a month or two of upload slots is a comfortable floor, generated well ahead of when you will need it. The point is not to plan a year in advance, since the best ideas often come from what is working right now, but to never be staring at a blank slot the night before a scheduled upload.
Why do reworded duplicate topics happen even with a spreadsheet?
Because a topic list is just words, and the same idea has many phrasings: "how ancient Rome moved water" and "Roman aqueduct engineering" read as different rows and are the same video. Catching that by eye means remembering every past title while writing every new one, which fails quietly as the archive grows.
Does automatic duplicate detection catch everything?
No, and it is worth knowing where it stops. Matching by meaning catches reworded repeats reliably, but two ideas that are conceptually similar while sharing almost no wording can still slip through the same way they would slip past a human skimming a list. Treat automatic checks as a strong first pass, not a replacement for knowing your own back catalog.
Should every generated idea in a batch get used?
No, generate more than you need and keep the strongest. Over-generating a batch and discarding the weak half costs nothing but a little compute, and it means the ideas that make the cut were chosen rather than simply the first ones that arrived.
Last updated July 18, 2026. Automated similarity checks are a production aid, not a guarantee; review a candidate topic yourself before committing real research time to it.