A YouTube idea generator that starts from your data, not a blank box

What is a YouTube idea generator that starts from your own data?

It is a tool that proposes videos without asking you for a topic first. UploadPack’s idea generator reads four independent sources — public outliers on the competitor channels you track, your own published catalogue, the search terms already reaching your channel, and live web research — ranks candidate subjects in code, then returns filmable titles with the evidence behind each one. Your saved ideas are used only to exclude repeats, never as a source.

Key takeaways

• No topic input: it answers what to publish, not only how to publish it.
• Four sources: competitor outliers, your catalogue, audience searches and live research.
• Ranking happens in code, so results stay consistent between runs.
• Each suggestion states its evidence — and states plainly what it cannot know.

What the idea generator actually is

The idea generator is a separate tool from the Pack generator, and it deliberately takes no topic input. Where the Pack generator answers “how should I publish this idea”, the idea generator answers the question that comes before it: “what should I publish at all”. It runs against four independent sources. The first is competitor intelligence — the public outliers across the channels you track, meaning videos that outperformed that channel’s own comparable-age median rather than videos that are simply popular. The second is your own channel: the classified catalogue of what you have already published, which is used to find genuine gaps rather than near-duplicates. The third is audience search evidence — the privacy-thresholded terms that already brought viewers to your channel. The fourth is live web research, which supplies current context so a suggestion is not built on a stale assumption. Crucially, your existing idea pipeline is used only to exclude ideas you already have. It is never a source. That distinction matters more than it sounds: a generator seeded from your own backlog will keep proposing variations on what you have already considered, which feels productive and teaches you nothing.

The pain point it solves

Creators rarely describe their problem as a shortage of ideas. They describe it as a shortage of confidence about which idea deserves the next two weeks. A backlog of forty loosely formed topics is not an asset; it is a decision cost that repeats every time you sit down to plan. The second, quieter pain point is blankness — the specific paralysis of opening a planning tool that immediately demands a topic before it will help you. That request assumes the creator has already done the difficult work. The third is the discovery gap. Every channel has subjects its audience is actively looking for that it has never covered, and formats that work reliably in its niche that it has never tried. These are invisible from the inside, because the creator’s mental model of their channel is built from what they have already made. The idea generator exists to surface exactly those two categories: demand you are not serving, and formats you are not using.

How ideas are selected before the AI is involved

The shortlist is computed deterministically, in code, before any model is called. Three signals are scored separately. A competitor gap is a tracked outlier whose subject does not overlap with anything in your published catalogue, weighted by how far it outperformed its own channel’s baseline. An audience-search gap is a search term that already reached your channel but has no upload answering it directly — the strongest signal available, because it is first-party evidence of demand you are already receiving and failing to serve. A format gap is a title structure or packaging route that repeats across several tracked channels and is absent from your recent uploads. Only after that ranking is complete does the model see anything, and its job is narrow: phrase each selected seed as a filmable title and explain why it fits. It does not choose which ideas appear. This is a deliberate design decision. Ranking inside a language model would make the output vary between runs for no gain in quality, and it would make cost scale with the amount of data rather than the number of ideas.

Why each idea carries its evidence

Every idea is labelled with the signal that produced it, and that label is not decorative. “A tracked channel reached 240,000 views with this angle, 6.7× its comparable-age median” is a different kind of claim from “viewers already reached your channel searching for this, but no upload answers it”. The first tells you a format is proven somewhere else. The second tells you demand is already arriving at your door. They justify different levels of confidence and different amounts of production risk, and blending them into a single score would destroy that distinction. The generator also states plainly what it cannot know. It does not predict views. It cannot see competitor click-through rate, retention or revenue, because the public API does not expose them. An idea scoring highly on evidence is a well-supported hypothesis, not a forecast — and the tool says so rather than implying otherwise through confident presentation.

How it fits the rest of the workflow

The idea generator is the front of a workflow you already have. Each proposed idea carries a single action: research it. Choosing one hands it to Opportunity Intelligence, which runs live research, scores demand, competition, audience fit and — where the topic is genuinely about buying or choosing something — commercial value. From there the approved direction flows into a complete publishing Pack: title routes, thumbnail direction, description, chapters, Shorts briefs and the upload checklist. Nothing about the existing workflow changes. What changes is that the first step no longer requires you to have already had the idea. The workflow statuses you use — approve direction, mark filmed, mark published, defer to archive — continue to work exactly as before, and ideas you have already saved are automatically excluded from future generated sets so you are never shown the same suggestion twice.

What it needs to work well

The generator improves sharply with connected data, and it is honest about that rather than producing filler when the evidence is thin. With no tracked competitors and no YouTube connection it will tell you it does not have enough signal, instead of inventing plausible-sounding topics. Tracking three to five channels in your niche transforms the competitor gap analysis. Connecting YouTube unlocks both the catalogue comparison and the audience search terms, which is where the most actionable suggestions come from. Saving your channel memory — niche, ideal viewer, voice, boundaries — shapes how ideas are phrased and which are excluded as off-brand. None of this is required to start, but each addition measurably improves what comes back. The set refreshes once a day by default. That is a deliberate choice: it keeps suggestions stable enough to plan against, and it keeps the tool affordable to use, rather than encouraging endless refreshing that costs money and rarely produces better thinking.

What it deliberately does not do

It does not chase trends for their own sake. A topic that is briefly popular but unrelated to your channel is not an opportunity, it is a distraction, and the generator weights channel fit accordingly. It does not propose ideas you already have saved. It does not copy competitor titles — an outlier tells you a subject and structure earned attention on another channel, and the suggestion is framed as your angle on that subject rather than a rewrite of their video. It does not present public view counts as proof that a market gap exists, because outperformance relative to one channel’s own baseline is not evidence about the wider market. And it does not pretend to certainty it lacks. Where a suggestion rests on a single source, or on evidence that will decay quickly, it says so. The point of the tool is to widen the set of ideas you are choosing between and to attach real evidence to each one — not to replace the judgement that decides which of them belongs on your channel.

Official and first-party references

• YouTube Data API v3 reference
• YouTube Analytics API metrics reference

Common questions about the idea generator