Somewhere on your machine there is a document. It runs to about nine thousand words, it has a section headed “OK so this is where it gets complicated,” and it contains a genuinely correct explanation of something the internet has been getting wrong for years. Could be why a 1997 console launch went the way it did, or the rules interaction three quarters of a playerbase misplays, or the production history that makes the last four episodes of a cancelled show land completely differently.
It has been sitting there since roughly last October. You keep meaning to make the video.
The usual explanation is that editing is hard, and editing is hard. I do not think that is what stops most people, though, because the same person who cannot ship a twelve-minute explainer will cheerfully spend four hours colour grading a thirty-second bit for a Discord server. Something earlier in the chain is jammed.
The Gap Nobody Warns You About
Research and video pull in opposite directions, and the handoff between them is where these projects go to die.
Doing research makes things bigger. Every source points at two more, the document grows because growing is what documents do. And there is always one more 2009 forum post that clarifies a detail nobody else caught. It feels productive, which it is, right up to the point where you have assembled something that cannot be turned into anything.
Making a video means throwing most of it away. Eleven of your fourteen sections have to go. The tangent you are proudest of turns out to be a tangent. Everything has to run in an order somebody with zero background can follow, which is rarely the order you learned it in. That is a different skill, it is much less fun, and it is exactly where the project stops feeling like a hobby and starts feeling like homework.
So the doc sits. Not because you cannot edit, but because the structural pass never happened, and the structural pass is invisible unpaid work that nobody claps for.
There is a second thing stacked on top of that for a lot of people, and it is the camera. Some of the sharpest researchers in any fandom are quietly awful on video, know it, and would rather not spend a Saturday confirming it. Knowing things and performing on camera are unrelated skills. The format has just historically insisted on both.
What Changes If the Structure Comes First
Treat the document as the input to the video rather than the quarry you eventually chip a video out of by hand.
That is what tools built to create educational videos with AI are doing. With Leadde you paste the text in, or upload the doc, the PDF, the deck, whatever shape the research ended up in. It comes back with an outline, an arrangement of what appears on screen for each beat, and generated narration. The structural pass you have been dodging since October takes about a minute.
Then you argue with it, which is the useful bit. The first draft will not be your video. It will be a structure, and reacting to a structure you disagree with is enormously easier than conjuring one out of nine thousand words. Half the time the reaction is “no, the timeline has to come before the mechanics or none of this lands.” That reaction is the editorial call you could not make while staring at the doc.
Worth touching a few settings. The audience field changes the output more than you would guess, and being honest in it matters. Something aimed at people already deep in a fandom can skip setup that newcomers need, and leaving the field blank produces something pitched at nobody. The narrative style options include Analytical and Explanatory, which are the right registers here. The persuasive ones read like an ad.
Length is the setting people get wrong. Six minutes is the number that gets repeated constantly in this space, originally out of engagement data on online course videos, and your audience is not a MOOC cohort, but the underlying point survives translation. Summary lands around one to three minutes and Balanced around three to five. If your gut says the topic needs twenty-two minutes, that is probably the same gut that kept the document growing.
For the camera problem there are over 200 built-in presenters, and you can generate one from a single photo. Someone who does not want to be filmed can still put a consistent face across a series, which matters far more for a series than for a one-off.
The Visuals Problem, Which Is Its Own Nightmare
Explainer content has an asset headache that most video advice ignores. Stock footage is useless for nearly everything nerds explain. Nobody has shot royalty-free b-roll of a cartridge lockout chip or an abstract concept from a novel.
The media panel covers part of this by generating assets from a prompt. You can produce images or short clips, pick which model handles it, and drop the result straight onto the canvas. That fills the awkward space between needing something on screen and not being willing to commission original art for a twelve-minute video about a discontinued handheld. There are motion assets as well, for the animated text moments that carry a surprising amount of this genre.
This does not replace a real artist on a project that deserves one. It does mean the absence of a visual stops being a reason to shelve the whole thing.
Measure the Right Thing
Views are a poor signal for educational content and most creators know it. A video that pulls eleven thousand clicks and loses most of them at ninety seconds did not explain anything. It got clicked.
The analytics on a shared video include completion rate and average watch time, which are the numbers that tell you whether the explanation worked. If people consistently bail at the point where the second mechanic arrives, the problem is how you introduced the second mechanic. That is fixable, and it is a different class of feedback from a comment section relitigating something you said in the opening.
Shared videos also carry a chat panel where a viewer can ask about what they just watched. For a rules explainer that is genuinely handy, because the questions people ask about rules are wildly predictable. And asking one directly beats scrolling for somebody who was confused in the same way.
If your topic has an international following, and most do, a finished video can be translated with the script and the on-screen text handled together, across 88 languages and 175 dialects. Fan translations of good explainers happen regardless. This version is at least accurate.
Series consistency is the other thing that quietly gets easier. Anyone who has come back to part four of something after a six-week gap knows the problem, which is that part four sounds like a different person made it. Because a different version of you did. Working from the same source settings and the same presenter across an entire run keeps the voice steady even when your enthusiasm for the topic is not. That is worth more on a long-running series than any individual episode improvement.
The Parts It Cannot Do
The output only knows what you handed it. A document that sounds certain about something you are privately unsure of produces narration that sounds certain in a clean. Even voice, which is worse than you being certain in your own voice where people can at least hear the hesitation. Read the script properly, and on contested topics put the uncertainty back in deliberately.
It also will not do the thing that makes explainer content worth watching, which is the connection nobody else made. Structure it can handle. Insight it cannot. A well-organised summary of material already sitting on a wiki does not become good because the production improved.
Look at a generated presenter for more than about ten seconds and it starts to wobble. Anything with your personality in it, a rant or a channel update, should be you. This handles the explanatory half of a channel and not the personality half, and confusing the two is how a channel loses whatever people subscribed for.
None of it helps with the rights on footage or images you do not own, which for this genre is frequently the actual constraint. Generated assets route around some of that. They do not route around the clip from the game.
Desktop and Chrome only, realistically, so do not plan on building anything from a laptop in a convention hall.
Just Do the One That Has Been Sitting There
Take the document you have been avoiding. Not a fresh idea. The old one, with the section heading that admits it gets complicated.
Run it through the free tier and set the length short enough that it feels wrong. Read the structure it hands back, and watch for the first moment you think “no, that ordering is backwards,”. Because that instinct is your video. Rebuild around it, cut the two sections you already privately know are tangents, and publish.
Then look at where people stopped. That number will teach you more about your explanation than another eight months of reading would, and reading was never the part that was missing.
