Long-form video is a Skill job: the AI turns your topic and your stories into an outline with talking points, and you record the video in your own words. Planning one video takes under an hour, and with recording it fits in an afternoon. The AI outlines; you talk. It never writes the sentences you say, because ten minutes read from a screen sounds like ten minutes of reading.
This page covers videos of roughly eight to twenty minutes: a YouTube explainer, a walkthrough, a webinar replay, a deep answer to one customer question. For videos under a minute, the short-form video page uses a different pattern.
- Shape
- Skill
- Sitting
- an afternoon
The shape: Skill
This job is a Skill because it fits in one conversation and follows one repeatable pattern. You bring material, you ask for an outline, you record from it. Nothing has to pass through another tool or wait for an approval before the next step.
The pattern is the skill: a promise, three to five sections, and talking points under each. The promise is what the viewer will be able to do or decide by the end. The sections are the steps to get there, in an order that makes sense to someone who does not do your job. The talking points are short notes, a few words each, that remind you what to say without saying it for you.
Long videos fail in two ways. Some wander, because the person on camera never decided what the video was for. Others sound stiff, because the person on camera is reading. An outline fixes the first and avoids the second. It gives you the structure of a planned talk and leaves the words to you.
That is the line the AI should not cross. It can order your ideas, suggest where a story goes and flag a section that drifts from the promise. It should not write a script for you to read aloud. Viewers came for your judgment, and they hear it in how you explain things, not in how well you read.
What you give the AI
The outline is only as good as the material behind it. "Make me a video about kitchen remodeling" returns the outline every remodeler gets. Gather these before you open the chat:
- One real question the video answers. A question a customer asked you, written the way they asked it. "How do I know if my roof needs replacing or just repair?" beats "roofing tips".
- Your rough answer. A few paragraphs, typed fast or dictated. This is the substance; the AI only organizes it.
- Two or three true stories from your work. A job that went well, one that went wrong, a customer who asked the same question. No names needed, just what happened.
- The mistakes you see. What people do before they call you that makes things harder or more expensive.
- Who the video is for. One line, such as "homeowners with a house built before 1980" or "first-time founders filing their own taxes".
- The one next step. Book a call, download a checklist, visit a specific page. Name the page or link if there is one.
- Where it will live. YouTube, your website, a course page, a webinar replay. This changes how the opening works.
- Roughly how long. Ten minutes and twenty minutes are different outlines.
- What you will not say. Prices you do not publish, claims your industry restricts, topics you avoid.
If your rough answer is thin, record a voice note explaining the topic to an imaginary customer for five minutes, transcribe it, and paste that. Most people explain better out loud than they write, and the transcript gives the AI your real reasoning.
What comes back
An outline you can read in under a minute and keep next to the camera. It has a working title, the promise, three to five sections with talking points, where each story goes, two or three moments marked as clips, the next step, and a short description for the upload page. None of it is a script.
Here is the brief to paste, with your material filled in:
Act as my long-form video planner.
Below is a real customer question, my rough answer, stories from my work and notes on who the video is for.
Use only the facts in these notes.
If something is missing, ask before you invent it.
The question the video answers:
Who should be watching:
Where the video will live:
Target length in minutes:
My rough answer:
True stories from my work:
Mistakes customers often make:
The one next step for viewers (with the page or link):
Things never to say:
Plan one video.
Give me:
1. Three working titles, each naming the question or the outcome plainly.
2. The promise: one sentence on what the viewer can do or decide by the end.
3. An opening idea for the first thirty seconds, as notes, not lines to read.
4. Three to five sections in a logical order, each with a plain heading and three to five talking points of under ten words each.
5. Which of my stories goes in which section, and why.
6. Two or three moments that would stand alone as short clips.
7. The closing step, matched to the next step above.
8. A description for the upload page of three or four sentences.
Do not write a word-for-word script.
Every section gets spoken from these notes, in my own words.
Never promise a result, a price or a timeline not given above.Read the outline once, out loud if you can. Wherever you stumble, the order is wrong or a section is doing two jobs. Ask the AI to fix that section only, rather than rerunning the whole thing.
When the outline reads clean, print it or put it on a second screen at eye level. Record section by section. A mistake costs you one section, not the whole take.
How to judge it
You can reject a weak outline in a minute or two. Look for these:
- A promise that could open anyone's video. "Everything you need to know about insurance" says nothing. A good promise names a decision the viewer will be ready to make.
- Sections that repeat. If two headings cover the same ground in different words, merge them.
- A script in disguise. Talking points that are full sentences, or whole paragraphs under a heading. Ask again for short notes.
- Stories you did not give it. An invented customer, a made-up case, a "studies show". Anything not in your notes comes out.
- Filler sections. "Why this matters" and "a quick recap" padding a ten-minute video to twenty. Cut them and let the video be shorter.
- An opening that delays. A minute of introduction before the first useful thing. The first useful thing goes first.
- A close with three asks. Subscribe, comment, book and download in one breath. Keep the one next step.
- Hype words. "Ultimate", "secret", "game-changing". Replace them with what the video actually covers.
If the outline is generic across the board, the problem is the notes, not the AI. Add another story or a sharper rough answer and run it again.
What stays with you
The AI decides the order of ideas. Everything the viewer actually hears is yours:
- The words. Explain each section the way you would to a customer sitting across from you. If a talking point feels wrong when you say it, skip it.
- The judgment. The reason someone watches twelve minutes of you is your experience. The outline holds it; it does not supply it.
- Every claim. Check any figure, rule or deadline against its source before you say it. Health, legal, financial and real estate work have rules about what you may claim, and they apply on video.
- Permissions. A customer named, shown or recognizable agrees first. Music and footage you did not make need the right to use them.
- The recording. Light from the front, the camera at eye level, sound checked before the first take. A clear voice matters more than a sharp picture in a long video.
- The upload. Title, description, thumbnail and chapters are checked by you before you publish. Our guide to YouTube thumbnail size covers the image.
- The comments. Questions under a long video are often the best ideas for the next one.
A worked example
This is an editorial example with an invented business. Say you run a small bookkeeping practice for contractors, and clients keep asking the same thing every spring.
The question in their words is "Do I need to send a 1099 to the guy who helped me on that job?" The rough answer is three paragraphs typed between calls. The stories are three: a client who paid a helper in cash and could not find the records, one who sent forms to everyone just in case, and one who set up a simple tracking sheet in January and had nothing to do in the spring. The next step is booking a year-end review on the practice's booking page.
The AI returns a promise: by the end, a contractor can tell which helpers need a form and what to collect now. It proposes four sections: who counts, what to collect before paying anyone, what happens when records are missing, and a simple tracking routine. The cash story lands in the third section, the tracking-sheet story in the fourth. It marks two clip moments: the cash story and the one question to ask before paying a helper.
One talking point under the first section is a full paragraph, so it goes back for short notes. One section titled "why this matters" repeats the opening, so it is cut. The owner checks the filing rules against the current official guidance, then records the four sections in one afternoon from the printed outline. The video makes no claim about how many clients it brought in, because there is nothing to prove that with.
Talking head or screen recording
Both routes use the same outline. The choice is what the viewer needs to see.
| Route | Who does the work | Best when |
|---|---|---|
| Talking head | You plan with the AI, then speak to the camera | Trust in you is why people buy: advice, coaching, local services |
| Screen recording with voice | You plan with the AI, then talk while showing your screen | The topic is visual: software, a spreadsheet, a form, a website |
| A mix | Camera for the opening and close, screen for the middle | You explain a process and also want viewers to know your face |
A long video also makes good raw material for the rest of your week. The content repurposing page covers turning one recording into many smaller pieces, and the full list of jobs like this one is on the DIY marketing hub.
Of every job on this list, social media is the one that comes back every week, and Boomp is software that does that one for you. See ten posts made from your website, free
