All Things AI · Season 3 · What You Can Build · Episode 7
You can produce thirty videos in one session using your own face and your own voice, without filming thirty times. Publishing consistently has always been a production problem rather than an ideas problem, and this is the thing that removes the production problem.
Consistency is a logistics problem, not a discipline problem
Every agent knows video works. Almost nobody publishes consistently, and the reason is not laziness or a shortage of things to say.
The reason is that filming has a fixed overhead that has nothing to do with the content. You have to look presentable, the room has to be tidy, the light has to be adequate, and you have to be in the right frame of mind to perform. That overhead is the same whether you are filming one video or eight, which means the sensible thing to do is film eight, which means you need eight scripts ready, which means you need to have written them, and the whole thing collapses back into a task you postpone.
So the week goes by. Then the month. And the agent who publishes twice a year concludes video did not work for them.
What is actually available now
Three separate capabilities that combine into one workflow, and it is the combination that matters rather than any one of them.
A presenter that looks like you
A recorded likeness of you that can be given new words to say. You film once, properly, in good conditions, and that recording becomes the presenter for everything afterwards. The quality of that single session determines the quality of everything it produces, which makes it worth doing carefully.
A voice that is yours
Your own voice, recorded once, used to speak new scripts. This is the part that most determines whether the result feels like you or feels like a product demonstration, and it is worth more attention than the picture.
Scripts that already sound like you
This is where the episode on training a model on your own writing becomes load-bearing. Thirty scripts written in a generic register produce thirty videos that sound like nobody. Thirty scripts from an adapter trained on your own work produce thirty videos that sound like you had a productive afternoon.
How the afternoon actually goes
The order matters, because most of the work happens before anything is produced.
- Decide the thirty topics first, in one sitting. This is the real bottleneck and it is a thinking job, not a production job. An hour with a list is worth more than any tool here.
- Draft all thirty scripts in one pass rather than one at a time. Batching keeps the register consistent and it is dramatically faster than thirty separate sessions.
- Read all thirty yourself before anything is produced. This is the step people skip and it is the only quality gate in the entire process.
- Produce them, which is the part that requires no attention from you at all and is the only part most people think about.
- Schedule them across the weeks rather than posting them together.
Getting the thirty topics right
The topic list is the whole ceiling on this, and it is the part that gets ten minutes when it deserves an hour.
The wrong way to build it is to sit down and think of thirty things, because what you produce that way is thirty things you already had opinions about, which is not the same as thirty things anyone is looking for. By about topic eighteen you will be inventing, and the last twelve will be the weakest of the batch.
The right way is to gather rather than invent. Take the questions you have actually been asked in the last three months. Take what people type into a search box, which you can see directly. Take the questions that come up in the same conversation every single time. Thirty real questions exist already and they are considerably better than thirty invented ones.
Then order them. A list that runs from the question a complete beginner asks to the question somebody asks right before they decide is a sequence rather than a pile, and a sequence gives anyone who finds one video a reason to watch the next.
The part that decides whether this works
Reading all thirty before they are produced.
The failure everyone has with volume production is that quality control does not scale with output, so it silently gets dropped. Thirty scripts is about ninety minutes of reading. That ninety minutes is the difference between thirty videos you are glad exist and thirty videos you will be quietly embarrassed by in six months, still sitting on your channel.
It is also the only stage where a mistake is cheap. A wrong sentence caught in a script costs a moment. The same sentence caught after production costs a reproduction, and caught after publishing it costs considerably more than that.
What to make thirty of
Not thirty variations of the same video, which is what happens when the topic list is written in ten minutes.
The strongest use of volume is coverage rather than repetition. Thirty genuinely different questions, each one a thing somebody actually types into a search box, each one answered properly in ninety seconds. That produces a library rather than a feed, and a library keeps working long after the feed has scrolled past.
The second strongest use is depth on one subject. Thirty short pieces that together cover one thing completely, published in order, so that anyone who finds the fourth one discovers there are twenty-nine more. That is a considerably better outcome than thirty unrelated videos, and it takes the same afternoon.
Where this goes wrong
Three failures, and the third is the one nobody warns you about.
Producing before reading. Covered above and worth repeating, because the temptation is enormous once the pipeline is working and the output is fast.
A weak source recording. Everything produced afterwards inherits the original session. Poor light, a distracted performance, or a hurried recording gets multiplied by every video that comes out of it, forever. Spend the extra hour on that one session.
Publishing all thirty at once. The point of an afternoon of production is a month of presence, not a single day of noise. Volume produced at once and released at once is worse than half the output released steadily, because consistency is what the audience and the platforms both respond to.
What it grows
Presence, which compounds in a way that individual videos do not.
One video is a piece of content. Fifty videos covering a subject completely is a position, and a position is what makes someone decide you are the person who knows about this. Nobody arrives at that conclusion from a single well-made video, and everybody arrives at it from a body of work.
The other gain is that it removes the correlation between how your month went and whether you published. That is the actual reason most agents fail at this, and it is a logistics fix rather than a character fix.
What it costs
This is one of the metered ones, so it is worth being clear.
The presenter and the voice are usually a subscription, and production is charged by the minute or the video. That makes an afternoon of thirty a real cost rather than a free one, and it is the reason the reading stage matters so much: reproducing something because nobody checked the script is paying twice.
The scripts, if they come from an adapter trained on your own writing, cost nothing. That is the half of the workflow that runs free, and it is also the half that determines whether the expensive half was worth it.
Questions agents actually type
Can I make videos without filming every one?
Yes. You film one good session to create a presenter that looks and sounds like you, and every video after that is produced from a script rather than a shoot. The quality of everything afterwards is set by that single original recording.
How many videos can I produce in one session?
Thirty is a realistic afternoon once the topics and scripts are ready. The production itself is unattended. The real constraint is deciding the topics and reading the scripts, which is roughly a couple of hours of actual thinking.
Will AI-produced video sound like me or like a robot?
That depends almost entirely on the script rather than the technology. Generic scripts produce generic videos regardless of how good the voice is. Scripts written by a model trained on your own writing are what make the result sound like you had a productive afternoon.
Should I publish a batch of videos all at once?
No. Schedule them across the weeks. The value of producing a month in an afternoon is a month of consistent presence, and releasing everything on one day converts that into a single day of noise followed by silence.
Season 1 is Claude Code for Real Estate Agents, filmed and on its own hub. All three seasons: AI for real estate agents.