Video content engine
Agents take a topic through the script, the avatar render, the thumbnail, the captions and the upload, then stop and wait for your approval. Long-form on Sunday, Shorts through the week. No camera day, no editor queue, no month where nothing ships.
Running on a US client channel since June 2026: 41 pieces published on a fixed weekly slate.
Why nothing ships
The usual way
With the engine
The machinery
Every idea passes a refinement step and is checked against everything the channel has already published, so the queue does not quietly repeat itself six months later.
Written against your voice document and the formats that have measurably worked on your channel, never from generic priors. Long-form in full, Shorts on a four-beat structure.
Your avatar renders the piece at 1080p from the start, so what you review is final resolution and approving it does not trigger another render.
Generated against your channel's thumbnail spec, three variants scored by a rubric, then composed on your own template so the channel stays visually consistent.
Word-by-word captions because a large share of Shorts viewers watch on mute, a visual change every few seconds, and the cut assembled ready to ship.
Uploaded unlisted with the title, description, tags, thumbnail and caption track already attached, waiting for someone to press schedule.
The week
The agent takes what is due off the researched topic queue and writes it. Shorts stop right here and wait for your OK, because we wrote them and you should see the words first.
A scheduled routine wakes up, reads the board and acts on everything sitting at production: the render, the thumbnail, the captions, the assembly and the upload. Nobody has to start it.
You watch the finished video and approve it. This is the thing the system will never do for itself, and nothing reaches YouTube before it happens.
Once approved, the piece goes up unlisted with full metadata, thumbnail and captions attached, so the last human step is scheduling rather than assembling.
Shorts on Tuesday, Thursday and Saturday afternoon, long-form on Sunday. Those times are not a guess; they moved to afternoons after the channel's own analytics showed what noon was costing.
The dashboard
Not a report about the system. The board is the control panel: approving a piece there is what releases it, and the agent reads the same board on its next run.
The verdict in plain language. What is working, what is not, and what to do about it next.
Channel and per-piece numbers pulled live from the YouTube API, down to the retention curve for each video.
What is waiting for approval, what is scheduled, what is live, and which slots are still open.
The running to-do list, so nobody has to remember where a piece stopped or who it is waiting on.
Fit
Proof
The system runs the agent recruiting channel of a sponsor at Real Brokerage, attracting agents across the US. We keep our clients and their channels unnamed, and we would do the same for yours. Every number below comes straight from the YouTube API rather than an estimate, and that includes the parts that are not working yet.
The channel is young: nine subscribers and roughly 5,900 views. The system worked out why from its own analytics, and the answer is not reach. Shorts average around 220 views and retention sits near 39%, so people do watch. Almost nobody subscribes, about one per 686 views. So the work moved to the conversion end, the hooks and the reason to come back, instead of making more of something already being watched. That diagnosis came out of the analytics pull with the evidence written down, and it is exactly the part an agency reporting on its own work would have left out of the slide.
How to start
Two parts, and a third only once the first is working. We stand the system up on your brand and your accounts, then it runs your slate every week. Neither part is billed by the hour.
Where everyone starts
A one-time build
Monthly, ongoing
Once one channel works
The price depends on the cadence, whether long-form is in the slate, and how much of your material already exists. You get a quote with an exact number within two business days.
Questions
It writes from your voice document, your brand thesis and what has measurably worked on your channel, and never from generic priors. That is a rule inside the system, not a preference we try to remember. It still needs your material to work with: on the running channel the client writes the long-form scripts himself and the system produces from them. If you have nothing to say, this will not invent it for you.
No, and that is most of the point. Your avatar is built once and renders every piece after that, at 1080p from the start. The running channel has had zero camera days. It is also disclosed rather than hidden: on that channel, one of the better performing Shorts is the one where the client says outright that the face on screen is an AI version of him. Pretending otherwise would be both dishonest and worse content.
Approve. Shorts wait for an OK on the script, because we wrote them, and every piece waits for your OK on the finished video before anything reaches YouTube. Both happen in the dashboard, and both take minutes. If you want the long-form words to be yours, you write those scripts and the system produces from them. Everything between those points runs without you.
You do. It all runs on your YouTube channel, your Drive and your accounts, and the avatar, the voice, the scripts and the thumbnails are yours. We are not a layer you have to keep paying in order to keep reaching what was already made.
We will not put a number on that, and anyone who does is guessing. What we do commit to is output and honest measurement: the slate ships, and every month you get a straight read of what the analytics say, including when the answer is that something is not working. On the running channel that read is currently "reach is fine, subscriber conversion is not", which is a far more useful sentence than a growth promise.
Yes. It is built as one shared engine and separate lines, where a line is a channel with its own brand, voice and published history. Lines never mix and never repeat each other's topics. Adding one is a copy of the template plus that line's own material, on the same dashboard.
Forty minutes looking at how things stand. It tells us how far we can take you towards being an AI-native company, and what that will cost you at the end of the day.
Jaro Rais
AI consultant and developer
Reach out:
jaro.rais@itashu.coGet in touch and we will come back to you within one working day.
By submitting you agree to your data being processed as described in the privacy policy.