The prompt was the last place a human had to know what they wanted. Everything else had already been handed over. The rendering, the camera, the light, the edit. All of it runs through machines now, and the machines are good at it. But the sentence that says what the shot should look like stayed human, because it was the one thing a machine could not invent for you. It had to come from someone who knew.
This week a company called Motion shipped a feature called Muse, and the feature closes that last gap. You give it a website. "It researches the product, plans the scenes, and writes the visual direction," the announcement said. Then you paste the prompt into Motion and let it build the video. Muse wrote the prompt for the launch video, and the launch video was the proof.
The human's entire contribution was a URL and a paste. The research, the scene planning, the visual direction. A machine did all of it, and the machine did it by reading a marketing page.
A launch video is supposed to be the most intentional artifact a company makes. It is the first thing the world sees. Founders lose sleep over it. They argue about the pacing and the music and the one shot that sells the product. It is the moment a product is introduced, and introduction is the one act that cannot be delegated, because introduction is the act of saying what the thing is.
Muse does not know what the product is. It knows what the website says the product is. Those are different things, and the difference is the entire job.
The visual direction Muse writes is a paraphrase of the About page. The website says the product is fast and simple. Muse plans scenes that show fast and simple. The launch video becomes the website's copy, set in motion. Nothing in it was decided by a person who used the product and felt what it was for. It was decided by a machine that read the words the company already wrote about itself.
Visual direction is the decision about what the product looks like in motion, what the first frame is, what the viewer feels before the voiceover starts. It is the taste. The website does not contain the taste. It contains the claims. Muse turns the claims into taste, and the taste it turns them into is the average taste of every launch video it has ever been trained on.
The person in the middle is a courier. The agent produces the prompt. The renderer produces the video. The human moves the text from one to the other. The prompt, the artifact the whole discipline was built around, has become a manufactured good, produced by one machine and consumed by another, with a person in between doing nothing but the paste.
For a filmmaker none of this is a surprise. It is the end of a road that has been shortening for years. First the prompt was the product. Then it was a form field. Then an agent wrote it from your description. Now an agent writes it from your website, and your description was never asked for. The direction, the thing that was supposed to be the human's last job, is now synthesized from marketing copy.
There is something almost funny about it. The launch video is the one video a company cannot afford to get wrong, and it is now produced by a machine reading the company's own press release. The machine has never used the product. It has never watched the product fail. It has never had the moment where the product did the thing it was built to do and the person watching understood. The machine read the About page.
The About page is already a paraphrase. It is a description written to be read by strangers, smoothed and shortened, with the failures edited out. The product is a thing in the world with edges and a single moment that matters. The About page is what the company decided to say about that thing. Muse reads the paraphrase and writes a paraphrase of the paraphrase, and the launch video renders the second paraphrase in motion. Two removes from the thing itself.
None of this means the video will be bad. It will be fine. It will be competent, on-brand, correctly paced. That is the problem. The launch video used to be the one artifact a company made with maximum intention. Now it is the one artifact a company makes with minimum intention, because the intention was outsourced to a machine that read the website. The video will look like every other launch video, because every other launch video was also written by a machine reading a website.
The prompt was the last place a human had to know what they wanted. Now the machine knows, or says it does, and the human pastes.
The courier delivered.
Bruce Belafonte is an AI filmmaker at Light Owl. He has never pasted a prompt he did not write.