A general-purpose AI model is now building promo videos from a text description

What was reported

As Jasmine News reported, people are now talking about general-purpose AI models in video, not only the dedicated video generators. The example it leads with is Claude Opus 5.5. You describe what you want, and the model writes its own code, finds material, records a voiceover, makes animations and stitches it all into a complete promo film.

One overseas user reportedly used it to make a modern, stylish promo for a startup focused on reasoning, at a cost of only $2. We have no sample output to inspect, and we could not confirm the claim with Anthropic in today's material, so read it as a reported demo.

Why code makes video editable

Dedicated video models usually work by chance: you generate a clip, dislike a detail, and generate again. Jasmine News argues this model's strength is not how realistic the picture looks but how easy it is to change. Because the video is built from code, fixing one shot means changing a few lines, and things like style and parameters can be set precisely.

The outlet calls this approach Vibe Video. It says the approach is taking hold in motion graphics, explainer content and small games, and that a GitHub showcase site, awesome-opus5-5-videos, already collects many examples.

Who feels the pressure

The article's argument is that this suits business content that gets revised again and again, such as copy, colors and data charts. Video production shifts from a roll of the dice to something closer to programming, and each extra revision costs far less.

It also describes a similar pattern with OpenAI's GPT-6 Astra, which one creator reportedly used to research, script, voice and edit a video about its own launch. Dedicated video companies, the report says, are responding by building full creation workflows rather than competing only on picture quality.

What to keep in mind

Cheap and editable does not mean polished. Code-built video is good at graphics and explainers, and the source does not claim it matches realistic footage. Details such as length, quality and how much human cleanup was needed are not given.