JSX compiles to typed function calls. The type safety is in the component definitions, not the syntax.
The real question is whether video composition is actually hierarchical enough to benefit from JSX. For simple clips, probably overkill. But once you're layering talking heads, captions, b-roll, and transitions, the tree structure starts making sense.
The hardest part isn't the generation. It's the orchestration. Caching, retry logic, stitching outputs from multiple providers. A declarative layer that handles that automatically is genuinely useful.
How many software engineers are also cinematographers or directors?
I know that AI will democratize these roles and everyone can be a director, but why does it make sense to use JSX as the means to do that? It would require people to learn a new skill.
There must be a better abstraction for creating video that provides the granularity of providing direction to individual objects in a scene that doesn't require someone to understand JSX.
I think the answer is in the tagline: AI Agent writes JSX, you get videos.
Sounds like a decent approach for today. LLMs are overtrained on JSX (Claude in particular, due to Artifacts feature IIRC being originally based on React), which makes them particularly good at translating from natural language to JSX, and that in turns makes JSX a decent choice for a structured description format.
JSX is just ugly Lisp anyway, so it's not half bad a choice for something that's structured, general-purpose, flexible and well-supported by tooling.
In other words:
[You]--natural language-->[LLM]--JSX-->[Vagrai]-->Video
How could a cached video possibly take 10 seconds? I would expect, at most, that it would cost whatever is necessary to read it from disk.
In the past I would say you should be ashamed of yourself but now I don't bother.
It's similar to remotion.dev, but focuses on generative video. Uses declarative JSX to orchestrate AI calls, which makes it much more readable!
We can build a better future with these tools, how about we build it instead of this garbage.
That is an example. Given the amount of AI-generated scam content exactly like that (and worse, have you seen the ED ones?) on the internet today, creating something in the same vein seems harmless.
On the other hand, unleashing the ability to generate massive numbers of variations cheaply, and experimenting to find the optimal mixture of insecurity, fear and avarice to exploit....that's dystopian. But Madison Avenue actually got most of that figured out long ago.