Why are there very few synthetic data startups in text/language as opposed to computer vision which is flooded with synthetic data startups?

I was doing a review of the synthetic data startup space and noticed that I couldn't find even one company which works on synthetic text generation in a significant way.

I suspect that CV/Object simulation have high costs which is why the incentive may be high for generating synthetic object simulation.

Is that the reason why?

computer computer vision data datascience language startups synthetic data text vision