Why is it challenging to apply diffusion models to text generation?

Comprehensive and Detailed In-Depth Explanation:

Diffusion models, widely used for image generation, iteratively denoise data from noise to a structured output. Images are continuous (pixel values), while text is categorical (discrete tokens), making it challenging to apply diffusion directly to text, as the denoising process struggles with discrete spaces. This makes Option C correct. Option A is false—text generation can benefit from complex models. Option B is incorrect—text is categorical. Option D is wrong, as diffusion models aren’t inherently image-only but are better suited to continuous data. Research adapts diffusion for text, but it’s less straightforward.

OCI 2025 Generative AI documentation likely discusses diffusion models under generative techniques, noting their image focus.