This is the true power of generative AI, enabling new functionality for the user with simple UX while doing all the heavy lifting in the background. Prompting as a UX should be abstracted away from the user.
The model need only recognize from the shape that it is a horse, and would know to extrapolate from there. It would presumably have some text encoding as residual from training, but it doesn't need to be fed text from the text encoder side to know that. Think of the CLIP encoder used in stable diffusion.