Image generation requires a text-to-image model, which is not part of this single-model (LLaVA 1.5 7B) build.