Amodal content completion is the reconstruction of the invisible region(s) of occluded objects by inferring their missing appearance, style, and texture from the available visual context. Despite recent advances in image completion, existing methods often struggle to accurately recover the content of the occluded areas and fail to generate contextually plausible completions or correctly capture long-range dependencies. To address these limitations, this paper proposes a novel amodal content completion network, that integrates semantic priors from a pre-trained diffusion model with local structural representations extracted by a gated convolutional encoder. A hierarchical transformer applies self-attention on the multi-resolution features to guide the decoder to reconstruct the occluded content. Experimental results on the COCOA dataset and two of its subsets demonstrate the effectiveness of the proposed approach in generating visually plausible and structurally coherent completion compared with the baseline methods.