Despite significant progress on current state-of-the-art image generation\nmodels, synthesis of document images containing multiple and complex object\nlayouts is a challenging task. This paper presents a novel approach, called\nDocSynth, to automatically synthesize document images based on a given layout.\nIn this work, given a spatial layout (bounding boxes with object categories) as\na reference by the user, our proposed DocSynth model learns to generate a set\nof realistic document images consistent with the defined layout. Also, this\nframework has been adapted to this work as a superior baseline model for\ncreating synthetic document image datasets for augmenting real data during\ntraining for document layout analysis tasks. Different sets of learning\nobjectives have been also used to improve the model performance.\nQuantitatively, we also compare the generated results of our model with real\ndata using standard evaluation metrics. The results highlight that our model\ncan successfully generate realistic and diverse document images with multiple\nobjects. We also present a comprehensive qualitative analysis summary of the\ndifferent scopes of synthetic image generation tasks. Lastly, to our knowledge\nthis is the first work of its kind.\n