This document presents a method for 3D object generation using a stacked generative adversarial network (3D StackGAN). It consists of an Inception generator and discriminator in Stage 1 to generate low-resolution objects from text, and a Batch Normalization generator and discriminator in Stage 2 to generate high-resolution objects. The generators map from a latent space to 3D objects while the discriminators classify objects as real or generated. Experimental results showed the model could generate novel and realistic 3D objects after training.