In recent years, text-to-image synthesis has drawn the attention of many researchers. Currently, GANs (generative adversarial networks) have been shown to deliver images from complex text. Though GANs provide reasonable images, obtaining accurate results is beyond the current artificial intelligence systems. On the other side, deep convolutional generative adversarial networks have been proven to compile images from specific types such as birds, cars, and dogs. In this work, we have combined two algorithms GAN-CLS from the authors Scott Reed and MS GAN (mode seeking generative adversarial networks) from the authors Qi Mao and Hisn-Yang Lee. One of the limitations of GAN-CLS is that while training the model is experiencing mode collapse problems. To overcome the mode collapse problem, we have combined MSGAN with GAN-CLS. We demonstrate the above-proposed algorithm by implementing it on the Birds Dataset. We have trained the model till 936 epochs.