For language generation, AFAIK there is no good model that follows this architecture. For image generation, Generative Adversarial Networks are strong contenders. See for instance:

https://github.com/Newmu/dcgan_code