A paper called Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networks (http://arxiv.org/abs/1604.04382 http://arxiv.org/abs/1604

A paper called Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networks (http://arxiv.org/abs/1604.04382 http://arxiv.org/abs/1604.04382) uses a different approach with similar results and runs about 500 times faster giving 0.25M pixel images at 25Hz. It also seems to result in less randomness from frame to frame https://www.youtube.com/watch?v=PRD8LpPvdHI https://www.youtube.com/watch?v=PRD8LpPvdHI Code at https://github.com/chuanli11/MGANs https://github.com/chuanli11/MGANs