The scarcity of historical financial data has been a huge hindrance for the development algorithmic trading models ever since the first models were devised. Most financial models assume as hypothesis a series of characteristics regarding the nature of financial time series and seek extracting information about the state of the market through calibration. Through backtesting, a large number of these models are seen not to perform and are thus discarded. The remaining well-performing models however, are highly vulnerable to overfitting. Financial time series are complex by nature and their behaviour changes over time, so this concern is well founded. In addition to the problem of overfitting, available data is far too scarce for most machine learning applications and impossibly scarce for advanced approaches such as reinforcement learning, which has heavily impaired the application of these novel techniques in financial settings. This is where data generation comes into play. Generative Adversarial Networks, GANs, are a type of neural network architecture that focuses on sample generation. Through adversarial training, the GAN can learn the underlying structure of the input data and become able to generate samples very similar to those of the data distribution. This is specially useful in the case of high-dimensional objects, in which the dimensions are heavily inter-dependent, such as images, music and in our case financial time series. In this work we want to explore the generating capabilities of GANs applied to financial time series and investigate whether or not we can generate realistic financial scenarios.