Train and integrate models for different styles

[ramongduraes]

Currently implemented model doesn't require pre-training a model per style. (See issue #12 )