Comment by icer2020

The very first run needs to wait for the machine translation model to be downloaded so it won't be quick. For the subsequent runs, the model will be reused as a cache.