What does an end-to-end TTS system take as input according to Machine Learning Yearning?

In an end-to-end TTS system as described in Machine Learning Yearning, the direct output of the model is audio.

In end-to-end TTS, _____ are used as input to the model to directly produce audio.

Match each element of the end-to-end TTS pipeline to its role in the system.

Order the stages of an end-to-end TTS pipeline from input to final output as described in Machine Learning Yearning.

Why does Machine Learning Yearning classify end-to-end TTS under 'Directly Learning Rich Outputs'?

According to Machine Learning Yearning, the TTS pipeline flows from audio input to text feature output.

Machine Learning Yearning (p. 103) describes TTS as mapping text features to _____ as its rich output.

Match each term to its description in Machine Learning Yearning's treatment of end-to-end TTS.

Order the reasoning steps for classifying an end-to-end TTS system as a 'directly learning rich outputs' problem.