Writing an article to compare Depp learning speech synthesis versus Festival Speech synthesis

I need to write about and compare three text to speech synthesis approaches. In which two of them are deep learning approaches (Tacotron2, VITS), and the other one is a traditional unit concatenation approach (Festival TTS). I want to compare the frameworks based on the following concepts:
- The implementation of each framework
--- How are they each implemented
--- What are the requirement of each framework
--- What skills or knowledge each framework needs the developer to have to use the framework

- Comparing their performance in regards of
--- Computational power and storage needed for the system
--- Cost in terms of time, effort, and money
- Comparing VITS framework with Tacotron2 for low resource language, in terms of
--- How much data is needed
--- Computational power needed to train a model
--- Time needed to train a model
--- Which have a better result, when small dataset is used

- In what case study Festival performs better
- In what case study Tacotron2 performs better
- In what case study VITS performs better