After some experiments with playing text to people around me, we decided that a huge factor in the perceptive quality of the voice comes from knowing who you are listening to before you get to listen, with the best perceived quality when the listener actually gets to see the picture of the voice's owner. Was it a deliberate choice to add those photographs for this reason?

Extremely fun little tool! Anywhere I can read about the techniques involve?

An interesting quirk: some words seem to get dropped entirely? for example the word "cleverer" and any word with a hyphen.

Really fun to start with a quote from one person and switch between voices to hear others recite the same line. Alan Rickman doing lines from Aladdin as Iago is pretty funny

You’ve clearly spent a good deal of your time creating this. Bravo. What steps can I take to find more time to dig in projects of my own? Assuming this is not what you make a living out of.

Hey! What do you think about running this on a mobile device? What are the current limitations? Since WaveNet we are able to get really good TTS but it's always online. Why aren't there any good TTS engines yet? E.g. for Pocket or similar services

Absolutely amazing. Are there any plans to make it open? Open Sourcing it? Paid API in the future? Something like that?