Improve AudioBook support

[mekarpeles]

Features we definitely want:

• A. Support for non-English (allowing our 600k non-English books to be read)
• B. Reduce bandwidth (generating mp3's is not accessible to all and slow to load)
• C. Pause button
• D. Playback speed controls
• E. Backwards compatibility with Festival (where possible)

Implementation

• Use Web Speech API (which provides options for many of these) where possible. Investigate how many of these features are also compatible Festival so we don't have to support two separate UIs if we have to fallback to Festival.
• Update the UI to provide access to these options.

Plan

(NOT CEMENTED YET)

1. Refactor plugin.tts.js to include class SpeechSynthesisEngine #129: Refactor plugin.tts.js to include an abstract class, SpeechSynthesisEngine, and a concrete class FestivalSpeechEngine (necessary for E)
2. Create/use BrowserSpeechEngine class in plugin.tts.js #130: Edit plugin.tts.js to include a class BrowserSpeechEngine which implements SpeechSyntehsisEngine, and which is used on supported browser (Completes B)
3. BrowserSpeechEngine should choose voice based on book language #132: BrowserSpeechEngine should choose voice based on book language (Completes A)
   • Create function to convert book metadata language to BCP47 language code #131
4. UX Planning - Determine what we want the audiobook experience to be like. This will dictate the next steps. Results of this planning will be posted on/used to close Design Read Aloud UI Options #118 .

Motivation

Each year millions of patrons use the Internet Archive’s BookReader to explore book-formatted content. Many readers are surprised to learn Bookdeader has a read-aloud button which enables any book to be read out loud by a friendly robot voice. While for many this feature is icing on the cake, others with sight and print-disabilities depend on this feature to access our books. Our current read-aloud offering gets the job done, but there are a handful of major problems and opportunities for improvement. By switching from our current system, which uses Festival text-to-speech and generating mp3 files on the fly, to using the browser’s native Web Speech API, we could fix 5 major issues:

• Generating mp3's is high bandwidth (not accessible to all users) and slow to load.
• Our current read-aloud has no way to pause and resume your position!
• There’s no UI for changing the playback speed of our Festival voices.
• In most browsers, the Web Speech API voices are superior to our Festival voices.
• Today, all our books are read with an English accent -- Web Speech API offers multiple voices and the ability for books to be read in their native language.

We have a working prototype of our books being read using the Web Speech API. The next step is to integrate the Web Speech API into our native Internet Archive BookReader, and we already have a start on the feature/native-audioreader branch!

Related to #58

[mekarpeles]

Discussion Requires:

• if in read-mode, do we still show images v. text?
• How/can a reader choose where to start reading on the page — always the top?
• What does he design/UI look like for start / pause / speed / voice (options we don’t really have today)?
• Can we dynamically choose the right (language) voice based on the book language field? This would make an amazing demo (imho)

[mekarpeles]

A basic API (which wraps/extends the Web Speech Synthesis native API) may look like.

• BookReader.start
• BookReader.stop
• BookReader.pause
• BookReader.startAt(sentenceNum)
• BookReader.setVoice(id)
• BookReader.setVolume(vol)
• BooKReader.highlight(sentenceNum)
• BookReader.setPitch(pitch)
• BookReader.setSpeed(speed)