WWDC21 Challenge: Speech Synthesizer Simulator

At a glance

Project structure

Speech Synthesizer Simulator/
├── ViewController.swift        # UI, AR scene, animation
├── Conversation.swift          # speech queue and delegate
├── ScriptReader.swift          # text-to-Quote parsing
├── Scene.swift                 # AR anchor placement
├── DialogueLabel.swift
└── SpriteHelper.swift

The split is by framework responsibility, not by formal layers.

Overall architecture

flowchart LR
    VC["ViewController"] --> Scene["ARSKView / Scene"]
    VC --> Conversation
    Script["conversation.txt"] --> Reader["ScriptReader"]
    Reader --> Quotes["[Quote]"]
    Quotes --> Conversation["Conversation"]
    Conversation --> Synth["AVSpeechSynthesizer"]
    Synth --> Conversation
    Conversation --> Bus["NotificationCenter"]
    Bus --> VC

Reference code

Speech Synthesizer Simulator/Conversation.swift:22 — the speech coordinator owns parsing and the framework delegate (abridged).

override init() {
    super.init()
    synthesizer.delegate = self
    quotes = ScriptReader().parseFile()
}

let synthesizer = AVSpeechSynthesizer()

Ownership and state

classDiagram
    ViewController *-- Conversation
    ViewController o-- ARSKView : outlet
    Conversation *-- AVSpeechSynthesizer
    Conversation *-- Quote : parsed list
    Conversation ..|> AVSpeechSynthesizerDelegate
    Conversation --> NotificationCenter : publishes
    ViewController --> NotificationCenter : observes

Ownership evidence

Speech Synthesizer Simulator/ViewController.swift:38 — the screen retains both script and speech helpers alongside presentation state.

var scriptReader = ScriptReader()
// Conversation abstracts AVSpeechSynthesizer details.
var conversation = Conversation()

NotificationCenter.default.addObserver(
    self,
    selector: #selector(updateLine(_:)),
    name: Notification.Name("AVLineData"),
    object: nil
)

Conversation is the speech mutation authority; the controller owns scene/animation state. NotificationCenter.default is shared infrastructure, not owned by either object.

Class and protocol design

Conversation: NSObject, AVSpeechSynthesizerDelegate isolates the framework protocol. ScriptReader and Quote are value types with deterministic parsing. Scene: SKScene owns anchor placement. No app-defined protocol is present; the sample uses framework inheritance and named notifications directly.

Access control

All declarations use Swift’s internal default; the source contains no explicit private, fileprivate, or public declarations. That keeps a short challenge easy to modify, but it means ownership is conventional rather than compiler-enforced. A production version could make synthesizer state and parser helpers private without changing collaboration.

Logic ownership and placement

Design patterns

Naming conventions

• Framework coordinators use role nouns (Conversation, ScriptReader, Scene).
• UI callbacks use action/update verbs (pressPlay, updateSpeaker, updateLine).
• Notification names use the AV...Data prefix, although typed static names would be safer at larger scale.

Architecture takeaways

• Wrap a stateful framework object in one owner that adopts its delegate.
• Parse external script data before it enters playback logic.
• Keep scene presentation and speech sequencing separate even in a small sample.
• The intentionally broad internal access reflects a challenge scaffold, not a recommended public API surface.

Source map

• Official Apple documentation