voiceover skill

Uses the oxframe skill (workflow, full API cheat sheet); install both with npx skills add oxygenframe/oxframe-sdk.

#[oxframe::scene]
async fn intro(view: View) {
    let square = view.add(Rect::new().size(200.0).fill(Accent).opacity(0.0));
    let line = voice("First we draw a square, then we turn it").start();
    line.at_word("square").await;
    square.fade_in().await;
    line.at_word("turn").await;
    square.rotation().to(90.0, 1.0).await;
    line.await;                       // the scene must not end before the line
}

• voice(text).start() speaks now; .at_word("word") waits for that word; awaiting the line waits for its end. check reports voice_cut and voice_overlap.
• oxframe does not call a speech service. Implement TtsProvider (id, synthesize → Synthesis { rate, samples, words }) and pass it with Project::tts(..); read the API key from the environment. Without a provider or key, timing is estimated and there is no audio.
• The first render caches audio in tts/; commit that folder so CI renders the real voice without a key.
• Music: Project::audio(Audio::track("assets/music.mp3").duck_under_voice()).
• Subtitles: oxframe render --subtitles out.srt (.vtt by extension).