v0.1.4

Parallel startup + streaming playback

Model loads in a background thread while text/G2P/voice resolve on the main thread. Audio chunks stream to speakers as they're generated — first chunk plays immediately.

Changes

• std::thread::spawn model loading, overlapped with G2P + voice loading
• Streaming playback via rodio Player::append per chunk
• Re-export mlx_rs::Array from voice-tts

Timing ("The quick brown fox...")

	Wall	Audio	First speech
say	3.03s	2.50s	~0.53s
voice	3.71s	3.28s	~0.43s

Previous

v0.1.3 — builtin voices, CI, clippy clean