Speaklone: On-Device Qwen3-TTS for iOS/macOS
Solo developer launches Speaklone app enabling on-device Qwen3-TTS inference (1.7B on macOS, 0.6B on iOS) via MLX-Swift, with voice cloning from 5-30s samples and natural language voice design. It supports streaming TTS without cloud dependency, overcoming memory constraints through quantization, cache limits, and chunked decoding. Now available on the App Store.








