speech-swift

Why we included this project

This Swift package is a strong fit for Mac and iOS teams that want real speech features without sending audio to the cloud. It bundles automatic speech recognition, text-to-speech, voice cloning, speaker diarization, and voice activity detection into one library that runs on Apple Silicon through MLX and CoreML, so recordings stay on the device. That makes it useful for dictation, transcription, and voice agent products where privacy is a hard requirement. The model selection is wide, with dozens of ASR and TTS options from lightweight CoreML builds tuned for the Neural Engine to larger MLX models, plus streaming variants with low real-time factors for live conversations. A handful of public apps already build on it, which gives you working reference code beyond the documentation.