apple-llm v0.3 Adds Local TTS and STT Features

On November 6, 2026, a developer posted an update to apple-llm in the Reddit community LocalLLaMA: version 0.3 brings local text-to-speech and speech-to-text to the package. apple-llm, which the developer had introduced to the community some time earlier, is a Node and Python package that calls the LLMs Apple ships with macOS, with no API key required. Version 0.3 extends the tool to cover the rest of what already exists on a Mac, namely the system voices and Apple's speech recognition, closing the loop between local LLMs and local speech. According to the developer, the new features need no sign-up whatsoever, and all audio is processed locally on the machine without leaving it. The narration in the accompanying video and every demo voice were produced by the package running on the developer's own Mac, and nothing was recorded in advance. Installation is a single command: npm install -g apple-llm, or pip install apple-llm for Python users, and speech output is invoked like apple-llm speak "Hello there". Performance-wise, the developer reports that on an M4 Air the package renders 30 minutes of audio in under 20 seconds. That M4 Air figure, from an Apple silicon laptop, underlines how far on-device speech has come. As of now the latest published state is this 0.3 release; the available material contains no further updates or stated plans for subsequent versions.