What is NovaHUD?
NovaHUD is a voice assistant for people who work with AI coding agents. On the Mac it listens for your voice, tells you what your Claude Code and Codex sessions are doing, and sends them instructions you speak or type. It also gives you dictation into any app, meeting recording with automatic notes, and a customizable notch at the top of your screen for music, shortcuts, files and sessions.
The iPhone app, Nova, records meetings and voice notes, imports call recordings, writes notes, and includes the Nova keyboard. The Apple Watch app is a portal onto your meetings, to-dos and Claude instructions.
Who is NovaHUD for?
- Developers who run one or more Claude Code or Codex sessions and want to keep them moving without switching windows.
- Anyone who wants fast, private dictation on a Mac — an alternative to Wispr Flow, Superwhisper or macOS Dictation.
- People who want meeting notes without sending audio to a cloud service.
What makes NovaHUD different?
- Local by design. Speech recognition (whisper.cpp), the voice (Kokoro on the Apple Neural Engine) and the model that summarises and routes (Llama 3.2 3B via Ollama) run on your Mac. On iPhone, transcription uses Apple's on-device SpeechAnalyzer and notes use Apple's on-device Foundation Models. Your audio is not sent to a server.
- It acts, not just listens. "Tell Claude to keep working on Life OS" reads the instruction back and delivers it to that session unless you say "stop".
- One system across devices. Mac, iPhone and Apple Watch share meetings and Claude instructions through iCloud, and the iPhone can use your Mac as its AI engine on the same Wi-Fi.
What does NovaHUD cost?
- $5 a month, or $3 a month with a referral code.
- It starts with a 3-day free trial, no card required.
- There are no API bills: the AI runs on your devices.
What do I need?
- A Mac with Apple silicon (M1 or later) running macOS 14 or later.
- For voice answers and summaries: Ollama with
llama3.2:3b. - For sending instructions to sessions: the Claude Code CLI, signed in.
- For the iPhone app: iOS 26 or later. For the Watch app: watchOS 26 or later.
Also available as Markdown · All docs in one file: llms-full.txt