Installation¶

Requirements¶

Before installing Hermes, make sure your system meets these requirements:

Requirement	Details
macOS	14.4 (Sonoma) or later — required for Core Audio Taps
Chip	Apple Silicon (M1 or later) — required for WhisperKit / Neural Engine
Disk space	~700 MB for the WhisperKit model (downloaded on first launch)

Download the DMG¶

The easiest way to install Hermes is from the GitHub releases page.

Go to the latest release.
Download Hermes-v*.dmg.
Open the DMG and drag Hermes.app to your Applications folder (or run it from anywhere you like).

Hermes is signed with an Apple Developer ID certificate, so macOS will not block it with Gatekeeper warnings.

Build from Source¶

If you prefer to build it yourself, you need Xcode 16+ and XcodeGen.

1. Install XcodeGen¶

brew install xcodegen

2. Clone and build¶

git clone https://github.com/TheWinterShadow/Hermes.git
cd Hermes
xcodegen generate
xcodebuild \
    -project Hermes.xcodeproj \
    -scheme Hermes \
    -configuration Release \
    -arch arm64 \
    build

The built app will be at build/Build/Products/Release/Hermes.app.

3. Run it¶

open build/Build/Products/Release/Hermes.app

Granting Permissions¶

On first recording, macOS will prompt you for two permissions:

Microphone Access¶

A standard system dialog will appear asking for microphone permission. Click OK. This is required to capture your voice.

Screen & System Audio Recording¶

CATap (Core Audio Taps) requires the Screen & System Audio Recording permission to capture audio from other applications.

Go to System Settings → Privacy & Security → Screen & System Audio Recording.
Find Hermes in the list and enable it.
You may need to restart Hermes after granting this permission.

Without this permission, Hermes can only capture your microphone

System audio capture (the "Them" channel) will not work until Screen & System Audio Recording is granted.

WhisperKit Model¶

On first launch, WhisperKit automatically downloads the transcription model (large-v3-v20240930_626MB). This is a one-time ~626 MB download. The model is cached locally and reused for all future sessions.

The model runs entirely on Apple Neural Engine — no GPU or CPU fallback needed on Apple Silicon.