WhisperKit CLI for Openclaw

A high-performance CLI for on-device Whisper transcription and Qwen3-TTS synthesis optimized for the Apple Neural Engine.

zachnagengast
v1.0.1
Mar 4, 2026
0
891
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install argmax-cli

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install argmax-cli using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is WhisperKit CLI?

WhisperKit CLI is a powerful tool designed for macOS that enables lightning-fast, private speech-to-text and text-to-speech capabilities directly on your hardware. By leveraging the Apple Neural Engine (ANE), these Openclaw Skills allow AI agents to process audio files without relying on external APIs, ensuring zero latency from network overhead and total data privacy.

The skill supports a wide range of Whisper models for transcription—from lightweight versions to high-accuracy variants—and utilizes expressive Qwen3-TTS for generating natural-sounding audio replies. Since all inference happens locally on Apple Silicon, it is the ideal choice for developers building secure, offline-capable voice interfaces.

WhisperKit CLI Use Cases

  • Transcribing user-provided voice notes or audio attachments in real-time within an agent workflow.
  • Generating expressive audio responses for AI agents using 9 built-in high-quality voices.
  • Processing meeting recordings or long-form audio files into text locally for summaries.
  • Building privacy-first voice assistants that function entirely without an internet connection.

How WhisperKit CLI Works

  1. The AI agent receives an audio file (m4a, wav, mp3, or flac) and saves it to a local temporary directory.
  2. The transcribe command is executed, invoking Whisper models that run on the Apple Neural Engine for efficient, low-power processing.
  3. For generating a response, the agent passes text to the TTS engine which synthesizes speech using the selected voice and language.
  4. The resulting transcription text or generated audio file is then returned to the user or passed to the next stage of the automation.

WhisperKit CLI Setup

To install the core components for these Openclaw Skills, use Homebrew on your macOS device:

brew install whisperkit-cli

On the first execution, the CLI will automatically download the necessary models from HuggingFace. After this initial setup, the skill functions entirely offline.

WhisperKit CLI Data Schema & Taxonomy

The skill manages audio input/output and model metadata as follows:

Data Type Description Supported Formats
Input Audio Source files provided to the agent m4a, wav, mp3, flac
Transcribed Text Clean text output from Whisper Plain text (stdout)
Generated Audio Synthesized speech output .m4a (default), .wav
Model Weights Locally cached inference models whisperkit-coreml variants
TTS Voices Pre-defined speaker profiles ryan, aiden, ono-anna, etc.

WhisperKit CLI Advanced Features

  • Local OpenAI-compatible API server support for integrating with existing transcription pipelines.
  • Word-level timestamps and context-aware prompts to improve technical term accuracy.
  • Support for style instructions in the 1.7B TTS model for dramatic or specific narrative tones.
  • Automatic model selection based on hardware capabilities to balance speed and accuracy.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*