A local code-generation and deployment skill that builds a full-stack, offline MLX Whisper audio transcription and translation Web GUI specifically for Apple Silicon Macs.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install audio2srtlocal
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install audio2srtlocal 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 audio2srt-localgen?
The audio2srt-localgen skill is a local project generation and deployment engine designed specifically for Apple Silicon Macs running macOS 14.0 or later. Unlike alternative deployment options that clone source files from remote git repositories, this unique skill runs offline by synthesizing the entire React and Python codebase from embedded local templates. This provides a secure, private, and highly reliable setup process without needing external network access to retrieve the project's source code.
Once generated, the system creates a full-stack application featuring a high-performance Python backend using aiohttp and mlx-whisper on port 8765, alongside an interactive frontend built with React, Vite, Tailwind CSS, and Zustand on port 3000. It is an outstanding example of how Openclaw Skills empower developers to rapidly provision complete, local machine learning pipelines on-device.
audio2srt-localgen Use Cases
- Generating a secure, private, and entirely local audio transcription workflow on Apple Silicon.
- Translating multi-language voice files to English or Chinese text offline using Qwen models.
- Bypassing network environments where Gitee, GitHub, or git cloning is restricted or unavailable.
- Building a native macOS speech-to-text GUI without dealing with complex cloud setups.
- Running rapid batch speech-to-text pipelines utilizing native Apple Silicon GPU acceleration.
How audio2srt-localgen Works
- Path Verification: The system asks the user for the target directory (defaulting to
~/audio2srt) and checks for existing files to avoid overwriting unless authorized. - Codebase Synthesis: The skill extracts and writes all backend configuration files, React modules, stores, stylesheets, and start scripts locally from template files.
- Python Dependency Installation: Sets up the local Python virtual environment and installs necessary libraries (
mlx-whisper,mlx-lm,aiohttp, andmodelscope). - NodeJS Installation: Pulls down NPM packages needed to drive the React, Tailwind, and Vite frontend architecture.
- Model Pulling & Server Initialization: The
start.shwrapper downloads lightweight Apple-optimized whisper-large-v3-turbo and Qwen2.5-3B models via ModelScope CLI and starts both servers. - Launch & Open: Runs the React app locally, allowing developers to navigate immediately to
http://localhost:3000to start transcribing.
audio2srt-localgen Setup
# Target your designated generation folder
mkdir -p ~/audio2srt && cd ~/audio2srt
# The AI agent executes this skill using Openclaw Skills architecture
# To manually install the dependencies after file generation:
python3 -m venv venv
source venv/bin/activate
pip install -r server/requirements.txt
# Install Node frontend dependencies
npm install
# Run the complete stack via the generated start script
chmod +x start.sh
./start.sh
audio2srt-localgen Data Schema & Taxonomy
| Directory/File | Purpose | Tech Stack |
|---|---|---|
server/transcribe_server.py |
Python Backend API service | Python, aiohttp, mlx-whisper |
server/requirements.txt |
Backend pip modules | dependencies list |
src/components/ |
Modular UI elements (DropZone, ResultPanel, SrtTranslatePage, etc.) | React, TypeScript, Tailwind |
src/store/queueStore.ts |
Frontend transcription queue management | Zustand state container |
models/ |
Downloaded weights (gitignored) | MLX Whisper & Qwen2.5 models |
start.sh |
Orchestration script for model verification and port starting | Bash |
audio2srt-localgen Advanced Features
- Automatic Model Resolution: Features native ModelScope integration to automatically resolve, fetch, and cache localized MLX-optimized weights.
- Apple Hardware Acceleration: Directly utilizes the Apple Silicon Neural Engine and GPU through MLX optimizations for ultra-fast, low-memory whisper transcription.
- Dual-Model Translation Pipeline: Connects Whisper voice transcription with a quantized Qwen2.5 model for high-fidelity subtitle translations directly on-device.
- Native Media Processing: Employs macOS
afconvertinternally to process compressed input formats like M4A and MP3 without heavy external binaries. - Self-Healing Port Recovery: Features built-in port management patterns to quickly terminate hung Node or Python server processes on ports 3000 or 8765.
SKILL.md
Loading
Related Openclaw Skills
fu-mu-gong-ke (Parents' Homework) - Parenting Psychology AI Agent Skill
An advanced parenting psychology AI agent skill powered by the HeartFlow cognitive engine to deliver empathetic, structured emotional support and boundary-conscious guidance for families.
Inventory Monitor
An automated stock monitoring tool that calculates remaining item lifespan directly from Excel and dispatches custom alerts to your enterprise chat channels.
Work Order Dispatch & SLA Monitoring
An automated facility management skill that routes maintenance tickets from communication logs into Excel, tracks SLA thresholds, and automates team notifications.
Fee Collection Management Skill
An AI-powered property fee dunning and collection automation skill based on standard enterprise management procedures.
Democracy in America AI Analysis Skill
An executable political philosophy toolkit for diagnosing groupthink, checking centralized power, and balancing liberty with equality within communities and teams.
Born a Crime: Trevor Noah's Survival & Identity Toolkit
An AI-powered life execution engine inspired by Trevor Noah's "Born a Crime" to help users navigate identity, defy boundaries, and build personal resilience.
Featured*
