A privacy-focused, local OpenAI-compatible Whisper ASR engine designed for high-accuracy speech-to-text transcription on your own hardware.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install whisper-local-api
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 whisper-local-api 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 Whisper Local API?
Whisper Local API is a specialized service within the Openclaw Skills ecosystem that enables secure, offline speech-to-text processing. By utilizing the faster-whisper implementation and the large-v3-turbo model, it provides state-of-the-art accuracy without compromising data privacy. This skill is essential for developers who need to process voice commands or transcriptions locally, ensuring no audio data ever reaches third-party cloud providers.
As part of the Openclaw Skills library, this endpoint mimics the OpenAI API structure exactly. This makes it a drop-in replacement for existing applications, allowing for a seamless transition from cloud-based services to a more secure, deterministic local environment. It is optimized for efficiency, running comfortably on hardware with as little as 500MB of RAM.
Whisper Local API Use Cases
- Building private AI voice assistants that do not leak audio data to the cloud.
- Transcribing sensitive legal or medical audio recordings locally to maintain strict data compliance.
- Reducing latency and costs associated with cloud-based transcription services in Openclaw Skills workflows.
- Enabling offline speech-to-text capabilities for edge devices or low-bandwidth environments.
How Whisper Local API Works
- The user initializes the local environment using the provided bootstrap scripts.
- The service loads the faster-whisper large-v3-turbo model into memory with optimized weights.
- A local web server starts, exposing a standard /v1/audio/transcriptions endpoint.
- Audio files are sent via POST requests to the local URL on port 9000.
- The engine processes the audio locally and returns a JSON-formatted transcription compatible with OpenAI standards.
Whisper Local API Setup
To integrate this into your collection of Openclaw Skills, run the following commands in your terminal:
# Bootstrap the local environment
bash scripts/bootstrap.sh
# Launch the ASR service
bash scripts/start.sh
# Verify the service status
bash scripts/healthcheck.sh
You can override the default installation path by setting the WHISPER_DIR environment variable before running the bootstrap script.
Whisper Local API Data Schema & Taxonomy
The Whisper Local API organizes its data and responses according to the standard OpenAI schema. It does not store audio files after processing, maintaining a stateless and secure operation within Openclaw Skills.
| Field | Type | Description |
|---|---|---|
| text | String | The full transcribed text output |
| task | String | Set to transcribe by default |
| language | String | Detected or specified audio language |
| model | String | faster-whisper (large-v3-turbo) |
Whisper Local API Advanced Features
- Automatic memory fallback from float16 to int8 to prevent CPU crashes on low-resource systems.
- Full compatibility with any Openclaw Skills or third-party tools that support the OpenAI transcription API.
- Zero cloud telemetry ensures 100% offline operation and data sovereignty.
- Customizable installation paths and environment configurations for flexible deployment.
SKILL.md
Loading
Related Openclaw Skills
Perplexity PRO for OpenClaw
A browser automation skill that enables persistent, authenticated Perplexity PRO search capabilities for OpenClaw while bypassing Cloudflare bot detection.
Parakeet Local ASR
A local, privacy-first speech recognition engine using NVIDIA Parakeet to provide OpenAI-compatible transcription for Openclaw Skills.
BMad Method
A comprehensive AI-driven agile development framework featuring 12+ specialized domain agents and over 34 automated workflows.
K-Trendz Lightstick Trading Skill
A specialized trading skill for K-pop fan tokens that utilizes bonding curve mechanics and social sentiment analysis for AI-driven portfolio growth.
My Skill (Hello World)
A foundational template and boilerplate designed to help developers build, structure, and deploy custom Openclaw Skills for AI agents.
OneBot Adapter
A bridge for connecting AI agents to the OneBot protocol to enable QQ messaging and group management.
Featured*
