← Back to Skills Marketplace
Transcribe Audio with Parakeet MLX
by
kylehowells
· GitHub ↗
· v1.0.0
1987
Downloads
1
Stars
2
Active Installs
1
Versions
Install in OpenClaw
/install parakeet-mlx
Description
Local speech-to-text with Parakeet MLX (ASR) for Apple Silicon (no API key).
Usage Guidance
This skill is an instruction-only wrapper around the parakeet-mlx CLI and appears to be what it says: a local ASR tool that downloads models from Hugging Face. Before installing, verify the 'uv' installer and the parakeet-mlx formula/source (review what uv will fetch), ensure ffmpeg is installed, and accept that model files will be downloaded into ~/.cache/huggingface and that network access is required. If you need to be cautious: run the install in a VM or isolated environment first, inspect the binary (checksum/signature) if available, and confirm the tool behaves as expected on your platform (note: description targets Apple Silicon). If you have Hugging Face credentials you don't want used, remove or restrict HUGGINGFACE_HUB_TOKEN/.netrc before first run.
Capability Analysis
Type: OpenClaw Skill
Name: parakeet-mlx
Version: 1.0.0
The skill bundle is benign. It defines a local speech-to-text skill using the `parakeet-mlx` CLI tool. The `SKILL.md` provides clear instructions for installation via `uv tool install` and usage examples, all consistent with the stated purpose. There is no evidence of data exfiltration, malicious execution, persistence mechanisms, prompt injection attempts against the agent, or obfuscation. The described actions, such as local file processing and model downloads from Hugging Face, are legitimate for this type of tool.
Capability Assessment
Purpose & Capability
The skill claims local ASR for Apple Silicon and instructs use of the parakeet-mlx CLI — that matches. Minor mismatches: SKILL.md says ffmpeg is required for audio processing but the registry's required binaries list does not include ffmpeg; the metadata lists no OS restriction even though the description emphasizes Apple Silicon (arm macOS), which may lead to installation on unsupported platforms.
Instruction Scope
Instructions are limited to running the parakeet-mlx CLI, installing it via 'uv tool install', and note that models will be downloaded from Hugging Face into ~/.cache/huggingface. The instructions do read/write the user's cache directory (expected for local model downloads) and require network access to fetch models; they do not request unrelated files, credentials, or arbitrary external endpoints beyond Hugging Face.
Install Mechanism
Install uses a 'uv' formula (uv tool install parakeet-mlx) which will place the parakeet-mlx binary on disk. This is not the lowest-risk 'no-install' case but is typical for CLI tools. The install spec does not include direct download URLs; however, 'uv' will fetch upstream artifacts — verify the uv package source and signatures if you need high assurance.
Credentials
The skill declares no required environment variables or credentials, which is proportionate. One practical note: Hugging Face model downloads may use any existing HUGGINGFACE_HUB_TOKEN in your environment or ~/.netrc if present — the SKILL.md doesn't mention that, so be aware model downloads could use existing local credentials. The tool will store model files under ~/.cache/huggingface (writes to your home directory).
Persistence & Privilege
The skill is not always-enabled and does not request ongoing elevated privileges. It does install a CLI binary into the system (via uv) and writes model files to the user cache — this is expected for a local model CLI and is limited to the tool itself.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install parakeet-mlx - After installation, invoke the skill by name or use
/parakeet-mlx - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of Parakeet MLX skill:
- Enables local speech-to-text transcription on Apple Silicon using Parakeet MLX ASR.
- No API key required; models are downloaded from Hugging Face on first use.
- Supports multiple audio formats and batch transcription with shell wildcards.
- Output available in txt, srt, vtt, json formats (or all at once).
- Simple installation via `uv tool install parakeet-mlx`.
- Requires `ffmpeg` for audio processing.
Metadata
Frequently Asked Questions
What is Transcribe Audio with Parakeet MLX?
Local speech-to-text with Parakeet MLX (ASR) for Apple Silicon (no API key). It is an AI Agent Skill for Claude Code / OpenClaw, with 1987 downloads so far.
How do I install Transcribe Audio with Parakeet MLX?
Run "/install parakeet-mlx" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Transcribe Audio with Parakeet MLX free?
Yes, Transcribe Audio with Parakeet MLX is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Transcribe Audio with Parakeet MLX support?
Transcribe Audio with Parakeet MLX is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Transcribe Audio with Parakeet MLX?
It is built and maintained by kylehowells (@kylehowells); the current version is v1.0.0.
More Skills