Openai Whisper

OpenAI Whisper provides local speech-to-text transcription directly on your machine without requiring API keys or cloud services. It wraps the Whisper command-line tool to convert audio files into text, subtitles, or translated transcripts — keeping all processing local for maximum privacy and zero ongoing costs.

Key Features

• Fully local processing with no cloud dependency, API keys, or recurring costs — all transcription happens on your machine
• Multiple model sizes from tiny to large, allowing you to trade off between speed and accuracy based on your needs (turbo is the default)
• Flexible output formats including plain text, SRT subtitles, VTT, and other standard caption formats
• Translation capability that can translate audio from other languages directly into English text
• Automatic model caching in

  ~/.cache/whisper

  so models only download once and are reused across sessions

Use Cases

• Transcribing meeting recordings, interviews, or voice memos as part of an agent workflow
• Generating subtitles for video content without sending audio to external services
• Processing sensitive or confidential audio where cloud-based transcription is not permitted
• Building local audio-to-text pipelines for content creation, note-taking, or accessibility

How It Works

The skill wraps OpenAI's open-source Whisper command-line tool. Users provide audio file paths and specify parameters like model size and output format. On first use, the selected model downloads and caches locally. Subsequent runs use the cached model for fast startup. The turbo model provides the best balance of speed and accuracy for most use cases.

Getting Started

Install the Whisper binary via Homebrew with

brew install openai-whisper

. The skill will then be ready to use — simply provide an audio file path and your preferred output format. No API keys or external service configuration is needed. Models cache locally in

~/.cache/whisper

and require several GB of disk space depending on the model size selected.