voice-recognition
by gykdly
AI & LLMsBrief
Local speech-to-text with OpenAI Whisper CLI.
Description
What This Skill Does
Runs OpenAI Whisper locally to transcribe audio files without sending data to any external service. Supports 100+ languages with options to translate speech to English or generate a text summary.
Runs entirely offline so audio never leaves your machine, unlike cloud transcription services that require API keys and per-minute billing.
When to Use It
- Transcribing recorded meetings or lectures
- Converting voice memos to text
- Translating foreign-language audio to English
- Getting a quick summary of a long recording
- Batch processing audio files from the command line
Example Workflow
User asks: transcribe this Chinese meeting recording and summarize it
- 1
Receives audio file path (e.g. meeting.m4a)
- 2
Runs voice recognition script with --zh and --summarize flags
- 3
Whisper loads the medium model and processes the audio locally
- 4
Script outputs full transcript followed by a generated summary
Full Chinese transcript with an English summary paragraph
Requirements
- openai-whisper installed via Homebrew (brew install openai-whisper)
Installation
clawhub install gykdly/voice-recognitionRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.