Overview

Voice Recognition uses the OpenAI Whisper CLI to convert speech to text locally, with no API key required. It supports 100+ languages, can translate output to English, and can generate a quick summary of the transcription. This makes private, offline transcription workflows fast and privacy-preserving.

How This Skill Works

The tool runs OpenAI Whisper locally, downloading the default medium model to ~/.cache/whisper on first use. You can force language output with --zh or --en, enable translation with --translate, and generate a summary with --summarize. It accepts common audio formats such as MP3, M4A, WAV, OGG, FLAC, and WebM.

When to Use It

• Transcribe private interviews or meeting recordings offline without sending data to the cloud
• Transcribe (and optionally translate) multilingual content to English for wider sharing
• Create concise summaries of long lectures or seminars
• Process offline voice notes when internet access is restricted
• Build a privacy-preserving, local transcription workflow for teams

Quick Start

1. Step 1: Install Whisper CLI and Python 3.10+ (e.g., brew install openai-whisper; ensure Python 3.10+ is available)
2. Step 2: Transcribe an audio file using the main script, e.g. python3 path/to/scripts/voice识别_升级版.py audio.m4a
3. Step 3: Add features as needed, e.g. python3 path/to/scripts/voice识别_升级版.py audio.m4a --translate --summarize

Best Practices

• Use the language flags (--zh for Chinese, --en for English) to improve accuracy when known
• Start with the default medium model; switch to larger models only if hardware permits and accuracy is insufficient
• Ensure input audio is in a supported format (MP3, M4A, WAV, OGG, FLAC, WebM)
• Only enable --translate and --summarize when you need them to save time
• Note that the first run downloads the Whisper model to ~/.cache/whisper

Example Use Cases

• Transcribing a Chinese podcast offline for local archiving using the script
• Transcribing a multilingual interview and translating output to English with --translate
• Generating a quick summary of a 90-minute lecture with --summarize
• Transcribing a private Zoom meeting locally to keep data on-device
• Setting up a private transcription workflow for a research team

Frequently Asked Questions