Speech-to-Text Transcription

You can transcribe audio files to text using the OpenAI Whisper CLI tool.

Prerequisites

The following must be installed on the system:

• openai-whisper: pip install openai-whisper
• ffmpeg: brew install ffmpeg (macOS) or apt install ffmpeg (Linux)

Verify installation: whisper --help

Usage

To transcribe an audio file:

whisper <audio_file> --model base --output_format txt --output_dir /tmp

The transcript will be saved as a .txt file in the output directory.

Common Options

Option	Description	Values
--model	Model size (larger = more accurate but slower)	tiny, base, small, medium, large
--language	Source language (auto-detect if omitted)	zh, en, ja, ko, etc.
--output_format	Output format	txt, srt, vtt, json
--output_dir	Directory for output files	Any writable path

Recommended Models

• tiny/base: Fast, suitable for clear speech in common languages
• small: Good balance of speed and accuracy
• medium/large: Best accuracy, recommended for noisy audio or uncommon languages

When to Use

When you receive a message containing an audio file path (e.g., [User sent a voice message: /path/to/audio.ogg]), use this skill to transcribe it:

1. Run the whisper command on the audio file.
2. Read the resulting .txt file.
3. Use the transcript to understand and respond to the user's request.
4. Clean up temporary output files.

If whisper is not installed, inform the user that speech-to-text requires installing openai-whisper and ffmpeg.