OpenAI Whisper CLI

What this skill does

• •Transcribe audio files into local text documents
• •Perform real-time speech translation into English
• •Output generated content as structured SRT files
• •Configure inference models from turbo to high-accuracy variants
• •Manage local model caching within the user directory

When to use it

• ✓Converting long project meeting recordings into searchable notes
• ✓Generating accurate subtitles for video content
• ✓Processing sensitive client interviews where data privacy is mandatory
• ✓Translating foreign language voice notes during development

When not to use it

• ✕High-throughput production apps requiring massive parallel processing
• ✕Low-power edge devices lacking sufficient RAM or GPU for model loading

How to invoke it

Example prompts that trigger this skill:

• “Transcribe this meeting recording to a text file.”
• “Convert the audio file at path/to/audio.mp3 into an SRT subtitle file.”
• “Translate this Spanish voice memo into English text.”
• “Run the whisper medium model on my latest audio file.”
• “Process all audio files in the current directory and save outputs.”

Example workflow

1. Store the raw audio file in your local project folder.
2. Invoke the whisper command pointing to the audio file.
3. Select the appropriate model size based on desired speed versus accuracy.
4. Define the output format as either plain text or SRT.
5. Review the generated transcription file created in your working directory.

Prerequisites

• –Python 3.8+
• –FFmpeg installed on the system path
• –Sufficient local storage for model cache (~2GB+)

Pitfalls & limitations

• !Large models significantly increase RAM usage during initial execution
• !First-time runs require a one-time download of the model weights
• !Accuracy varies heavily depending on background noise levels

FAQ

How it compares

Unlike cloud APIs that require network connectivity and billing, this local approach keeps your data entirely within your infrastructure and avoids usage costs.

Source & trust

From the source: “# Whisper (CLI) Use `whisper` to transcribe audio locally. Quick start - `whisper /path/audio.mp3 --model medium --output_format txt --output_dir .` - `whisper /path/audio.m4a --task translate --output_format srt` Notes - Models download to `~/.cache/whisper` on first run. - `--model` defaults to `t…”

# Whisper (CLI)

Use `whisper` to transcribe audio locally.

Quick start

- `whisper /path/audio.mp3 --model medium --output_format txt --output_dir .`
- `whisper /path/audio.m4a --task translate --output_format srt`

Notes

- Models download to `~/.cache/whisper` on first run.
- `--model` defaults to `turbo` on this install.
- Use smaller models for speed, larger for accuracy.

Quoted from openclaw/openclaw for reference — see the original for the authoritative, latest version.

Recommended Rules

Recommended Workflows

Recommended MCP Servers