Guide · Agent skill
How to use the Local Whisper Transcription agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Local Whisper Transcription skill (skill name openai-whisper) teaches an AI agent to run OpenAI's Whisper speech-recognition program on your own computer and turn audio files into text. It is useful for interviews, lectures, and video subtitles when you want the audio processed locally. After this guide you will know what the skill does, how to install it, how to ask for transcriptions in plain language, and what to check before using it on sensitive recordings.
Local Whisper Transcriptionspinabot/brigade · MIT Download the skill Quick start
- 1Install whisper. Run brew install openai-whisper so the whisper command is available.
- 2Download the skill ZIP. Get the ZIP from this page and unzip it to get the openai-whisper folder with SKILL.md.
- 3Place the skill folder. Copy it into ~/.claude/skills/ or a project's .claude/skills/, or upload the ZIP under Settings → Capabilities → Skills.
- 4Ask in plain language. Name the audio file and the output you want, such as text, SRT, or English translation.
- 5Wait for the first model download. The first run downloads a model to ~/.cache/whisper, so allow time and network access.
- 6Check the output. Look in the output folder for the transcript or subtitle file and confirm the result.
What is an agent skill and what does this one do?
An agent skill is a folder containing a SKILL.md file that gives an AI agent instructions for a specific task. When the skill is installed, the agent can use it when your request matches the skill's description.
This skill makes the agent use the whisper command-line tool to transcribe audio locally. According to the skill, it can produce text or subtitle files from audio, and it can translate speech into English. The skill's description is: local speech-to-text with the Whisper CLI, with no API key.
Install the Whisper command-line tool first
The skill cannot transcribe without the whisper command-line tool. The skill's install option is Homebrew, using the formula openai-whisper, which provides the whisper binary.
Install the tool before you install the skill, or the agent will have nothing to run.
- Required binary: whisper
- Homebrew formula: openai-whisper
- Install option listed by the skill: brew
brew install openai-whisperInstall the skill from the ZIP
Download the ZIP for this skill from this page and unzip it. The skill is delivered as a folder that contains SKILL.md. Place that folder in one of these locations:
- Personal: ~/.claude/skills/
- Single project: the .claude/skills/ folder inside your project
- Claude apps: upload the ZIP under Settings → Capabilities → Skills
Trigger the skill with natural requests
You do not need to type a command. Describe the job in plain language and mention the audio file and what you want back, such as text, an SRT subtitle file, or an English translation. The agent uses the skill when the request fits its description.
Give the full path to the audio file when you can, since the skill's examples use full paths.
- Transcribe /Users/me/interviews/interview1.mp3 to a text file in this folder.
- Make an SRT subtitle file from lecture.m4a.
- Translate the speech in this Spanish recording into English.
Understand the commands the skill uses
The skill's quick-start examples show two patterns. The first transcribes to a text file in the current directory with the medium model. The second translates speech to English and writes SRT subtitles. The skill notes that smaller models are faster and larger models are more accurate, and that --model defaults to turbo on the install it describes.
- --model medium: a chosen model size
- --output_format txt or srt: the output type
- --output_dir .: write output to the current folder
- --task translate: translate speech into English
whisper /path/audio.mp3 --model medium --output_format txt --output_dir .
whisper /path/audio.m4a --task translate --output_format srtPlan for the first run and disk space
Whisper models are downloaded on first run and stored in ~/.cache/whisper, so the first transcription needs network access and free disk space. No API key is required, according to the skill.
- Models download to ~/.cache/whisper on first run
- Larger models are generally slower but more accurate, per the skill's notes
Safety notes before you use it
This skill runs scripts and uses the network. The network use is at least the model download on first run. The skill does not explain every file the tool may create, so check the output folder after a run. For confidential recordings, confirm where the tool's downloads and outputs go before you rely on it.
- Runs scripts: yes
- Uses network: yes
- Review the output directory and model cache before processing sensitive audio
What people use it for
- Transcribing interview recordings into text for notes or quoting
- Turning lecture audio into searchable text notes
- Generating SRT subtitle files for videos
- Translating non-English speech into English text
Frequently asked questions
How do I install the openai-whisper skill?
Download the ZIP from this page and unzip it. Put the folder in ~/.claude/skills/ or a project's .claude/skills/ folder, or upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
Do I need an API key?
The skill says no API key is needed because it runs Whisper locally. The first run downloads a model to your machine, which requires network access.
Is the Whisper skill free?
The skill is published under the MIT licence. Pricing details for other services are not covered in the source, so check those separately.
Why does the skill fail to transcribe?
The most likely cause is that the whisper command-line tool is not installed. The skill lists the Homebrew formula openai-whisper as its install option, so install that first.
Where are Whisper models stored?
Models download to ~/.cache/whisper on first run. Make sure you have free disk space there.
Can it transcribe live microphone audio?
No. The skill only covers processing existing audio files, not real-time or live transcription.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Choose this skill if you already work with audio files and want transcripts, subtitles or English translations generated on your own machine without an API key. Pick something else if you need live microphone transcription, cannot install the whisper command-line tool, or need full visibility into every file created with highly sensitive recordings.
Download the skill