Overview

Transcribes audio locally using whisper-cli and the whisper.cpp large-v3-turbo model. This setup keeps data on your machine, reducing latency and improving privacy for transcription tasks.

How This Skill Works

Use either the wrapper script scripts/whisper-local.sh <audio-file> or call the whisper-cli binary directly. The process loads the model from /usr/share/whisper.cpp-model-large-v3-turbo/ggml-large-v3-turbo.bin and runs with -f <file> -l auto -nt; GPU acceleration can be enabled via whisper-cli for faster results.

When to Use It

• Offline transcription of audio without sending data to cloud services
• Privacy-sensitive projects where data must stay on-device
• Transcribing single audio files quickly with GPU acceleration
• Building local transcription workflows or batch processing pipelines
• Using whisper.cpp large-v3-turbo for higher accuracy on local hardware

Quick Start

1. Step 1: Download the model to /usr/share/whisper.cpp-model-large-v3-turbo/ggml-large-v3-turbo.bin using the provided wget command
2. Step 2: Run the local transcription, e.g., scripts/whisper-local.sh <audio-file> or whisper-cli -m /usr/share/whisper.cpp-model-large-v3-turbo/ggml-large-v3-turbo.bin -f <file> -l auto -nt
3. Step 3: Retrieve the transcript from the output path and validate accuracy

Best Practices

• Download and store the model once at /usr/share/whisper.cpp-model-large-v3-turbo/ggml-large-v3-turbo.bin and reuse it
• Prefer the wrapper script scripts/whisper-local.sh for simple usage, or replicate the whisper-cli -m ... -f ... -l auto -nt invocation in automation
• Enable GPU if available to speed up transcription via whisper-cli
• Test with representative audio to verify language detection and accuracy (-l auto)
• Log commands and outputs to support reproducibility and auditing

Example Use Cases

• Transcribing a team meeting recorded locally on a workstation
• Converting an in-person interview to text for analysis without uploading audio
• Generating podcast transcripts offline to publish alongside episodes
• Archiving lecture recordings for accessibility in education
• Automating captions for local video projects during post-production

Frequently Asked Questions