Whisper-Transcription

Transcribe audio files to text via a local whisper.cpp server.

Updated Dec 22, 2025
One-click install
npx skills add https://github.com/lawless-m/Gwen --skill whisper-transcription
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: Whisper-Transcription
Source: https://github.com/lawless-m/Gwen/tree/main/.claude/skills/Whisper-Transcription
Command: npx skills add https://github.com/lawless-m/Gwen --skill whisper-transcription

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill enables local speech-to-text transcription using whisper.cpp with GPU acceleration, avoiding cloud services and preserving privacy.

Core Features & Use Cases

  • Local whisper.cpp server on port 5555
  • GPU acceleration for faster transcription
  • HTTP API support for integration with scripts and apps

Quick Start

Install whisper.cpp, run the whisper-server with a large model, and POST audio files to /inference on the local server.

Frequently Asked Questions about Whisper-Transcription

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I transcribe audio files locally without using cloud services?

Local audio transcription using whisper.cpp with GPU acceleration preserves privacy by running speech-to-text on your machine. Set up a whisper-server on port 5555 with the ggml-large-v3 model, then POST audio files to the /inference endpoint to receive JSON responses with transcribed text.

Can I transcribe MP3 and WAV files with GPU acceleration?

Yes, whisper.cpp transcription supports WAV, MP3, and other common audio formats with CUDA GPU acceleration. Optimize input as 16kHz mono WAV for best performance, POST to the local HTTP server, and receive text output in JSON format.

What do I need to set up before transcribing audio with whisper.cpp?

Prerequisites include a CUDA-enabled GPU, whisper.cpp installed, the ggml-large-v3 model downloaded, and a running whisper-server listening on port 5555. Audio input should be prepared as 16kHz mono WAV files for optimal transcription results.

How do I integrate audio transcription into Python scripts or shell workflows?

The local whisper-server exposes an HTTP API at /inference accepting POST requests, enabling integration with Python, shell utilities, and other applications. Send audio files to the endpoint and parse the returned JSON object containing the transcribed text field.

What are the limitations of local GPU-accelerated transcription versus cloud services?

Local transcription requires upfront GPU hardware investment and model setup but eliminates cloud dependency, latency, and privacy concerns. Performance depends on GPU capability; the ggml-large-v3 model provides accuracy at the cost of resource consumption.