One-click install
npx skills add https://github.com/Probably-Group/Dev-AID --skill speech-to-text-probably-group
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: speech-to-text
Source: https://github.com/Probably-Group/Dev-AID/tree/main/.dev-aid/skills/expert/speech-to-text
Command: npx skills add https://github.com/Probably-Group/Dev-AID --skill speech-to-text-probably-group

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill solves the problem of turning spoken audio into accurate text reliably, while minimizing privacy and security risks during transcription.

Core Features & Use Cases

  • Local Faster Whisper transcription: Convert WAV/MP3/M4A/OGG/FLAC audio to text using production-oriented settings (VAD, decoding parameters).
  • Safety-first validation and resource management: Enforce file validation (format/size/duration), singleton model pooling, and concurrency limits to prevent overloads.
  • Security and privacy protections: Apply transcript handling guidance, sanitize outputs against crafted audio content, and avoid unsafe storage patterns.

Use case examples: Transcribe recorded meetings into searchable notes, implement voice command features in an app, or build a real-time transcription pipeline that only processes speech segments.

Quick Start

Use the speech-to-text skill to transcribe the attached audio file into text with Faster Whisper and VAD-based chunking while following the provided validation and privacy rules.

Frequently Asked Questions about speech-to-text

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

FAQPage Schema
How do I transcribe audio locally without sending data to external APIs?

Local transcription processes WAV, MP3, M4A, OGG, and FLAC audio files into text entirely on your machine. The workflow uses a local Faster Whisper model with VAD-driven chunking to avoid transmitting sensitive audio data over the network.

What is VAD chunking and how does it work for speech recognition?

VAD chunking uses Voice Activity Detection to identify and isolate actual speech segments within an audio stream. By processing only these validated speech segments for transcription, the workflow filters out silence and optimizes batch processing efficiency.

Can I use Faster Whisper for real-time transcription pipelines?

Yes, Faster Whisper supports both real-time and batch-oriented transcription patterns. The workflow applies singleton model pooling with concurrency limits and VAD-driven chunking to manage streaming audio inputs without overloading system resources.

How does local speech-to-text handle language detection for multi-language audio?

Local speech-to-text handles multi-language audio by applying language detection confidence thresholds during the transcription process. This ensures the model accurately identifies the spoken language before converting the validated audio segments into readable text.

What file formats are supported for local audio transcription?

Local audio transcription supports WAV, MP3, M4A, OGG, and FLAC file formats. The workflow enforces strict file validation for format, size, and duration before processing to ensure safety and prevent system overloads during the transcription task.