Linuxでライブオーディオ転写にささやきAIを使用する方法

This guide details how to install and use Whisper AI for real-time speech-to-text transcription on Linux systems. Whisper AI, an OpenAI creation, offers high-accuracy transcription across multiple languages. While primarily designed for batch processing, this guide demonstrates real-time capabilities.

Understanding Whisper AI

Whisper AI is an open-source, deep-learning based model excelling at:

• Multilingual speech transcription.
• Handling accents and background noise.
• Translating spoken language to English.

Its applications include live transcription services, voice assistants, and audio file transcription. While not inherently optimized for real-time use, it can be adapted with additional tools.

System Requirements

Before proceeding, ensure your Linux system meets these requirements:

Hardware:

• Multi-core CPU (Intel/AMD)
• At least 8GB RAM (16GB+ recommended)
• NVIDIA GPU with CUDA (optional, but significantly improves speed)
• 10GB+ free disk space

Software:

• Linux distribution (Ubuntu, Debian, Arch, Fedora, etc.)
• Python 3.8 or later
• Pip package manager
• FFmpeg

Installation Steps

1. Update System Packages:

   sudo apt update     # Ubuntu
   sudo dnf update -y  # Fedora
   sudo pacman -Syu    # Arch

2. Install Python and Pip:

   sudo apt install python3 python3-pip python3-venv -y     # Ubuntu
   sudo dnf install python3 python3-pip python3-virtualenv -y # Fedora
   sudo pacman -S python python-pip python-virtualenv       # Arch

3. Install FFmpeg:

   sudo apt install ffmpeg     # Ubuntu
   sudo dnf install ffmpeg     # Fedora
   sudo pacman -S ffmpeg       # Arch

4. Install Whisper AI: Create a virtual environment and install the package:

   python3 -m venv whisper_env
   source whisper_env/bin/activate
   pip install openai-whisper

   Verify installation with: whisper --help

   

5. Transcribing an Audio File: Use the command whisper audio.mp3 to transcribe an audio file.

   

Real-time Transcription Script

To enable real-time transcription, create a Python script (real_time_transcription.py) with the following code:

import sounddevice as sd
import numpy as np
import whisper
import queue
import threading

# ... (Code as provided in the original input) ...

Run the script using python3 real_time_transcription.py. Speak clearly into your microphone; the transcription will appear in the terminal.

Conclusion

This guide provides a comprehensive walkthrough of setting up Whisper AI for real-time transcription on Linux. Remember that a GPU will significantly enhance performance. Optimize your system for real-time processing for the best results.

以上がLinuxでライブオーディオ転写にささやきAIを使用する方法の詳細内容です。詳細については、PHP 中国語 Web サイトの他の関連記事を参照してください。