Podcast Translator Project

This guide explains how to set up and run the fully open-source, locally-run MediaTranslator application which requires no cloud services.

Prerequisites

Required Software

• Java JDK 21 or higher
• Maven for dependency management
• FFmpeg (for audio extraction and processing)
• Ollama (for LLM capabilities) - https://ollama.com

Speech-to-Text Option:

• OpenAI Whisper locally installed:
  • Install whisper.cpp:
    • Clone from https://github.com/ggerganov/whisper.cpp
    • In the directory, run make
    • Download a model, e.g. sh ./models/download-ggml-model.sh large-v3-turbo-q8_0

Choose One Text-to-Speech Option:

• eSpeak: sudo apt-get install espeak (Linux)
• Pico2Wave: sudo apt-get install libttspico-utils (Linux)
• Mac's built-in 'say' command (MacOS only)
• Python pyttsx3: pip install pyttsx3

Setting up Ollama

1. Install Ollama from https://ollama.com
2. Pull a model for translation and text enhancement:

   ollama pull mistral    # Basic, fast model
   ollama pull llama3     # Alternative option
   ollama pull mixtral    # More capable for complex tasks
   

3. Ensure the Ollama service is running before using the application
4. Check the model is available:

   ollama list
   

Building and Running the Application

mvn clean package
java -jar target/media-translator-1.0-SNAPSHOT.jar

Usage Instructions

1. Create a settings file settings.properties, based on src/main/resources/settings-example.properties
2. Run the application
3. The application will process the file and generate the translated audio

Troubleshooting

Speech Recognition Issues

• Make sure either whisper.cpp is properly installed

Text-to-Speech Issues

• The application will try multiple TTS options in order
• If using pyttsx3, ensure you have compatible voices installed for your target language

Ollama Connection Issues

• Check that the Ollama service is running on http://localhost:11434
• Verify you have pulled the model you're trying to use

Supported Languages

The application supports translation between any language that Ollama models can handle. Speech synthesis capabilities may vary depending on your chosen TTS system, but commonly supported languages include:

• English (en)
• Spanish (es)
• French (fr)
• German (de)
• Italian (it)
• Portuguese (pt)
• Russian (ru)
• Chinese (zh)
• Japanese (ja)

To add more language support for Text-to-Speech, install additional language packs for your chosen TTS system.