GlitchAI

GlitchAI is a real-time speech-to-speech voice assistant designed to revolutionize your interactions with technology. With its advanced natural language processing capabilities and intuitive voice interface, GlitchAI understands your spoken requests and responds in a clear, human-like voice.

Features

• Speech Recognition : Accurate and fast conversion of spoken words to text.
• Text-to-Speech: Natural and clear speech generation from text.
• Contextual Awareness: Understanding of conversation context for tailored responses.
• Entertainment: Storytelling, jokes, poetry, content - recommendations, trip planning.
• Productivity: Task management and reminders.

Technologies

• Speech-to-Text: OpenAI's Whisper base English model
• Language Model: Gemini 1.5 Flash integrated with LangChain
• Local Deployment: Ngrok
• Text-to-Speech: Edge-TTS
• Programming Language: Python

Installation and Usage

Prerequisites:

• Python 3.10.x
• Required libraries (install using pip install -r requirements.txt)

STEP 1: Clone the repository

git clone https://github.com/nilotpal-basu/GlitchAI.git

STEP 2: Set up environment variables

GOOGLE_GEMINI_KEY: Your Google Gemini API key

NGROK_AUTHTOKEN: Your Ngrok authtoken

STEP 3: Setting up the LLM server in Google Colab

• Open Google Colab: Go to https://colab.google/.
• Upload the file: Upload the llm_server.ipynb file to Google Colab.
• Run the cells: Execute the cells in the notebook, following the instructions provided.
• Obtain Ngrok URL: The notebook will output the Ngrok URL once the tunnel is established.
• Copy that URL and paste it in main.py in place of your_public_ngrok_url.

STEP 4: Run the application

python main.py

STEP 5: Usage

Speak into your microphone to interact with GlicthAI. The assistant will respond verbally and in text.

 

Contributing

Contributions are welcome! Please follow these guidelines:

• Fork the repository.
• Create a new branch for your feature or bug fix.  
• Make your changes and commit them.
• Submit a pull request to the main branch.

License

This project is licensed under the MIT License.