Real-Time Speech-to-Speech Chatbot Based on Ollama & Kokoro - Install Locally

Real-Time Speech-to-Speech Chatbot Based on Ollama & Kokoro - Install Locally

---

---

AI Summary

Overview

• Project: Real-time speech-to-speech chatbot
• Integration of:
  • Speech recognition
  • AI-driven reasoning
  • Web information access
  • Natural sounding voice output
• Powered by Whisper, Llama 3.1, Kukuro, and Eggno agent framework

Installation Steps

1. Olama Installation
   • Visit ola.com and download.
   • For Linux, run command in terminal.
   • For Mac/Windows, download executable.
2. Create Virtual Environment
   • Use Conda to create a virtual environment.
3. Clone Project Repository
   • Clone the Vocal Agent repository.
4. Install Dependencies
   • Run pip install -r requirements.txt from the repo’s root.
5. Text-to-Speech Synthesis
   • Use Kukuro models available in the repository.
6. Download Olama Model
   • Run lama list to check for models.
   • Download Llama 3.18 billion model.

Running the Tool

• Execute the command python3 main.py from the project root.
• First run may require additional model downloads.
• Handle input format issues (ensure input is a valid list).

Observations

• Initial errors related to input format; require conversion of string to list.
• Functionality confirmed to work after code adjustments; improvements needed in interface design.

Conclusion

• Project works but has bugs that need addressing.
• Suggestion: Develop a cleaner interface, possibly using Gradio.
• Repository link to be provided in video description for access.