davery92 / sara

Sara: A Voice Chat AI

Welcome to Sara, an AI-powered voice chat platform that lets you have conversations like never before. With advanced natural language processing (NLP) and machine learning capabilities, Sara is designed to understand and respond to your thoughts and emotions in a way that feels human-like.

Getting Started

To get started with Sara, follow these steps:

Step 1: Install OpenVoice

OpenVoice is an open-source speech recognition library that provides the foundation for Sara's voice processing capabilities. To install OpenVoice, run the following command:

pip install openvoice

Step 2: Download Checkpoints from OpenVoice

After installing OpenVoice, you'll need to download the necessary checkpoints from their GitHub page. These checkpoints are pre-trained models that enable Sara to recognize and respond to your voice input.

1. Visit the OpenVoice GitHub page.
2. Follow the instructions for the V2 option.

Step 3: Download and Install Vosk Model

Vosk is an open-source speech recognition model that Sara uses for voice-to-text conversion. To download and install the vosk model, follow these steps:

1. Visit their page(https://alphacephei.com/vosk/).
2. Follow their installation instructions.
3. Click on the "models" button.
4. Download the file "vosk-model-small-en-us-0.15".

Step 4: Install Faster-Whisper

Faster-Whisper is an open-source library that enables Sara to recognize and respond to voice input in real-time. To install Faster-Whisper, run the following command:

pip install faster-whisper

Step 5: Create a Python 3.9 Virtual Environment

Sara is built on top of Python 3.9, so we need to create a virtual environment for our project.

python3.9 -m venv env

Step 6: Install the Requirements

pip install -r requirements.txt

Setting Up Sara

Make sure to have the following folders in the root directory:

1. checkpoints_v2
2. docs
3. env
4. openvoice
5. outputs
6. processed
7. resources
8. vosk-model-small-en-us-0.15

Running Sara

To run Sara, simply run

python speak.py

Using Your Wake Word

At this point, feel free to say the demo wake word "sarah" (or whatever wake word you've chosen) to initiate the conversation. To select your own wake word, simply modify the speaking.py file where it says response_word = "sarah" and enter your preferred keyword.

How It Works

Once Sara recognizes your wake word, she'll respond with "Hotword detected! Recording..." and wait for your input. Speak freely until you're finished, then automatically, Sara will process the text and generate a response using the LLaMA model (which can be changed within the speaking.py file). The generated response is then passed to our TTS program, which plays it back once generated.

Conversation Loop

The conversation loop continues as follows: after the response is played, Sara will display the word "speak" to indicate she's listening again. You'll have 2 seconds to speak before the system loops back and displays the word "Listening..." again. If you see this message, please say your wake word once more to continue the conversation.

Tips for Seamless Conversations

To ensure a seamless conversation experience:

• Make sure to say your wake word within the 2-second time limit.
• Speak clearly and at a comfortable pace.
• Feel free to ask Sara questions or share your thoughts – she's designed to respond thoughtfully and engage in meaningful conversations!

I hope you enjoy using Sara, and I'm always looking for ways to improve her capabilities. Happy chatting!

Future items to be added

1. Function calling
2. Memory Database
3. Homeassistant integration
4. Note taking
5. Timer functionality
6. Cell Phone Push Alerts