TowerAI v2 by yAI

LLaMA3.2-based TowerAI v2 3B for better RC plane simulation. It acts like a tower, and with the scripts, you can speak and talk to TowerAI. The script adds radio effects to the output sound file, so it feels like you are really talking to a tower.

There’s also a Google Gemini 1.5 Flash API script, but TowerAI can work locally on an average laptop, making it portable because you probably don’t have WiFi access in the place where you are flying the RC plane, right?

Also, it’s at least 37% faster than using the Gemini API.

Here is the GitHub page.

Here is the HuggingFace page.

Here is the YouTube video.

Here is the yAI website.

What Scripts Do? TowerAI script converts Speech-To-Text first. Then it sends the result to the chosen AI model (Gemini API or TowerAI) as a prompt. After that, it converts the response into a regular sound file (as mp3). Then it adds a radio effect to it to improve the experience. So it feels like you are really talking with the tower.

Installation

1. Clone the repository:

   
   git clone https://github.com/MYusufY/tower-ai-v2.git
   cd tower-ai-v2
   

2. Install the requirements

   pip install -r requirements.txt
   

3. (OPTIONAL) Switch to “myenv” which has all the required libraries and setup

You can install the venv to directly have all required libraries installed instead using requirements.txt.

   source myenv/bin/activate

4.Run the AI/API script

TowerAI custom model:

   python3 tower-ai-custom.py

Google Gemini 1.5 Flash API:

   python3 tower-ai-api.py

5.Set API key (if you are using Gemini) To get an API key, you can follow the steps of my LVGL Gemini project.