seandadonntech/ai-vtuber

AI-Vtuber

This code is designed to read chat messages from YouTube and then utilize OpenAI's GPT-3 language model to generate responses. The output from GPT-3 is then read out loud using a TTS (Text-to-Speech) engine provided by ElevenLabs.

Setup

Install dependencies

git clone https://github.com/Koischizo/AI-Vtuber/
cd AI-Vtuber
pip install pytchat openai pydub pyttsx3 simpleaudio

It also requires ffmpeg to be installed

Usage

Edit the variables video_id, EL_key, EL_voice and OAI_key in run.py

video_id is the ID of the Youtube stream found in the Youtube link

EL_key is the API key for ElevenLabs. Found in Profile Settings

EL_voice is the voice ID for ElevenLabs. Found in the docs in Get Voices

OAI_key is the API key for OpenAI. Found here

Then run run.py

python run.py

then you're set

Live Demo

Livestream 1

Livestream 2

Other

I used This VTS plugin and VB Audio cable to make her mouth move and be able to play music at the same time

Please note that this project was created solely for fun and as part of a YouTube video, so the quality and reliability of the code may be questionable. Also, after the completion of the project checklist, there won't be much activity in updating or improving this repository. Nonetheless, we hope that this project can serve as a source of inspiration for anyone interested in building their own AI Vtuber.

• Clean up
• GUI
• Executables (exe, bat or sh)
• Extra features (maybe) (Prompt injection protection, questions only mode, virtual audio)

License

This program is under the MIT license