BuddyGenAI (demo)
A downloadable game
Demo Release -- Next release expected July-August
This is a demo where you chat with AI virtual buddies.
No internet connection is needed thanks to Llama.cpp, Stable-Diffusion.cpp, Piper TTS and Whisper.cpp
- Customize your buddies' personality.
- Generate profile pictures for your buddy using dynamically-generated appearance options.
- Use Speech-to-Text to speak instead of typing your message.
- Text-to-Speech lets buddies speak back to you.
- Buddies can send pictures in the chat.
Source code is available on GitHub
Discord server if you need help (unfinished)
Feedback is appreciated and welcome!
I'm not sure when I'll have the next update out -- this project's taken a lot of my time the past couple months & I need to catch up on things.
Let me know in the comments/discord/anywhere else issues you'd like me to prioritize.
Compatibility
- This requires Windows 10 or higher.
- Currently NVIDIA GPUs only. Developed on an RTX 3060, also tested to work on GTX 1650 TI
Troubleshooting / Known Issues
- This is an early demo -- be prepared to lose messages (tip: right-click messages to Copy or Edit them).
- Most issues can be fixed by restarting the game.
- There are issues navigating between pages (e.g. viewing 1 buddy and then a different one). Pressing Ctrl+R usually fixes this.
- Pay attention to the "Chat Offline/Online" status in the top of the Chat tab. If Chat is Offline then you won't be able to use most of the game. You can click the status to get manual controls, or try restarting the game.
- Chat Images are disabled by default (turn on in Settings -> Image AI Options -> Enable chat images).
- Settings get reset sometimes.
- Reloading the last response can be buggy.
- Lots of areas need improvement.
Download
Install instructions
Getting Started
Setup
You'll need Microsoft Visual ++ Redistributable 2015-2022
First, download and install BuddyGenAI-Setup-0.2.3.exe (expect a warning from Windows SmartScreen, hope to fix soon).
Next, you'll need models to run the AI. A zip file with the models you need is provided for convenience -- for this, download the AI Model Pack zip (if you already have the needed models, skip to "Import Models" below and refer to the section "Bring Your Own Models").
Once the file is done, find it in the File Explorer, right click and Extract All. You can delete the zip file.
Import Models
Open BuddyGenAI, choose to Import Models -> Choose File and then navigate to the models you extracted. Select all of the files in the folder and click Open. Finally, click Move. You can now select each of these models during setup or through Settings ("Chat AI Options", "Image AI Options", etc.)
Make a Buddy
After setting a name for yourself, give your buddy a name and some optional characteristics. Then you set their appearance using dropdown menus and generate profile picture(s) for them (pictures are saved and accessible by going to Edit that buddy).
Bring Your Own Models
The game has the ability to import models to customize your experience further.
The following is a brief explanation of the models + their file types that the game expects, as well as links to the models included in the AI Model Pack.
- LLM/Chat models should be .gguf
- Stable Diffusion/Image models should be .safetensors (see here for SD version support)
- Piper TTS Voices should be a .onnx file with an adjacent .json file with the name of the first file (including .onnx)
- Whisper/Speech-to-text models should be .bin
Development log
- Demo Launch-ish16 days ago
Comments
Log in with itch.io to leave a comment.
Does this app not run on Windows 7? I seem to be having some windows 7 issues with launching the installed package.
I believe Windows 10 and up is needed to run.
Sorry about that, I'll note this in the description.
I'll research and try to support 7 in the future.