- ai_pngtuber_ollama
- Table of content
- Video example HERE
- What is Ai PngTuber Ollama?
- Do i need to pay something?
- How to install Ollama and what model?
- How to run Ollama as server?
- How to setup this app?
- How to close the app?
- How to change the avatar of the PngTuber?
- How to have substitles of the PngTuber's speech on my stream?
- How to change the lore of the PngTuber?
- Mediagraphy
Video example HERE
This app is made from Python scripts to generate a PngTuber that use Ollama system and api has AI language model. In another words, that use a locale, free and private ChatGpt-like in your network as Ai of this PngTuber.
This app is made to offer a full experience free and customizable. But if you want a better voice than the basic Google one. You need to use the api named ElevensLabs. P.S.: There is a free plan with ElevenLabs, but you will maybe touch the limit of characters per month of it with the free one.
This app use as default the model of llama2 who is easily downlodable by Ollama.
To download and install Ollama there is a great video to explain how Ollama is used for and how to install it. Video Here P.S.: You don't need necessary the fine tuning or custom models. Just the first part of the video explain the necessary.
It's important for this app to be able to communicate with Ollama. For that you need to tell to Ollama to run as a local server.
Terminal command to run Ollama as server:
OLLAMA_ORIGINS=* OLLAMA_HOST=127.0.0.1:11435 ollama serveFirst, rename all the files in the ./env folder by removing the .example of them. For the custom settings, it's in them that will be done. And all the explaination of each setting is in these files. But, the default settings will make it work.
Second, install python if it's not already done. The recommended version is Python 3.12.2. Also, be sure to install pip too when you install Python.
Third, install all the dependencies with pip. You can use ./setup/setup.bat (Windows) or ./setup/setup.sh (Mac/Linux) to install all the dependencies.
Fourth, run the app.py (Linux/Windows/Mac) script in a terminal or use the app.exe (Windows Only) in the release section.
P.S.: There is for now only user's messages by terminal input, but I will try to add a Twitch integration in the future.
You can close the app in 2 ways:
- Doing CTR+C in the terminal window.
- CLicking on the X bouton the the Avatar window.
The 2 images of the avatar is in the ./res/img folder. I recommend to use 480x480 images. The app doesn't work with gif files. If you need to change the path of the imgs for the software, you have to edit the ./env/avatar.env.
P.S.: I recommend also to colorize the background of your images to be able to make it transparent in OBS.
P.S.2: The default avatar if from Avataaars, so it's not recommend to use it for stream or recording.
There is ./res/content/subs.txt that you can use as file source for the text to show in OBS.
You change the lore in the ./res/content/lore.txt. I suggest to write the lore in english because the models of Ollama are generally build in english first. So there will be better responds from the chatbot.
That is also why there is a translator in the app.