>hmm it sounds a bit too much for a single program though I cannot tell what kind features you are going for that needs so much RAM.
Well, the renderer itself takes about 500 MB RAM and like Em Elle said, the speech synthesis module takes about another 500 MB RAM, and then you have the AI model that takes about 5-6 GB of RAM.
Now since this is a fully offline
architecture, everything is run locally
on your system directly. We are not using any online based server to process all the AI stuff (very much unlike
how other assistants like Siri, Cortana or Google Assistant), since we value your privacy first and foremost
Now, it's also possible to run an AI without loading the model to the RAM, but then you would have to wait about 20 s for each response. In order to make it run fast and respond to you in real time, we require to keep it loaded in the memory, so that you get an immersive experience just like talking to a real person.
So the app itself would take at most about 8 GB RAM to operate (or maybe slightly less than that) and that is the minimum requirement. 16 GB of RAM is what we recommend though, since you will want to keep her on your desktop wallpaper, and you would want to keep doing other regular stuff at the same time, which is why the extra RAM. Hope that makes sense. :)
We however plan to make the AI models smaller and more compact (so that they take less RAM) by distilling and retraining them once we have more resources to do so, cause at present we have been working on this with a zero budget and a very small team of volunteers only.
>Because DirectX is available only for Windows OS and I'm running GNU/Linux.
We plan to support Linux in the future, but it will not be able to have a wallpaper mode. Linux support is currently low priority in our development roadmap, due to far more people using Windows, which is why we want to focus on getting this up for Windows first, which I hope, is understandable.
I hope that clarifies the doubts that were there. :)