This is a big update with a lot a new features and content, I hope you'll like it.
As always, please feel free to contact me (contact@daimonia.app) if you encounter any strange behavior, bugs, or want to share feedback.
And of course thanks to all of you who have supported the development of eudaimonia in any way!
Special mention to a community member who sent me his great TTS integration, saving me a lot of time, which you can use locally!
Eudaimonia's normal session now features a new interactive mode. Your Mistress can now give you simple in-game tasks, which I believe makes the session more interesting and interactive.
Using the Speech Recognition feature with this mode adds another layer of immersion.
In this mode, you now have a behavior score that will be modified by your actions during the session, leading to consequences...
I still have a lot of ideas and scripts that I couldn't add in time, but will do so in the coming months.
To create an even more immersive experience for both normal mode (with interactions) and EstimAI (prompts and interruptions), a new feature is available in the subscription:Speech Recognition.
This feature can be used when your input is required, either by holding down the microphone icon or by holding down the shortcut key 'ctrl + e'.
All subscribers benefit from this feature in their current subscription. All informations about usage here.
A community user wrote a great Python script for synthesizing voices with eudaimonia. It's simple, lightweight, fast, and works locally! I adapted it slightly and added support for it in eudaimonia.
You can find all informations to use it here.
The installation process may not be the easiest for non-developers, so I'll likely make a video tutorial later.
I plan to add more voices and models later, as well as voice cloning. I'll also try to simplify the installation process.
Lots of improvements and especially support for another realistic model: BigLove.
Make sure to to check the model's page on civitai on how to use it. Default settings are: CFG 1-1.5, steps: 8, sampler: LCM, scheduler: exponential.
I have tried the pony2 version of BigLove several times with good results, the SDXL version also looks very good.
This model produces high quality realistic photos (very NSFW), while being fast and supporting sdxl and pony Loras.
A new feature: Intelligent image sync. It tries to add relevant images to the AI's output. It works with image generation and intelligent images sets, as they are tagged more precisely than the other.
This slightly increases generation time because two requests are performed (one to get the result, a second to add images).
Also, two failed experiments: I've tried to work on an MCQ auto-play mode, but it wasn't good. I've also tried to improve the way estim works in EstimAI, but didn't see any real improvements while it made everything more complicated. I'm still working on this.
And many different improvements, the biggest one is probably the possibility to edit all stims now.