December 2025 - Release notes

The last update of the year, and an important one for EstimAI, as the stimulation handling has been completely reworked.

🎁 This feels like a good moment to thank you all for the support over the past year. I'm really grateful, so I wanted to give a little something back: a 15% discount on the subscription for one month. 🎁

This release contains:
- 15% Discount!
- Multi-Device
- zImage Support
- EstimAI saved sessions

As always, feel free to reach out (contact@daimonia.app) if you run into any weird behavior, bugs, or just want to share feedback.

15% Discount

It’s with great pleasure that I’d like to thank you and celebrate the end of the year by offering a 15% discount on your subscription. More details here.

If you’re already subscribed, your next automatic renewal will be 15% cheaper.
If you’re not subscribed yet, your first month will cost 15% less.
This discount applies only once (for a single month) and is valid from 2025/12/23 to 2026/01/22 (inclusive).

Multi-Device (EstimAI)

Stimulation handling has been reworked to allow users to use multiple stimulation devices at the same time. This means you can now combine a 2B and a Coyote 3 to stimulate up to four different areas simultaneously (see example illustration below).
AI now understands and controls your specific setup: a detailed description of your configuration is automatically added to the prompt. You can create scenarios tailored to your setup, and they should work well.
Each area is now stimulated independently, unlike before. The AI can increase the intensity on one area while sending a painful stimulation to another at the same time for example.
You can also configure whether an area can be subjected to pain, or mark it as an exclusive pain area.

Example:

Limitations: While you can create complex setups, adding more areas may increase the risk of confusing the model. I’ve had good results with up to four different areas using models like DeepSeek, GLM 4.6, Gemini, etc.
Smaller models, however, can struggle to stay consistent with more complex configurations, so you may want to reduce the number of areas in those cases.
You can also switch back to the legacy stimulation handling if this setup doesn’t work well for you.

In the next update: I intend to add the following features:

Support for Sex toys (allowing to mix E-stim and classic sex toys togeter)
An intelligent E-stim feature. I’ve been working on it for a while, and the results haven’t been great so far — but I’m hopeful I can get it right.

Z-image Turbo Support

This month, a new and groundbreaking base model was released: Z-Image Turbo. In my opinion, this is an important release, as the model is open source, fast, and lightweight enough to run on your own GPU (with at least 12 GB of VRAM).
It's a realistic model with an impressive prompt adherence (I had really good results with the 'Intelligent Image genereation feature'). It's also uncensored (for the most parts).

You can find help on how to use it here.

Limitations: The model is uncensored in the sense that it supports full nudity (except for penises) and has not been poisonned with censorship afterwards. However, it cannot generate images depicting explicit sexual acts.
There are currently very few LoRAs available, so you’ll likely need to use it without a character LoRA. This means character consistency may be limited.
Style consistency is also a bit shaky. The model performs best with realistic styles, but maintaining a semi-realistic or anime style becomes difficult when prompts get more complex or overloaded.

The good news is that many fine-tuned models and LoRAs will probably be released over time to address these issues. There are already quite a few interesting ones available on Civitai.

EstimAI saved sessions

A smaller but important feature: you can now save multiple sessions and resume them later.
Sessions are stored online (on Eudaimonia’s servers), so this feature is disabled by default.
If you choose to enable it:

Your last 10 sessions are automatically saved.
You can permanently store up to 30 sessions.
You can easily resume a saved session.
You can fully edit a session.

This is useful if you want to play a scenario across multiple days or need to resume a session.
It’s also interesting because you can now edit sessions to replay a scenario in a different way, guide the model in a specific direction, or fix issues like repetitions, etc.
Additionally, this can be handy for scenarios that require a "setup." Normally, you’d have to redo the full introduction each time, which can take a bit of time. Now, you can start from a saved point and choose not to overwrite it, letting you skip the initial setup messages.

Miscellaneous:

Lots of different things of different importance:

EstimAI Debug / Textual mode has been improved a bit. You can now use the "intelligent image generation" feature with it, it will generate one image by message.
HTML Support in EstimAI Debug Mode.
Improved in-game history management. You can now edit the scenario / persona / OOC message.
Image generation: Mediums (styles) are now linked to global settings instead of character.
Intelligent image generation: few improvements, especially regarding speed. If you're using Z-image model, then it uses natural language.
Serial Link 2B: Pain stims intensity has been changed with an higher base value. This means that a 100% intensity pain stim will be the same, but a 70% will be higher.
In the state management part of the scenario editor, you can now toggle if a property is visible to the user or not.
I've added empty system prompts. This can be useful if you want to create a complete system prompt.
In EstimAI, you can now hide the answer module to see the full image.
MCQ mode will now display the different paths in a random order. Ai tend to order them, which makes some scenarios easier as the right answer is usually the first.
You can now use "Shift + enter" or the Num pad while typing a message.
The denial option was not working properly in normal mode. This is now fixed.