SentientSims
patreon
3.13.0 Lots of Little Updates and Whats Coming Soon
🕑 Added 2025-06-18 20:51:56 +0000 UTCHello, this post includes many of the updates and changes I have made since the last update.
The mod is about using AI, but I write these posts without AI. If anything is written by AI, I will explicitly state it. This way you know what is AI and what is not.
Buff/Moodlet Updates and Fixes
I made a few fixes to the buffs you receive after chats. They should work a lot better for Sentient Sims and OpenAI.
Buffs after chats now run in the background. Previously it would lock up the game while it was running and freeze. Now it runs in the background and the game runs normally while the AI is thinking.
Mac OS Universal Build
Mac players now use a “Universal” build. This means that the sentient sims app now includes all the files for both Intel and the newer Macs, so there is only one version for everyone.
I have requested a Mac developer license. Once I receive it, the installer will be trusted by Apple. This will allow auto updates on Mac! It will also not try to quarantine the install when you try to install Sentient Sims.
Let me know if you run into any Mac issues in Discord.
Minor Fixes and Updates
Continue should now work a little better and actually keep continuing, especially if you use the Mythomax model.
Switching locations sometimes makes the AI think the sims are at the previous location. I put in a fix that prefixes each prompt with the location of the sims. I have found this has greatly improved the ability for the AI to get the location correct. Let me know if you have noticed it being better about this.
Fixed a bug in Kokoro TTS where “Isabella” voice didn’t work. This voice is now removed from the options.
Scout and Maverick AI models are currently in test on Sentient Sims AI. Try these models out and let me know what you think of them in Discord!
The memories UI in game was slightly tweaked so that there is a newline in between messages. This makes it so the whole thing doesn’t look like one giant block of text.
VLLM support has been added to the Sentient Sims App. VLLM is an AI inference engine. If you want to host your own AI to play with Sentient Sims, now you can point the app at your own VLLM instance running Mythomax or RP-max and play without sending your data anywhere. This requires GPUs in order to run which could be hosted locally or rented from somewhere like runpod.io.
Coming Soon
Coming soon, your sims will “reflect” after a good nights sleep and receive a buff based on how their previous day went, just like chats. The AI will take all of their conversations from the previous day and give a moodlet based on how they thought the day went. Your Sim will wake up with a boost based on the previous day.
Orpheus voice TTS will be coming soon as another option to Sentient Sims AI in addition to Kokoro. If you aren’t familiar with Orpheus TTS check out this sample.
Eleven labs TTS will also be coming soon as a third party option to use for TTS. This can be used in conjunction with any other AI options.
I am working on fine tuning an AI model to the prompt format of Sentient Sims. This is ongoing and very slow work. I don’t really know anything about how to train AI so maybe some time in the future there will be something to show for this. To create the dataset I am taking a bunch of different books and working backwards to create the Sentient Sims prompt. That includes the characters, location info, and all the details about each block of text so that it essentially is working backwards to create a huge dataset that is in the same format as Sentient Sims. More on this in the future!
I am also creating an “auto” mode in Sentient Sims AI that will use a collection of AI models to do different things in Sentient Sims. Some models are better at doing different things, so being able to use whichever model is best for each part of the job automatically can greatly improve Sentient Sims.
Reasoning models are coming soon, I am hooking up Deepseek R1. Reasoning models are different because it takes a bunch of time to “think” about what it’s going to say before it gives an answer. This is highly desirable for Sentient Sims because they usually follow the rules of the prompt better. R1 has fairly good writing style too. Look for this coming soon!
I will be adding an “impersonate” button in the chat window, so that it can generate what it thinks you are going to say next.
In general Sims modding terms, I have made a lot of progress breaking things down for Sentient Sims. I am able to track when a Sim sleeps, and get all the Sentient Sims memories from that time span and build off of that. I also am able to generate Drama Nodes which opens up a lot of possibilities. I also figured out how to compile Python for Mac and Windows, a special version that allows me to bring in third party libraries and binaries. This means I can add a lot of functionality that wasn’t previously possible. This hasn’t been done before and will allow things like custom overlay and UI and memory modding outside of the existing python running The Sims 4. It also allows python things that weren’t possible before like including SSL binaries for making secure HTTPS requests.
