- Sep 14, 2020
- 79
- 183
You must be registered to see the links
You don't have permission to view the spoiler content.
Log in or register now.
Ren'Py Influencing [v0.1.17.8] [Golden Crow]
- Thread starter BaasB
- Start date
-
- Tags
- 2dcg adventure anal sex animated big ass big tits character creation cheating corruption dating sim exhibitionism female protagonist footjob futa/trans futa/trans protagonist gay harem interracial lesbian male protagonist management masturbation mobile game multiple protagonist netorare oral sex prostitution romance rpg sandbox simulator vaginal sex voyeurism
- Jan 21, 2021
- 1,337
- 2,593
- Jan 30, 2023
- 304
- 283
- Jan 21, 2021
- 1,337
- 2,593
I would love for the trans characters to have the male sex positions included, kind of weird how they don't have access to some of them.
- Apr 16, 2020
- 128
- 386
- Aug 12, 2023
- 55
- 29
- Jun 29, 2019
- 15
- 68
I thought I'd explain how the AI chatting works. The recommended LLM is a 11b which could be ran easily on 16gigs of VRAM. I'm using Kobold.
It seems the initial prompt is around 1000 tokens long. I suspect running the LLM at a context limit of ~8000 is fine. The prompt included all the stats of the characters. You can type in whatever you want in your communication with the characters in game and the LLM will generate the response. Obviously an 11b is not SUPER smart but it's good enough.
Sometimes it works fine and other times it crashes the game.
Whatever you type into the chatbox is weighed against the character's likes/dislikes and adds or takes away from the friendship/love points you have with the character.
They used the LLM in clever ways in the game so all the information you get in regular game chat could be gained in AI chat. If you ask for a phone number and the AI responds by saying yes, you get that character added to your contacts list in game + a friendship point boost. The same happens when you ask for interests. However, it's not a back and forth conversation. You're the one always asking the questions due to how the AI works.
I'm a little too lazy to organize this prompt, but this is the information that's sent to the AI so they can generate a response.
You can describe your character's actions and the other character you're talking to will respond with described actions of their own.
It seems the initial prompt is around 1000 tokens long. I suspect running the LLM at a context limit of ~8000 is fine. The prompt included all the stats of the characters. You can type in whatever you want in your communication with the characters in game and the LLM will generate the response. Obviously an 11b is not SUPER smart but it's good enough.
Sometimes it works fine and other times it crashes the game.
Whatever you type into the chatbox is weighed against the character's likes/dislikes and adds or takes away from the friendship/love points you have with the character.
They used the LLM in clever ways in the game so all the information you get in regular game chat could be gained in AI chat. If you ask for a phone number and the AI responds by saying yes, you get that character added to your contacts list in game + a friendship point boost. The same happens when you ask for interests. However, it's not a back and forth conversation. You're the one always asking the questions due to how the AI works.
I'm a little too lazy to organize this prompt, but this is the information that's sent to the AI so they can generate a response.
You can describe your character's actions and the other character you're talking to will respond with described actions of their own.
Last edited:
- Jul 5, 2017
- 327
- 410
How do you configure it to make it work though? I'm getting a "something went wrong generating the answer" error.
- Jul 5, 2017
- 327
- 410
Was trying aihorde since its apparently the free one, even selected all models. All other ones require a base IP and port.
Yeah, it doesn't look like the horde is working for me either. I'm not getting any sort of error log to hint at what's wrong.
- Dec 21, 2016
- 183
- 317
Local LLMs are free too, aside from the VRAM requirements and your power bill. On the bright side if it's winter where you are your PC can be an effective space heater!
Several backends, including KoboldCPP which I recommend, can split layers between GPU and CPU making the VRAM requirements a little easier, but it'll generate a good bit slower if you do split. Usually it's best to pick a LLM with a parameter count and a quant that allows you to fit all layers on your GPU or at least most of them. For KCPP you'll want a GGUF quant which is nice an easy since it's just one file.
It's worth learning how to use a VRAM calculator to pick an appropriate LLM quant:
Several backends, including KoboldCPP which I recommend, can split layers between GPU and CPU making the VRAM requirements a little easier, but it'll generate a good bit slower if you do split. Usually it's best to pick a LLM with a parameter count and a quant that allows you to fit all layers on your GPU or at least most of them. For KCPP you'll want a GGUF quant which is nice an easy since it's just one file.
It's worth learning how to use a VRAM calculator to pick an appropriate LLM quant:
You must be registered to see the links
You can run Kobold on a mid-level PC. Download Kobold
You must be registered to see the links
, download the model
You must be registered to see the links
. Run Kobold.exe, click [browse] and select the model. Wait for it to run (will take a few minutes, a new browser window will pop up when it's done), then you can play. You might not be able to run it if you're on a laptop or a low-end PC though.