Others Multiic [v0.5.04] [Stvoler]

[myxlmynx]

Is this game set up for Vulkan/AMD on Linux? Tried doing local kobold instead of web tunneling to try and get faster responses but the log is all messed up and it just loads forever. Maybe I did the setup wrong but I Launched Kobold with my model, confirmed everything was working in the web ui, launched multiic, went to settings and picked local and looked up my model from the menu on the left, set a context value, then started the game. Then when I get to the bedroom description the log looks like this, tries to run CPU only with 4k context instead of using my GPU.

As I understand it, "Koboldcpp local" will try to use the built-in Koboldcpp server, which is a CUDA binary (no Vulkan acceleration). To use your own local server, launch it as usual first, then choose "Koboldcpp remote" with your local URL, typically "http://localhost:5001" (I'm also on Vulkan + AMD + Linux, and that's exactly how I set up Multiic here). You'll see the game requests on your Koboldcpp server logs.

 

[kurwa99]

so for image generation i need to get grok or openAI api keys and looking into it they cost money per promt. is there a way to get immage generation for free ?

 

[IsaacOtto]

so for image generation i need to get grok or openAI api keys and looking into it they cost money per promt. is there a way to get immage generation for free ?

Great question, also is there a way to generate porn images based on the context ? The only results I've had so far with the OpenAI image generation is horrendous, doesn't even look like a person lmao

 

[random fella]

I also am not sure how to even generate an image with grok, always says it can not generate images

 

[Jinkxy]

A few questions for the sandbox mode.
Is there a good way to initiate scenes with multiple people?
I had some success to get one additional person into it, with the prompt at the top saying "X wants to join", but more then one i could never make work.
Also even with one sometimes i get funny results, like for example asking for Fiona and prompt comes up as "Jo wants to join" and if you click the model of Jo would join but the Chat would refer to her as Fiona. Also in those cases most AI models seem to get confused and try to fit Fiona AND Jo into the conversation.

Is there any prompt i can use to make someone join that doesnt fail 70-80% of the time?

 

[Recouper]

Does that mean I can use any gguf model that I want to with the game?

Yes. I've been trying to find the best one.

 

[Jinkxy]

Yes. I've been trying to find the best one.

Let me know if you find something good as im doing the same.

My current winner is L3-8B-Stheno-v3.2-Q8_0-imat.gguf so far.


Edit1:
Tested "cognitivecomputations_Dolphin3.0-R1-Mistral-24B-Q6_K.gguf" and "dolphin-2.5-mixtral-8x7b.Q4_K_M.gguf" today , those had also pretty nice results.

 

[jager20150]

new version on patreon 0.4.92 for Multiic

 

[PGU]

Nice girls...


*le hard pass*

THX for sharing...

This is just like VR only games, it's pandering to such a small audience. Just how many people does the dev think have internet access nowadays? It's a shame, i really wanted to try it. Maybe one day i get the dial-up hooked up and can play all these fancy new games.

 

[Cryptist]

10 Gs? Freakin' forget it. Compress to under 3 Gs, or die. I literally have 2 dozen games in my pending folder. I don't have to waste time downloading Moby Dick or War And Peace.

10 Gs indicates crappy game design to me. Some of the games I have followed through years of updates are still only a couple of Gs in size.

 

[chrono1337]

10 Gs? Freakin' forget it. Compress to under 3 Gs, or die. I literally have 2 dozen games in my pending folder. I don't have to waste time downloading Moby Dick or War And Peace.

10 Gs indicates crappy game design to me. Some of the games I have followed through years of updates are still only a couple of Gs in size.

then use the internet ai instead of the local model, its 2gb. He seems to have upgraded the LLM model which explains the increase in size.

 

[Jinkxy]

Got a problem on my first playthrough with saiga_nemo.
During the dinner conversation everything switched to russian until im back in my room again.

 

[NDogmeat]

Yeah, I'm also playing in russian now. It just switches from English randomly. Great!

 

[jager20150]

Yeah, I'm also playing in russian now. It just switches from English randomly. Great!

It's just a learning experience for the AI. Just keep trying to type English and it will solve itself. I've had that problem many times and I've learned how to deal with it. think of it like a kid just learning on how to speak and read there are many mishaps along the way until the lessons are learned

 

[Daijoubudesu]

I just started and sure it's very quirky but my god the potential long term is just outstanding. As the associated technologies for this project become more fine-tuned and accessible, I can't help but imagine how great games like this will be to get immersed into. Thanks for sharing and updating this project!

 

[itiosolis]

I think a walkthrough explaining how to get all of this to work properly is neccisary. I cannot figure out how to get online or offline to work whatsoever.

 

[hotte]

Man, no walkthrough or guide. I am spending 80% of the time shooing away the girls to avoid sexy time in hopes to maybe explore more in act 0. Still, I get no free roam time.

And I think doing stuff outside the predetermined path also kind of breaks the game.
I formed a Special Task Force with Josephine and Fiona and we unleashed Hell over the City. City and NPC's weren't present.
At this time I managed to peek into the secret door in Lauras room, it said it is a a Lab but there wasn't anything of interest. Is it empty because I found it outside the regular game loop?
Struggling to reach the Yoga scene, I am always getting cut off by Laura. I've found no way to check Lauras room during within the regular path.
Some rooms also have no background images. I don't know if the AI doesn't create them or if there are none. I'm using a paid version of gemini, so no limitations, and image generation is turned on. I want to concentrate on the next acts after I have all the clues in act 0.
A token counter would also be pretty cool. So I can see if the AI generates and image and it isn't displayed or such.

 

[panalok]

Let me know if you find something good as im doing the same.

My current winner is L3-8B-Stheno-v3.2-Q8_0-imat.gguf so far.


Edit1:
Tested "cognitivecomputations_Dolphin3.0-R1-Mistral-24B-Q6_K.gguf" and "dolphin-2.5-mixtral-8x7b.Q4_K_M.gguf" today , those had also pretty nice results.

You must be registered to see the links

This is the one I can really recommend. If you have 12 GB VRAM use the Q8_0 Quant. It doesn't preach, it's pretty descriptive and realistic(*) and fairly liberal.

(*)You should be aware though that many LLMs that are somewhat fine tuned for roleplaying are kinda unrealistic once you really "engage" with them. At least the 50+ or so I came across.

It's is also really important to flash out your characters with custom context. I haven't played the game in a while but the character definitions the author provides are somewhat generic and already preconditioned towards the player.

If you really like the sandbox mode you might enjoy using KoboldAI lite a lot better. It doesn't have fancy pictures by default but the customization options are way better and you can even rollback conversations and a whole bunch of other things like, voice input, image generation, voice output and so on. The only thing limiting you is your imagination and time.

 

[aleksandr unconditional]

Got a problem on my first playthrough with saiga_nemo.
During the dinner conversation everything switched to russian until im back in my room again.

I see the issue - the system context in that part wasn't translated into english, which could have caused the lang switch on some models. thx, I'll fix it

 

[aleksandr unconditional]

so for image generation i need to get grok or openAI api keys and looking into it they cost money per promt. is there a way to get immage generation for free ?

for now, yes - and imgs generated via the grok/openai api are heavily censored, so it's more for fun than anything serious. but integration of local image models (stable diffusion) into the game is in development and should be available here as soon as next week •ᴗ•
a

You must be registered to see the links

, a bit more detail