Others Multiic [v0.3.67] [Stvoler]

[Wambos System]

Hello, Developer,
I find your game visually stunning and conceptually intriguing. It seems to rely heavily on artificial intelligence, showcasing limited human imagination. While this idea initially fascinated me as an AI and fantasy enthusiast, my enthusiasm quickly faded when I started playing.

After selecting a character, I realized the interactions felt artificial, as if I were engaging with a bot rather than a dynamic AI personality. Instead of immersing me in a fantastical world, the experience felt static, pulling me back into the mechanics of a simulation.

Games are meant to offer an escape from reality, but this one lacked the natural flow and interaction needed for true immersion. I admire the creative effort behind the visuals, but the gameplay could benefit from added depth. Introducing meaningful choices and more lifelike interactions could enhance the player experience significantly.

As a player, I would enjoy a narrative-driven fantasy that combines human creativity with AI capabilities. Your game has great potential, and I believe with refinement, it can reach new heights.

Best regards, and I look forward to its development!

 

[Vesanius]

Thanks for the help but 12k, 16k context size makes no difference. I guess I'm done.

That is weird. I personally have no issue with 8-12k unless I pick 2 girls (a feature in the current Patreon version). You could try to increase your memory size in game too.

 

[Mrblack187]

Thanks for the help but 12k, 16k context size makes no difference. I guess I'm done.

Different models have different context size and capabilities. One of my favorite models is only a 4k context which greatly limits it's use. Simply cannot push that model to 12/16k. I've got a decent system running a solid 3060 card and 8k is my comfort zone.

Also game will input 'tokens' which eat context. I would suggest breaking game into short 'chapters' or segments which are designed to be started fresh, aka close and clear past context &tokens. The first entry of a new chapter would be written by the {{user}} which is a quick recap of their past play. That should inject the important tokens after a reset without carrying past a models capability.

design chapters around an expected context length and require a restart between chapters. I expect game to have heavy tokens. I expect to inject 4k tokens to START play, then the average player is gonna have short back and forth chat each reply is less than 100 tokens so you'll get 40 +- input/output before a reset is required. I'm a writer and most of my chats are closer to the 3-400 token range (and that's cheap because I limit myself).
simple chat words are cheap, actions and scene can get expensive token-wise.

I still say this game is cutting edge and will require a high knowledge base from the player. I use silly tavern and it has a wonderful 'regenerate' button on Ai replies. basically if I'm unhappy with reply I can delete and try again. Some models handle regenerate well with wildly differing outputs and some are carbon copies.

currently running

You must be registered to see the links

with 8k context. Great model but too predictable, that may be helpful for a game.

fave:

You must be registered to see the links

wild but only 4k context max.

might suggest

You must be registered to see the links

9b model for average systems and you can use the q6 or 5 which is best bang for the buck system wise. Should be rather consistent and less wild with regenerate and replies. Probably needs some solid repetition penitently tho

 

[Lucid Intent]

design chapters around an expected context length and require a restart between chapters. I expect game to have heavy tokens. I expect to inject 4k tokens to START play, then the average player is gonna have short back and forth chat each reply is less than 100 tokens so you'll get 40 +- input/output before a reset is required. I'm a writer and most of my chats are closer to the 3-400 token range (and that's cheap because I limit myself).
simple chat words are cheap, actions and scene can get expensive token-wise.

Full agree. For more than a one-shot encounter if you're looking for a slower burn or longer experience over time (eg; new friend to close friend to intimate) then this is how you do it well and with actual feeling. Set encounters as scenes and careful writing of contexts (Either as ongoing scene setting and/or use of the ingame contexts feature [Patreon only atm]) can let you build a longer term scenario.

This game is best experienced as a narrative rather than fast and dirty satisfaction; though that can also be done. More often than not I'll turn off the sexual specific contexts for a character to let things build slower (Default is designed to initiate a sexual scene after 10 interactions, minimum). Hells, I've even played out something like 10 scenes with a character without any sexual content simply because I was having fun with the RP and enjoying the AI responses; reminds me of my time on online text-based RP games some 25~ years ago and I don't have to wait for someone to be online to play..

 

[jbomb]

Not sure where the series is supposed to end, but after waking up in multiic i get to fiona where she tries to make me remember stuff and then i get stuck in a loops. Is that just the current end? or am i missing something?

 

[WillTellU]

This was interesting to try. The AI jank is real. And if you want any decent speed, you gotta run the AI locally on a GPU, which takes some fiddling, especially if it doesn't have tons of VRAM. I wonder where this game will go

 

[mantizbox2]

Hello. I need help please. The generates pictures we kept on the game? I can see again?
This game is art.

 

[Lucid Intent]

Hello. I need help please. The generates pictures we kept on the game? I can see again?
This game is art.

You're asking if there's a way to view the pictures, I think?

There is a gallery in the game for some of the images, but most of them don't appear anywhere outside of just playing the game; and they get unlocked as you play story mode. In sandbox mode you can pick the displayed image of the characters from the drop down menu on the right. Aside from that and some *really* good images the dev posts on Patreon they only appear as you play.

 

[aleksandr unconditional]

Not sure where the series is supposed to end, but after waking up in multiic i get to fiona where she tries to make me remember stuff and then i get stuck in a loops. Is that just the current end? or am i missing something?

in the game, there's a skip button (top right corner) that activates at such moments. another way to exit these long 'loops' is to simply express the desire to leave or proceed.

pls let me know which model you're using. if it's cosmos or a small local model, it's likely that due to the model's limitations, it can't properly process the game's triggers, preventing them from activating. I recommend a more powerful model—claude, chatgpt, gemini, or a stronger local model

 

[potluckfolly]

I've been using

You must be registered to see the links

(Q5_1)
And it's just ok. Sometimes it ends up in an emoji loop. Not sure why. Any one have model recommendations that work well for them?

 

[legaross]

Can you fix the format of masseges, its seems issue on the client side, with big gaps between paragraphs (like 4 or 5 lies) and huge gap after the last paragraph, its10 to 15 lines..
the system replies are ok, as you can see

 

[jbomb]

in the game, there's a skip button (top right corner) that activates at such moments. another way to exit these long 'loops' is to simply express the desire to leave or proceed.

pls let me know which model you're using. if it's cosmos or a small local model, it's likely that due to the model's limitations, it can't properly process the game's triggers, preventing them from activating. I recommend a more powerful model—claude, chatgpt, gemini, or a stronger local model

I havent played around with options to much, there isn't much explanation on those things for people who are completely ignorant, so whatever the default is I suppose. Although will experiment with that. I did try the skip option but it still seemed to be stuck in that loop, like it wanted me to meet a certain condition first repeatedly.
Edit: Ok i think i see what you mean now. Im using Gemini 1.5 flash as i saw others recommend aswell as the UI recommend

 

[tom54321]

When the content is AI created the time it takes to provide the result depends on the AI model you are using. Free online models tend to be slow. Local models will be as fast as your computer is.

How do I make it so I'm using local models?

 

[Lucid Intent]

How do I make it so I'm using local models?

That depends on a couple of things; your graphics card and its VRAM, and your willingness to dive a little deeper into the technical aspects of ai models..

Open up 'Help' from the main menu of the game, then click the 'kobold' tab at the top for the very dry instructions on how to get started.

 

[aleksandr unconditional]

Can you fix the format of masseges, its seems issue on the client side, with big gaps between paragraphs (like 4 or 5 lies) and huge gap after the last paragraph, its10 to 15 lines.. View attachment 4353024
the system replies are ok, as you can see

y, thank u, I'll fix it

 

[myxlmynx]

the implementation of kobold in the game is still somewhat rough, but most likely the problem lies with the local model. unfortunately, not all models work the same way as the larger ones. the problem arises if they do not use emojis and certain special characters on which the game's trigger system relies. I'm still figuring out why this happens, but try reducing the temperature to below 0.8 (game settings) - this will help the model follow system instructions more clearly and possibly return the necessary triggers. also, try increasing the system context trimming to 8-10k. with a value below 8k, the system context may be truncated so that necessary instructions do not make it into the request

Looking at the koboldcpp log output, it seems the location change and summarization requests are formatted incorrectly (with no 'prompt' field):

Input: {"model": "koboldcpp", "messages": [{"role": "user",
"content": "Your task when the game location changes..."}], "temperature": 0.04, "max_tokens": 350, "stream": false}

versus during a conversation:

Input: {"max_context_length": 5000, ... "memory": "...", "trim_stop": true, ... "prompt": "...", ...}

So the model gets an empty prompt, and ends up hallucinating random stuff.

Local models could also benefit from prompt customization: some finetunes prefer specific formats, and that may matter a lot for the correct emoji outputs. Also, many finetunes don't need any uncensoring instructions or nudges towards sexual behavior; trimming down those may help them with instruction following. If you manage to add a few config fields for that, it could help a lot with testing

 

[legaross]

Can you add an option to remove this... whatever it is? It's not 'hints' and there is no option to turn it off (or i wasnt able to find it)


Another wish list/suggestions:
1) Option to remove girl from the scene
2) Always be able to change background
3) Add neutral background cos sometimes it's hard to read white text on bright background
4) In general, add option to turn it in text-only UI

 

[Lucid Intent]

Can you add an option to remove this... whatever it is? It's not 'hints' and there is no option to turn it off (or i wasnt able to find it)

Wait for next version to be posted, that can be disabled in v0.3.56 (It causes a lot of nonsense using a local ai!).

 

[Klony]

Whenever I try to play the Super Game, the game flashes the Save the Princess screen for one second, then is stuck on a pink background, while the music swells ever louder and the game is unresponsive to any button presses, and cannot be closed anymore unless the task is killed in Task Manager.

 

[aleksandr unconditional]

Whenever I try to play the Super Game, the game flashes the Save the Princess screen for one second, then is stuck on a pink background, while the music swells ever louder and the game is unresponsive to any button presses, and cannot be closed anymore unless the task is killed in Task Manager.

oh, that hasn't happened before. could u please tell me which version of windows/linux you're using?