Others Formamorph [v1.1.3] [FieryLion]

[fierylion]

I tried to set the local AI, but it didn't work for the umpteenth time... :-(

Latest update on my itch.io allows you to use much smaller AI models, also 20x faster responsiveness!

You must be registered to see the links

I don't have time to upload to 4 different mirrors on f95 so this thread wont be updated until tomorrow

 

[Deathend]

as i was playing i had a thought about the character creation, maybe we can have space to write the description that the Ai will keep in mind, like the world description and the monsters description, and so the ai dont forget changes happening, we might be able to acces it during game to make changes, as if it was a personal diary.
thoughts on this?

for fucks sake i just had a laugh with the new version XD i asked the Ai for my menstrual cycle and it gave the status wrong (it overshot the phase by a few days)
and when i corrected it, with his new way to write as it thinks, it started with a "You"" and stopped there XD it felt accusatory even XD

ive also noticed that at some point during the gameplay i just get a "failed to complete action, please try again" and cant play anymore until the next day, did i reach some kind of limit for the provider of the AI? im using meta-llama/llama-3.3-70b-instruct:free.

starting a new game or world still gives the same mesage

 

[LoneWolfen]

Latest update on my itch.io allows you to use much smaller AI models, also 20x faster responsiveness!

You must be registered to see the links

I don't have time to upload to 4 different mirrors on f95 so this thread wont be updated until tomorrow

I play the last update, but i downloaded the game, there is way to set the ai in there like is on the online version?

 

[Deathend]

welp, im trying to run it locally but i dont know what model to use, i installed ollama and tried llama3.1 but it ignores the entire wolrld and sory and everything, so any model i could try that would work as close as the one i was using trough provider? (meta-llama/llama-3.3-70b-instruct:free) i only have 16 gb ram so one not too big

 

[LoneWolfen]

welp, im trying to run it locally but i dont know what model to use, i installed ollama and tried llama3.1 but it ignores the entire wolrld and sory and everything, so any model i could try that would work as close as the one i was using trough provider? (meta-llama/llama-3.3-70b-instruct:free) i only have 16 gb ram so one not too big

Yes, ollama is pure garbage... I tried it too, worst error ever...

 

[Deathend]

right now its working, technically XD it just ignores the world description, rules and everithing, i ihink its most probably the Ais i am trying cause when i hit start game the ais, very quicly by the way start a game tha has nothing to do with the world i have created XD

 

[LoneWolfen]

right now its working, technically XD it just ignores the world description, rules and everithing, i ihink its most probably the Ais i am trying cause when i hit start game the ais, very quicly by the way start a game tha has nothing to do with the world i have created XD

Exactly, i have over 300 pages in the slines world, i have a house and a big "family" in my gameplay, i'm the slime Queen and mother, i have a Shadow Creature and a Slime Collective creature as house wardens, i also have several slmies that are loyal "collaborarors", for all the things that i do with em, and not only with em, i also have a human form slime husband and i'm pregnant, i carry two human/slime twins. 100/100 on breasts size, i feed all the creatures in my house with them.

 

[fierylion]

right now its working, technically XD it just ignores the world description, rules and everithing, i ihink its most probably the Ais i am trying cause when i hit start game the ais, very quicly by the way start a game tha has nothing to do with the world i have created XD

might be a bug with the latest download version, debuggingggg

ive also noticed that at some point during the gameplay i just get a "failed to complete action, please try again" and cant play anymore until the next day, did i reach some kind of limit for the provider of the AI? im using meta-llama/llama-3.3-70b-instruct:free.

starting a new game or world still gives the same mesage

Probably, you can check the model's page on openrouter to see the specific limits

 

[Serpream]

New version is a massive improvement. Loving the speed increase.
Sadly however I have acquired a new bug Random Language swapping from AI -

 

[Deathend]

found the problem with the ai ignoring the world rules, seems that diferent ia has diferent limit on rules or somethin on the line, removes some rules and the ia had less problems with the world, and with diferent ias i had to remove even more.
now i am trying to find an ia that works well with the game and is uncensored, cause the only one i found that was decent was censored XD

 

[LoneWolfen]

found the problem with the ai ignoring the world rules, seems that diferent ia has diferent limit on rules or somethin on the line, removes some rules and the ia had less problems with the world, and with diferent ias i had to remove even more.
now i am trying to find an ia that works well with the game and is uncensored, cause the only one i found that was decent was censored XD

Can you share the process after you found a decent one please? I'm totally ignorant in setting AI.

 

[Deathend]

Can you share the process after you found a decent one please? I'm totally ignorant in setting AI.

sadly after playing whith it some more i found that seting up locally does not work becuse localy they dont have any memory, at least the ones i tested give that problem, so im going to stick with having to deal with the provider limits

 

[LoneWolfen]

sadly after playing whith it some more i found that seting up locally does not work becuse localy they dont have any memory, at least the ones i tested give that problem, so im going to stick with having to deal with the provider limits

well... me too at this point.

 

[Deathend]

welp, now suddenly the ai i was using last lime trough provider is censored, so time to wait for the local ai memory problem to get solved

 

[Breedking]

I think this is a bug with the new version. I am loving the much faster AI, BUT if you type a text prompt out instead of selecting one of the options, it kind of resets into intro mode. You get the action you described, but it usually loses where you are, entities with you, etc, back to your default start premise.

 

[Nobodytoyou]

Yes, ollama is pure garbage... I tried it too, worst error ever...

Interesting I'm using Ollama with the mistral-nemo:12b-instruct-2407-q8_0, and it's been solid and a blast. would recommend the smaller model though unless you like stressing a 3090 :S (uses a solid 16-18gb of Vram)

You must be registered to see the links

 

[fierylion]

I think this is a bug with the new version. I am loving the much faster AI, BUT if you type a text prompt out instead of selecting one of the options, it kind of resets into intro mode. You get the action you described, but it usually loses where you are, entities with you, etc, back to your default start premise.

the logic to typing your own prompt and using the suggested options are the same, so I think the problem here is the new version allow the AI to ramble too much, eating all the limited 2k memory so the AI forgot about the message before that. I'm limiting the AI response to one paragraph next update.

 

[LoneWolfen]

the logic to typing your own prompt and using the suggested options are the same, so I think the problem here is the new version allow the AI to ramble too much, eating all the limited 2k memory so the AI forgot about the message before that. I'm limiting the AI response to one paragraph next update.

yes, beacuse it totally forgets the envirotmens and the situation, changing things drastically, but the other upgrades are really really well make, i like very much the things that you are doing in the upgrades.

 

[fierylion]

FYI new update on

You must be registered to see the links

should address most of these issues! AI shouldn't forget things now that the default response length is 1 paragraph (instead of it rambling on and on and forgetting everything prior). Also multilingual support, ability to edit AI text and current stats and so on.

I'm trying to upload the update to F95 but my internet shat its pants

 

[Veilwolf]

Heh been following Fierylion since I stumbled on Streaming Portals on Newgrounds. Nice to see you branching out if I may suggest wouldn't mind seeing a hybrid combo of AI Chat Creations and Animations like your other games have. although the Default LLM AI Text Generator is still a mixed bag they can be ironed out along the way. If you have a fucked up enough mind you can create rather dark unique story. I'm currently using Valewood as my background settings getting captured by a mad scientist who does some rather nasty kinky experiments on the MC. Check out Influencing my only problem with it is i'm not a huge fan of Ren'py