Vore AI Dungeon

Forum for the Vore games, and other downloads
Forum rules
Don't ask about updates. If there is an update, it will simply be posted. If someone is committing to a timeline, they will just tell you the timeline without needing anyone to ask.

Use Looking for master thread when you are not posting about an existing game.


You use this forum and website at your own risk for all links and uploads. There is no quality control or malware scanning or testing done here. Proceed with caution and use a virtual machine (VM) for any uploads strongly recommended. Virus/malware scanners alone are generally not good enough.

Vore AI Dungeon

Postby DragOnDude » Fri Dec 02, 2022 10:43 am

Okay! We all know that this AI has by the months if not years started to become more and more shit. But recently I'd say it's been doing half-good answers. Admins, feel free to delete this post if it doesn't check into the criteria but for those who are willing to give AI Dungeon a second chance, here are a few scenarios and worlds that are fun for those who want vore, micro, and macro. Worth to mention that it's free but with a shit ton of ads. There's that 7 days trial that with time, I've managed to claim 3-4 times by now because of the constant renewals. If you're like me and want to explore AI Dungeon again and are willing to do it for a long time, 9.99 USD isn't THAT much of a big deal.

Gaia: Furry NSFW. Vore can be done here as well, I've tried it.
https://play.aidungeon.io/main/worldSta ... 2fd658a2b6

Macromancy Sandbox: The memory is pre-written to remember commands such as shrink, shrink other, grow, grow other. You can also "DO" "grow to 20 feet" or "shrink to 5 inches".
https://play.aidungeon.io/main/scenario ... cab85ba7c9

Voros: The World of Vore: Customizable appearance, race, setting, and everything. This is the most vore friendly world as it also allows reformation.
https://play.aidungeon.io/main/scenario ... 5506502ef9

LuaTech: This is a creator who focuses on size stuff.
https://play.aidungeon.io/main/profile?username=LuaTech

All of these scenarios/worlds can do the same stuff but sometimes the pre-written stuff teaches the AI what to focus on. Also, there's unlimited "Try" and ad-free if you do the trial/9.99 USD. Sometimes you gotta need to use "Edit" because I'd say 70% of the time it's been close but no cigar for me.
DragOnDude
New to the forum
 
Posts: 19
Joined: Wed Jan 19, 2022 5:04 pm

Re: Vore AI Dungeon

Postby JDandAliProductions » Sat Dec 03, 2022 11:52 am

Voros is actually terrible, it asks you to customize everything and then tells you that it's actually gonna make up all the stuff for you, it will then force you down the route it wants to go by claiming you've done and said things that you didn't
If you wanna support what I do then please check out my Patreon, I'm working on a novel series too https://www.patreon.com/AliRenegade
User avatar
JDandAliProductions
Participator
 
Posts: 180
Joined: Sat Dec 28, 2019 6:51 pm

Re: Vore AI Dungeon

Postby fieldmousse » Sat Dec 03, 2022 1:01 pm

You should check out https://chat.openai.com/chat
while it lasts
It's not built to roleplay with you, but it is fully capable of writing stories for you and editing them
User avatar
fieldmousse
Participator
 
Posts: 218
Joined: Tue May 16, 2017 12:34 am

Re: Vore AI Dungeon

Postby MasterMage13 » Sat Dec 03, 2022 5:39 pm

As far as AI goes, I've had much better luck with NovelAI. (https://novelai.net/)

Especially if you write something with the AI, re-write it how you want it in proper detail (by default the AI has a really hard time understanding vore, it seems) and then take the story and run it through the module trainer and then try a new story with that module. It works pretty good once the AI figures out what you want.

Use the memory box to help keep the AI on track of where characters are and what is happening to them. I've gotten it to be... pretty graphic and mostly coherent with only minor adjustments.

And it's pretty privacy oriented. Everything is either cached locally or encrypted on their servers, supposedly.

I didn't like AI dungeon that much. And OpenAI has a LOT of content rules, and it's live moderated sometimes. I signed up for one of OpenAI's APIs and they have a pretty strict content policy.
User avatar
MasterMage13
Been posting for a bit
 
Posts: 42
Joined: Tue Jan 25, 2022 8:02 pm
Location: Nevada

Re: Vore AI Dungeon

Postby AlluringPredation » Sat Dec 03, 2022 8:48 pm

I still remember when Latitude/AI Dungeon announced they were going to do live moderation of people's stories when they introduced a filter for content, anyone who tripped the filter would have their content reviewed.

One person got banned for having a boy in the cockpit of a plane, another for trying to make babies with their wife, and another for riding a horse in one of the default story prompts.

OpenAI, who provided AI hosting for AIDungeon began sharing story context to foreign workers on TaskUp to evaluate and classify for potential harm.

You'd think that the end of it, but no. Someone hacked the AI server and downloaded everyone's stories for an entire month, as nothing was encrypted or properly secured.

And there was more, so much more. Basically all they do is create front-end to interact with companies that host the AI, they do not know how to do anything of real value.
Discord for artists and those who commission them! https://discord.com/invite/F3UDqqAx53
User avatar
AlluringPredation
Participator
 
Posts: 292
Joined: Fri Mar 27, 2015 11:10 am

Re: Vore AI Dungeon

Postby AlluringPredation » Sat Dec 03, 2022 8:48 pm

I still remember when Latitude/AI Dungeon announced they were going to do live moderation of people's stories when they introduced a filter for content, anyone who tripped the filter would have their content reviewed.

One person got banned for having a boy in the cockpit of a plane, another for trying to make babies with their wife, and another for riding a horse in one of the default story prompts.

OpenAI, who provided AI hosting for AIDungeon began sharing story context to foreign workers on TaskUp to evaluate and classify for potential harm.

You'd think that the end of it, but no. Someone hacked the AI server and downloaded everyone's stories for an entire month, as nothing was encrypted or properly secured.

And there was more, so much more. Basically all they do is create front-end to interact with companies that host the AI, they do not know how to do anything of real value.
Discord for artists and those who commission them! https://discord.com/invite/F3UDqqAx53
User avatar
AlluringPredation
Participator
 
Posts: 292
Joined: Fri Mar 27, 2015 11:10 am

Re: Vore AI Dungeon

Postby DragOnDude » Sun Dec 04, 2022 6:29 am

MasterMage13 wrote:As far as AI goes, I've had much better luck with NovelAI. (https://novelai.net/)


Maybe I should try it out...
DragOnDude
New to the forum
 
Posts: 19
Joined: Wed Jan 19, 2022 5:04 pm

Re: Vore AI Dungeon

Postby ian66613 » Mon Dec 05, 2022 1:23 pm

The problem with AI is that it doesn't like to interact with small sized characters, often making them back to normal human size within two sentences. It also doesn't remember items inside items that are inside other items, so it's very terrible at recursive memory, which is terrible for vore and size content. NovelAI is pretty good at this, especially with the algorithm that does second-person better than the others, but you have to be aggressive when it comes to writing, you have to make it remember more often.

I'd recommend trying to make yourself a fairy, and literally act out a story. You will VERY quickly realize even at a few inches tall, the story makes you average human-sized for sexual interactions. Especially on AI Dungeon.
Ideal Vore Games · Vore Preferences · My OC (4 inch (10cm) fairy boy.)
I am NOT into Furries/Bestial/Animals.
User avatar
ian66613
---
 
Posts: 1304
Joined: Mon May 16, 2011 9:12 pm
Location: Wisconsin

Re: Vore AI Dungeon

Postby MasterMage13 » Mon Dec 05, 2022 9:21 pm

AI Dungeon isn't very good.

But NovelAI will figure it out eventually. Especially if you put a lot of time into tweaking the lorebook and memory and training the modules. You do still need to interact with it and push it in the right direction. Sometimes adding vore related terms to the phrase bias helps and adding some sex terms to the phrase ban helps. Part of the problem seems to be that the AI often translates "eating" someone to mean "eating someone out" or oral sex of some kind. So it needs to be trained around that.
User avatar
MasterMage13
Been posting for a bit
 
Posts: 42
Joined: Tue Jan 25, 2022 8:02 pm
Location: Nevada

Re: Vore AI Dungeon

Postby Takoma » Wed Dec 07, 2022 5:23 am

the only options are NovelAI KoboldAI and that one steam named ai rouge like, two of them require you to probably have a PC built for something that of a crypto miner *just make sure you don't try buying a GPU that turns out to be a used one from a miner whose likely gone banana's since the recent crashes* and the other one is...a decent price tag for subscription to an extent *with a side note with image gen is a separate pricing, but the lads with novel don't glow bright and have all that data somewhere in a location that may or may not be a NSA facility* but jokes aside those three are basically the only options since the Mormon gutted the golden goose ai.
(oh and character ai is also a lost hope since they also did an ai dungeon.)
Image
User avatar
Takoma
New to the forum
 
Posts: 8
Joined: Mon Sep 04, 2017 9:06 am

Re: Vore AI Dungeon

Postby ian66613 » Wed Dec 07, 2022 11:17 am

Takoma wrote:and that one steam named ai rouge like


I'd highly recommend against this, especially considering it costs 7.99 USD and is extremely restrictive on how it sets up the world/encounters. For instance, if you manage to somehow get yourself trapped in a "vial" it will consider it an actual location like a town or field.

KoboldAI is a little too advanced I think for most people, especially when you have to install a specific module for NSFW, which I don't think even I know how to do despite reading the documentation.
Ideal Vore Games · Vore Preferences · My OC (4 inch (10cm) fairy boy.)
I am NOT into Furries/Bestial/Animals.
User avatar
ian66613
---
 
Posts: 1304
Joined: Mon May 16, 2011 9:12 pm
Location: Wisconsin

Re: Vore AI Dungeon

Postby Takoma » Thu Dec 08, 2022 5:35 am

ian66613 wrote:
Takoma wrote:and that one steam named ai rouge like


I'd highly recommend against this, especially considering it costs 7.99 USD and is extremely restrictive on how it sets up the world/encounters. For instance, if you manage to somehow get yourself trapped in a "vial" it will consider it an actual location like a town or field.

KoboldAI is a little too advanced I think for most people, especially when you have to install a specific module for NSFW, which I don't think even I know how to do despite reading the documentation.


I mean it's not like the lads at /aids/ and NAI have already prsented work arounds/ways to put a monkey wrench into the steam one over in the guides section as I won't specific as there might be lurking goons of the AI dungeon trying to hammer out fun like its bedrock edition Minecraft.

And kobold is more of the fun sandbox of text gen and image gen from what I've been told from an unspecified user who had abandoned crypto mining.
User avatar
Takoma
New to the forum
 
Posts: 8
Joined: Mon Sep 04, 2017 9:06 am

Re: Vore AI Dungeon

Postby MrXenomorph » Sun Dec 11, 2022 6:14 pm

Regarding KoboldAI, you can use Colab option instead. You just choose the model you wish to use, and wait for it to start... The only issue will be to get the TPU to start it, since otherwise, big models won't work. It is a bit of a gamble to get one at times.
I can share settings for KoboldAI if someone is willing to try it out, all the adjustments that I made caused to be... quite, detailed and descriptive.
User avatar
MrXenomorph
Been posting for a bit
 
Posts: 57
Joined: Tue Mar 26, 2019 8:41 pm

Re: Vore AI Dungeon

Postby MasterMage13 » Sun Dec 11, 2022 8:53 pm

Takoma wrote:the only options are NovelAI KoboldAI and that one steam named ai rouge like, two of them require you to probably have a PC built for something that of a crypto miner]


You need really great hardware to get your own AI to run on your own personal system. And it pretty much requires a very high end GPU. Like a 3090, because of the VRAM requirements.

I have a very powerful and high resource server at home and I tried to get GPT-J running on it. And while it DOES run... without a GPU it takes over a minute to generate responses. And forget about fine tuning it.

I might look into one of the smaller models, but even then, I think you would need something specifically built for it.

You can rent GPU cloud instances designed for AI "research" though. https://lambdalabs.com/service/gpu-cloud/pricing they also sell boxes for local development, but the cheapest one is like 5 grand.
User avatar
MasterMage13
Been posting for a bit
 
Posts: 42
Joined: Tue Jan 25, 2022 8:02 pm
Location: Nevada

Re: Vore AI Dungeon

Postby Airinnann » Tue Dec 13, 2022 2:39 pm

MrXenomorph wrote:Regarding KoboldAI, you can use Colab option instead. You just choose the model you wish to use, and wait for it to start... The only issue will be to get the TPU to start it, since otherwise, big models won't work. It is a bit of a gamble to get one at times.
I can share settings for KoboldAI if someone is willing to try it out, all the adjustments that I made caused to be... quite, detailed and descriptive.


Would be interested in the details you mentioned, as It's been a little....weird when I've fumbled around trying to get it to participate in vore in general. Either dancing around it or confusing the characters in the scenes.
Airinnann
New to the forum
 
Posts: 15
Joined: Fri Dec 18, 2015 5:48 pm

Re: Vore AI Dungeon

Postby MrXenomorph » Thu Dec 15, 2022 7:18 pm

Airinnann wrote:
MrXenomorph wrote:Regarding KoboldAI, you can use Colab option instead. You just choose the model you wish to use, and wait for it to start... The only issue will be to get the TPU to start it, since otherwise, big models won't work. It is a bit of a gamble to get one at times.
I can share settings for KoboldAI if someone is willing to try it out, all the adjustments that I made caused to be... quite, detailed and descriptive.


Would be interested in the details you mentioned, as It's been a little....weird when I've fumbled around trying to get it to participate in vore in general. Either dancing around it or confusing the characters in the scenes.



PM me if you wish.
Also, which model are you using, as the starting question? Erebus 20B does the job TOO well on my side, with the settings I've setup.
User avatar
MrXenomorph
Been posting for a bit
 
Posts: 57
Joined: Tue Mar 26, 2019 8:41 pm

Re: Vore AI Dungeon

Postby ian66613 » Fri Dec 16, 2022 10:14 am

MrXenomorph wrote:
Airinnann wrote:
MrXenomorph wrote:Regarding KoboldAI, you can use Colab option instead. You just choose the model you wish to use, and wait for it to start... The only issue will be to get the TPU to start it, since otherwise, big models won't work. It is a bit of a gamble to get one at times.
I can share settings for KoboldAI if someone is willing to try it out, all the adjustments that I made caused to be... quite, detailed and descriptive.


Would be interested in the details you mentioned, as It's been a little....weird when I've fumbled around trying to get it to participate in vore in general. Either dancing around it or confusing the characters in the scenes.



PM me if you wish.
Also, which model are you using, as the starting question? Erebus 20B does the job TOO well on my side, with the settings I've setup.



How do I set the GPU Memory Allocation manually? it always over-allocates when trying to load:

Code: Select all
       DEVICE ID  |  LAYERS  |  DEVICE NAME
               0  |      22  |  NVIDIA GeForce RTX 3060
             N/A  |      11  |  (Disk cache)
             N/A  |      11  |  (CPU)
INIT       | Loading model tensors:  61%|######1   | 406/664 [00:39<00:34,  7.57it/s]Exception in thread Thread-15:
Traceback (most recent call last):
  File "B:\python\lib\site-packages\transformers\modeling_utils.py", line 399, in load_state_dict
    return torch.load(checkpoint_file, map_location="cpu")
  File "C:\KoboldAI\torch_lazy_loader.py", line 295, in torch_load
    callback(retval, f=f, map_location=map_location, pickle_module=pickle_module, **pickle_load_args)
  File "aiserver.py", line 2429, in lazy_load_callback
    model_dict[key] = model_dict[key].to(device)
RuntimeError: CUDA out of memory. Tried to allocate 288.00 MiB (GPU 0; 12.00 GiB total capacity; 11.17 GiB already allocated; 0 bytes free; 11.19 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "B:\python\lib\site-packages\transformers\modeling_utils.py", line 403, in load_state_dict
    if f.read().startswith("version"):
  File "B:\python\lib\encodings\cp1252.py", line 23, in decode
    return codecs.charmap_decode(input,self.errors,decoding_table)[0]
UnicodeDecodeError: 'charmap' codec can't decode byte 0x81 in position 1866: character maps to <undefined>

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "aiserver.py", line 2604, in load_model
    model     = AutoModelForCausalLM.from_pretrained(vars.model, revision=vars.revision, cache_dir="cache", **lowmem)
  File "B:\python\lib\site-packages\transformers\models\auto\auto_factory.py", line 463, in from_pretrained
    return model_class.from_pretrained(
  File "aiserver.py", line 1822, in new_from_pretrained
    return old_from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)
  File "B:\python\lib\site-packages\transformers\modeling_utils.py", line 2326, in from_pretrained
    model, missing_keys, unexpected_keys, mismatched_keys, error_msgs = cls._load_pretrained_model(
  File "B:\python\lib\site-packages\transformers\modeling_utils.py", line 2536, in _load_pretrained_model
    state_dict = load_state_dict(shard_file)
  File "B:\python\lib\site-packages\transformers\modeling_utils.py", line 415, in load_state_dict
    raise OSError(
OSError: Unable to load weights from pytorch checkpoint file for 'cache\models--KoboldAI--GPT-NeoX-20B-Erebus\snapshots\1a80940a290452af71caf17a8e520955eb338e0f\pytorch_model-00014-of-00023.bin' at 'cache\models--KoboldAI--GPT-NeoX-20B-Erebus\snapshots\1a80940a290452af71caf17a8e520955eb338e0f\pytorch_model-00014-of-00023.bin'. If you tried to load a PyTorch model from a TF 2.0 checkpoint, please set from_tf=True.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "B:\python\lib\threading.py", line 932, in _bootstrap_inner
    self.run()
  File "B:\python\lib\threading.py", line 870, in run
    self._target(*self._args, **self._kwargs)
  File "B:\python\lib\site-packages\socketio\server.py", line 731, in _handle_event_internal
    r = server._trigger_event(data[0], namespace, sid, *data[1:])
  File "B:\python\lib\site-packages\socketio\server.py", line 756, in _trigger_event
    return self.handlers[namespace][event](*args)
  File "B:\python\lib\site-packages\flask_socketio\__init__.py", line 282, in _handler
    return self._handle_event(handler, message, namespace, sid,
  File "B:\python\lib\site-packages\flask_socketio\__init__.py", line 826, in _handle_event
    ret = handler(*args)
  File "aiserver.py", line 466, in g
    return f(*a, **k)
  File "aiserver.py", line 3917, in get_message
    load_model(use_gpu=msg['use_gpu'], gpu_layers=msg['gpu_layers'], disk_layers=msg['disk_layers'], online_model=msg['online_model'])
  File "aiserver.py", line 2607, in load_model
    raise RuntimeError("One of your GPUs ran out of memory when KoboldAI tried to load your model.")
RuntimeError: One of your GPUs ran out of memory when KoboldAI tried to load your model.


I manually set the layers to 22 of 44, and that didn't help. Filed an issue on the github: https://github.com/KoboldAI/KoboldAI-Client/issues/192

Even after getting it to load with a lower GPU Layer setting, it gives this issue: https://github.com/KoboldAI/KoboldAI-Client/issues/193
Ideal Vore Games · Vore Preferences · My OC (4 inch (10cm) fairy boy.)
I am NOT into Furries/Bestial/Animals.
User avatar
ian66613
---
 
Posts: 1304
Joined: Mon May 16, 2011 9:12 pm
Location: Wisconsin

Re: Vore AI Dungeon

Postby MrXenomorph » Fri Dec 16, 2022 6:14 pm

Is this even Google Colab? ^^; I use colab exclusively, as my pc won't handle big models (and thx to the AID, I have a fear of small models now lol), therefore I can't help with that technical stuff. Sorry.
User avatar
MrXenomorph
Been posting for a bit
 
Posts: 57
Joined: Tue Mar 26, 2019 8:41 pm

Re: Vore AI Dungeon

Postby Airinnann » Sat Dec 17, 2022 3:22 am

MrXenomorph wrote:
Airinnann wrote:
MrXenomorph wrote:Regarding KoboldAI, you can use Colab option instead. You just choose the model you wish to use, and wait for it to start... The only issue will be to get the TPU to start it, since otherwise, big models won't work. It is a bit of a gamble to get one at times.
I can share settings for KoboldAI if someone is willing to try it out, all the adjustments that I made caused to be... quite, detailed and descriptive.


Would be interested in the details you mentioned, as It's been a little....weird when I've fumbled around trying to get it to participate in vore in general. Either dancing around it or confusing the characters in the scenes.



PM me if you wish.
Also, which model are you using, as the starting question? Erebus 20B does the job TOO well on my side, with the settings I've setup.



Dunno if you got the PM or not, was saying I was using the erebus 13b for my starting model. Thought it might be adventure mode not playing nice with it.
Airinnann
New to the forum
 
Posts: 15
Joined: Fri Dec 18, 2015 5:48 pm

Re: Vore AI Dungeon

Postby ian66613 » Thu Jan 05, 2023 9:01 pm

I tried Google Colab using Erebus 6B (Since that's the highest version colab offers), and gave it the following prompt:

Code: Select all
The world has 90 percent females, five percent effeminate males, and five percent hermaphrodites that look like women. Most women and hermaphrodites are at least six feet tall, with the average being six and a half feet tall. The world has a medieval tone, with some steam machines being used here and there in larger cities. Magic is dominant here, and many try to study it in great detail. Magic can be anything from shooting fireballs, to shrinking others, or even turning people into objects. The world is ruled by women, and men tend to be either playthings or used for menial chores and labor. You are a small four inch fairy boy, and have found yourself in a small town. You currently find yourself inside of a buxom woman's home, inside a dresser drawer. You are unsure of how you got here, and are unsure if you will be able to get free of this woman.


It immediately disregarded it and made up its own story after a line break:

Code: Select all
Personality: You are an extremely quiet little boy, who tends to hide in corners and behind things when he isn't trying to sneak up on someone. You are very shy, and don't like to speak unless spoken to. You have been raised by your mother, and have never seen a man before. You are extremely loyal to your mother, and would do anything for her.

Backstory: Your parents were killed when you were only three years old, and you were taken in by your grandmother, who was the only person you knew at the time. She took care of you until you were eight years old, when she passed away as well.


I don't know what I expected. But NovelAI is definitely better than this.
Ideal Vore Games · Vore Preferences · My OC (4 inch (10cm) fairy boy.)
I am NOT into Furries/Bestial/Animals.
User avatar
ian66613
---
 
Posts: 1304
Joined: Mon May 16, 2011 9:12 pm
Location: Wisconsin

Next

Return to Vore game

Who is online

Users browsing this forum: Account132, aceinthehole, aem, alberth122, animegeek, aperson1234567890, Bantha1979, BepisBepis, BirdBrain, Dankwank91, dima123, disposable914, DJMemphis, DkA95, Dnagi, Dracon_Kitsune, DragonFire03272, KaZe_DaRKWIND, kevin888, meaning4, Microbastard, Otherperson, pigest, Pingu2023, RageofRagaki, reeneee, SceptileSamurai, Seclusive1, SkyStrikerAceRaye, Something101, TBVAS2, TGTyros, theskorpionkid, tomatus, ViskaWolf, VoidInVoid, wariodude128, Xman3114, Yandex [Bot], Yujacy, Zirqx