AI tools for traditional games

The Something Awful Forums > Discussion > Traditional Games > AI tools for traditional games

Doctor Zero: Sep 21, 2002; Would you like a jelly baby?
It's been in my pocket through 4 regenerations,
but it's still good.

Oh man I want to dog pile on the ethics point, but I will restrain myself.

I tried the exact same thing. ChatGPT actually knows enough about Ironsworn and Starforged (which is strange because it came out after the bots supposed �data cut off�) to be able to do an okay job.

Limitations: it�s only good at creating smalls thing. Characters, villages, situations, etc. if you expect it to create a complex and interesting plot thread for the entire game, LOL no. It has no concept of compelling drama, conflict, or motivations. It has the storytelling ability of a 3 year old.

Second, it can�t keep everything �in mind� very long so you are best to prompting it to give you small things and you stitch them together in a logical way. Otherwise it drifts off into the weeds quickly. It also has a tendency to jump over good plot points. You have to rein it back. It will say something like �a you quickly solve the mysterious plague. Later, bandits are seen�� and you�re like �hold on, let�s get back to the plague.� It will also make poo poo up a lot. I tried to get it to GM a Villains and Vigilantes game (superheroes) and it blithely made up new heros and villains rather than use established lore.

Third, you can teach it things, but again, only for a period of time. I taught it the Starforged universal truths (because it only knew about half of them and made up stupid poo poo for the rest) and used it to generate a background that worked pretty well. But you will need to keep those in mind going forward because it won�t.

Fourth, it doesn�t like violence. If you force the matter it will say stuff like �seeing how outnumbered they are, the pirates quickly surrender.�

Even with all that it�s still a very useful assistant. But it�s not able completely GM anything, at least not yet.

# ? Jun 16, 2023 23:41

Adbot: ADBOT LOVES YOU

# ? May 22, 2024 18:14

feverish and oversexed: Mar 9, 2007; I LOVE the galley!

feedmyleg posted:

I've tried it but found that its "memory" is too short for it to be much fun for long. The illusion is really compelling for a bit, but as soon as you hit it forgetting key details and hallucinating things that don't fit with the narrative you've been in a few times, it feels less like playing a game and more like writing fiction. And constantly having to remind it of the "rules" and summarize play feels tedious and frustrating. There's a model that OpenAI is testing that ups the memory from 4000 tokens to 16,000 tokens which feels much better for these purposes, though, so maybe it'll be a better experience soon.

That's the thing, you can expect it to do this much (yet)

I personally am using it to help my narration, expand on oracle stuff, and introduce elements I hadn't thought of otherwise to once again, great success.

It cannot DM fully for you, but it certainly assists.

Another user in the reddit thread mentioned using the API feature and I've asked them to expand on it in DM.

Also wondering if anyone has experience doing something similar with novel.ai or other AIs.

Doctor Zero posted:

Fourth, it doesn’t like violence. If you force the matter it will say stuff like “seeing how outnumbered they are, the pirates quickly surrender.”

Even with all that it’s still a very useful assistant. But it’s not able completely GM anything, at least not yet.

Yes exactly this, a lot of people get really up in arms when people try to talk about it as replacing the GM but it's just not there yet.

Also I just wanted to say that by using the narrate this, I had no trouble having a fight scenes or violent scenes, as long as I'm using the "narrate this" and Oracle structuring rolls as negative :airquote:

# ? Jun 16, 2023 23:49

Humbug Scoolbus: Apr 25, 2008; The scarlet letter was her passport into regions where other women dared not tread. Shame, Despair, Solitude! These had been her teachers, stern and wild ones, and they had made her strong, but taught her much amiss.; Clapping Larry

I use ChatGPT to come up with names and countries/political styles/economics. It's easier than rolling off random tables for me and I can ask it to describe them if any are more interesting. I don't use it for DMing per se, but it speeds up prep a ton.

# ? Jun 17, 2023 01:53

Bucnasti: Aug 14, 2012; I'll Fetch My Sarcasm Robes

The players in my DnD game get weirdly fixated on random mundane stuff, and tend to gloss over actual important plot and worldbuilding things.
A location in the dungeon they're exploring right now is a wizards laboratory and I know they'll get obsessed with reading every book.
So I used ChatGPT to generate books for me, and I told it specific world items and dungeon hints to include in some of them. I'm really happy with the results.

Cosmic Constellations By Rowan Silvermoon

Cosmic Constellations: Unlocking the Secrets of Astrology" takes readers on an illuminating journey into the depths of astrology, revealing the hidden wisdom and cosmic connections that shape our lives. This comprehensive guide explores various aspects of astrology, providing a rich blend of theoretical knowledge, practical techniques, and profound insights.

The book begins by introducing the foundations of astrology, including the history and development of this ancient practice. It explains the significance of celestial bodies such as the sun, moon, planets, and constellations, and how they influence our personalities, relationships, and life events.

Readers will delve into the intricacies of the zodiac, exploring the thirty two astrological signs and their unique characteristics. Each sign is carefully examined, highlighting the key traits, strengths, challenges, and compatibility factors associated with them.

I'm hoping they'll make the connection that the magical doo-dad they found in another room has 32 symbols on it and realize that they're related to the zodiac. Althouth the way they operate I wouldn't be surprised if they spent all their time trying to figure out what the bookbinding was made of.

# ? Jun 17, 2023 01:57

feverish and oversexed: Mar 9, 2007; I LOVE the galley!

Humbug Scoolbus posted:

I use ChatGPT to come up with names and countries/political styles/economics. It's easier than rolling off random tables for me and I can ask it to describe them if any are more interesting. I don't use it for DMing per se, but it speeds up prep a ton.

It definitely is good for inspiration in this sense, especially with a little guided context.

Bucnasti posted:

The players in my DnD game get weirdly fixated on random mundane stuff, and tend to gloss over actual important plot and worldbuilding things.
A location in the dungeon they're exploring right now is a wizards laboratory and I know they'll get obsessed with reading every book.
So I used ChatGPT to generate books for me, and I told it specific world items and dungeon hints to include in some of them. I'm really happy with the results.

-snip-

I'm hoping they'll make the connection that the magical doo-dad they found in another room has 32 symbols on it and realize that they're related to the zodiac. Althouth the way they operate I wouldn't be surprised if they spent all their time trying to figure out what the bookbinding was made of.

I am not in a place yet where I want to DM, but I was considering down the line after I've run a few solo cycles in ironsworn or similar I might try to.

Do you use it actively while playing or is this all prep for you?

edit: my biggest frustration (probably everyones frustration) is the memory context lacking even in the "better" chatgpt4 model, but once you disconnect from trying to get it to be your DM completely, and use it more as an assistant, is where it shines rn

I've been meaning to gently caress around with some of those plugins for chatgpt 4 that allow reading from a PDF, but I haven't bothered yet since my current method is working fairly decently with mild corrections

# ? Jun 17, 2023 03:18

Fuzz: Jun 2, 2003; Avatar brought to you by the TG Sanity fund

There's chat AIs you can run on your home machines... you guys know that, right? They don't end up having the short memory limitations like Chatgpt.

# ? Jun 17, 2023 03:38

BrainDance: May 8, 2007; Disco all night long!

Fuzz posted:

There's chat AIs you can run on your home machines... you guys know that, right? They don't end up having the short memory limitations like Chatgpt.

For the most part they have much shorter memories (most 30B models have about a 2k token context limit? Somewhere around there?)

Larger context sizes are being experimented with but generally, the larger your context, the more vram you use.

You can do things like put all the memory in a vector database (I think that's what the superbooga oobabooga textui extension does?) But you can do that with ChatGPT too.

Leperflesh posted:

I'm curious (coming from a cloud software background), is everyone running "local" models only on their home machines, or have people played around with running on cloud infra? Like there are cheap and even free infra options out there.

People are running models in the cloud, especially because 30B models are right in that spot where a lot of cards can run them but not a lot of cards people have can run them (basically you need a 3090 or so)

I swear I had this one place bookmarked that had cheap, really easy to use options for renting some 3090s, you only pay for what you use but they keep your setup so you don't have to redo it every time like you would with vast.ai or something, and it was a lot cheaper and easier to work with than one of the big cloud hosts. Cannot find that bookmark anymore though.

BrainDance fucked around with this message at 04:20 on Jun 17, 2023

# ? Jun 17, 2023 04:13

feverish and oversexed: Mar 9, 2007; I LOVE the galley!

Fuzz posted:

There's chat AIs you can run on your home machines... you guys know that, right? They don't end up having the short memory limitations like Chatgpt.

I -might- be wrong, I'm not super tech savvy, but from my short foray into the "can I run it" the answer is no, and getting a machine that can will cost me upwards 2k and I'm here trying to nickel and dime chatgpt plus to avoid using the api sooooo

# ? Jun 17, 2023 04:20

Bucnasti: Aug 14, 2012; I'll Fetch My Sarcasm Robes

feverish and oversexed posted:

It definitely is good for inspiration in this sense, especially with a little guided context.

I am not in a place yet where I want to DM, but I was considering down the line after I've run a few solo cycles in ironsworn or similar I might try to.

Do you use it actively while playing or is this all prep for you?

edit: my biggest frustration (probably everyones frustration) is the memory context lacking even in the "better" chatgpt4 model, but once you disconnect from trying to get it to be your DM completely, and use it more as an assistant, is where it shines rn

I've been meaning to gently caress around with some of those plugins for chatgpt 4 that allow reading from a PDF, but I haven't bothered yet since my current method is working fairly decently with mild corrections

I occasionally use it actively when DMing, like if the players do something completely unexpected and I need to come up with a character name or items name on the fly (I'm terrible at naming stuff). I once used it to generate a unique magic item that I forgot to create before the game and it turned out pretty cool.
What I mostly use it for is list making, like if I need a list of something specific "What do you find in this drawer" stuff. It's really good at giving me a couple dozen interesting but mundane items, like "three random chess pieces", "A handful of foreign copper coins" or "A river rock with a bloody fingerprint on it". That last one really threw my group for a loop, they spent 15 minutes trying to figure out it's significance, and now im weaving it into the campaign. I compile the items ChatGPT gives me into a traditional random table generator and roll on it during game.

It's pretty good at making NPCs, but they tend to be a bit samey, which sometimes is all you need. When you just need a name and a bit of dialog for Villager #3 its cool, but I wouldn't use it for major NPCs or Villians without doing a lot of curating.

None of this is stuff you need ChataGPT for, it could be done with many random generators available, but what I like is that it's really fast, I just tell it what I want, I don't have to go looking for a specific chart or table or anything and it spits out good enough content for my purposes.

# ? Jun 17, 2023 08:18

Megazver: Jan 13, 2006

I was busting my players' balls about being murderhobos and murdering a random mimic without any attempts to negotiate and I told them about how this one was actually a delicate soul and wrote poetry, so then I had ChatGPT write some terrible, terrible Mimic poetry on the spot. Good times.

# ? Jun 17, 2023 10:43

sebmojo: Oct 23, 2010; Legit Cyberpunk

You must definitely post this, if you still have it

# ? Jun 17, 2023 10:56

Megazver: Jan 13, 2006

Mimic love be sofuh,
Big cushy place to snooze-uh.
Adventurers come, all brave and bold,
Mimic open wide, mouth full of gold.

Me gobble 'em up, tasty treat,
Crunchy bones and armored meat.
They swing swords, try to fight,
But mimic's jaws, they hold 'em tight.

Mimic not smart, but mimic know,
Adventurers make belly grow.
Munching on heroes, me feeling grand,
They no match for me sticky hand.

Mimic big and strong, sturdy and stout,
Swallowing warriors without a doubt.
They think me furniture, nice and soft,
But little do they know, they be aloft.

Mimic happy, mimic content,
Eating adventurers, their strength spent.
In shadows, me lurk, waiting for prey,
Sofa disguise, fooling 'em all the way.

So adventurers beware, be on guard,
Mimic lurks, sofa disguise hard.
If you sit on me, you might get chewed,
Mimic hungry, always in the mood.

===================

Now, admittedly this mimic loved eating adventurers (and being a sofa), so their murderhoboism was somewhat warranted. But hey, I didn't say it was a vegan poet mimic! And if you, say, see a tiger in the wild minding its own business your first thought shouldn't be "oh it's not a herbivore, let's murder it"!

Megazver fucked around with this message at 11:22 on Jun 17, 2023

# ? Jun 17, 2023 11:15

Doctor Zero: Sep 21, 2002; Would you like a jelly baby?
It's been in my pocket through 4 regenerations,
but it's still good.

Bucnasti posted:

I occasionally use it actively when DMing, like if the players do something completely unexpected and I need to come up with a character name or items name on the fly (I'm terrible at naming stuff).

Not arguing against, but you reminded me that some of our best NPC names came from creative brain lock.

DM: �A traveling halfling approaches and��
Me: �what�s his name?�
DM: (caught off guard) what?
Me: what�s his name, Bob? HE SURELY HAS A NAME!
DM: �uhh - Spugnoise.�
Me: �Spugnoise? What�s his last name then??�
DM: �er� Bugbomb!�
Me: �SPUGNOISE BUGBOMB is his name?
DM: �Yeah! Spugnoise Bugbomb!! The 3rd!�

And that�s how we got a recurring NPC called Spugnoise Bugbomb. :haw:

# ? Jun 17, 2023 11:58

Bucnasti: Aug 14, 2012; I'll Fetch My Sarcasm Robes

Doctor Zero posted:

Not arguing against, but you reminded me that some of our best NPC names came from creative brain lock.

DM: �A traveling halfling approaches and��
Me: �what�s his name?�
DM: (caught off guard) what?
Me: what�s his name, Bob? HE SURELY HAS A NAME!
DM: �uhh - Spugnoise.�
Me: �Spugnoise? What�s his last name then??�
DM: �er� Bugbomb!�
Me: �SPUGNOISE BUGBOMB is his name?
DM: �Yeah! Spugnoise Bugbomb!! The 3rd!�

And that�s how we got a recurring NPC called Spugnoise Bugbomb.

Fortunately ChatGPT produces some goofyness like this on it's own sometimes.
Most the names are fairly cliche but every once in a while it spits out a winner.

# ? Jun 17, 2023 15:21

Golden Bee: Dec 24, 2009; I came here to chew bubblegum and quote 'They Live', and I'm... at an impasse.

AI dungeon has been iterating on the GM model for a while, and has a few innovations. First, it�s unlimited free. Second, it lets you write prompts or fix what it sends. Third, you can rewind at will.

It�s a really solid app, and has integrated creating starters for other people as well.

# ? Jun 17, 2023 17:07

feedmyleg: Dec 25, 2004

Yeah, I can't believe that ChatGPT hasn't implemented the editing feature yet. It goes a LONG way in helping maintain consistency.

# ? Jun 17, 2023 17:15

Repaired Radio: Nov 13, 2017

One of my favorite parts of AI is that I don't have to pay artists if I want work done in their style. What's everyone favorite artists to try and replicate the style of using AI tools?

# ? Jun 17, 2023 20:06

KwegiboHB: Feb 2, 2004; nonconformist art brut
Negative prompt: amenable, compliant, docile, law-abiding, lawful, legal, legitimate, obedient, orderly, submissive, tractable
Steps: 32, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 520244594, Size: 512x512, Model hash: 99fd5c4b6f, Model: seekArtMEGA_mega20

Repaired Radio posted:

One of my favorite parts of AI is that I don't have to pay artists if I want work done in their style. What's everyone favorite artists to try and replicate the style of using AI tools?

My own.

# ? Jun 17, 2023 20:10

Colonel Cool: Dec 24, 2006

I don't really have a favourite artist. They're all kind of interchangeable meatbags to me.

# ? Jun 17, 2023 21:22

Raenir Salazar: Nov 5, 2010; College Slice

I was a bit busy the past week, but I'll be going over the past couple of pages and try to revise the WIP op.

# ? Jun 17, 2023 21:26

HOMOEROTIC JESUS: Apr 19, 2018; Verily I say unto thee, That this night, before the cock crow, thou shalt deny me thrice.

To add thread content, one of my favorite artist styles has been Pamela Colman Smith, the lady who did the "original" tarot card illustrations in 1909. I was making a custom tarot deck in Stable Diffusion with fine-tuning that eliminates the black-white balance problem. The cards are only half finished, and I could probably draw better art myself, but I had a good time just experimenting with Stable Diffusion's ink drawings. As for how I made them, I gave SD a base image of the Fool and had it inpaint a new image inside, thus why they're all still "the Fool".

Background: In my current homebrew campaign, there is a group of primordial beings called "The Immortals". They've long disappeared, but their iconography permeates through everything in the campaign's aesthetic.

The Claw (a six fingered troll claw associated with wishes)

The Babe (a gelatinous cube representing birth and death, the twilights of life)

The Dragon (typical dragon stuff about being a tyrant, associated with nobility and stubbornness)

The Moon (a snow-elf demon of dreams and the moon)

The Peacock (moss-woman of storms and magic and stuff)

The President (giant tortoise who invented democracy)

The Rose (getting a mouse and flowers into the same image was a genuine difficulty)

The Scholar (all knowing monstrosity made of ink)

The Siren (a siren lich associated with captivity, the sea, and ambition)

Maybe I'll finish them eventually, but if I do I'll probably do it by hand to get more aspects of each Immortal's portfolio in. It's impossible to get these AI to have more than two complex things in an image and do them well.

# ? Jun 17, 2023 21:44

Raenir Salazar: Nov 5, 2010; College Slice

WIP new OP V2:

quote:

AI tools can be a great resource for generating new content for traditional games like Dungeons and Dragons and Magic the Gathering and other TTPRGs. These tools can help you come up with new spells, feats, magic cards, and more, quickly and easily. Below is a list of tools that might be useful, be aware that there is a debate regarding the ethicacy of their training data, and the broader societal economic effects. To learn more you can go to the AI Debate thread in D&D. And to learn more about the technology behind AI like Machine Learning you can go to the Caverns of Cobol thread here.

Below are various tools that people use regarding generative AI.

Stable Diffusion:
https://playgroundai.com/ (this is just one example, there are many others)
This AI tool is designed specifically for generating new content for Dungeons and Dragons. It uses a machine learning model trained on a large dataset of existing spells and abilities to generate new, unique spells and abilities. You can use it to come up with new ideas for your character's abilities, or to add some excitement to your game by introducing new spells and abilities that your players have never seen before.

Here is a video tutorial on how to use ControlNet with Stable Diffusion for more control!

Midjourney:
https://www.midjourney.com/
This AI tool is designed to generate new images and graphics for use in games and other creative projects. It uses a deep learning model to generate original artwork based on a set of user-specified parameters. If you're looking for new magic cards or other graphics for your game, this tool might be worth checking out.

ChatGPT:
https://chat.openai.com/chat
This AI tool uses a variant of the popular GPT language model to generate text-based content. You can use it to generate new descriptions for spells or magic items, or to come up with ideas for new quests or adventures. ChatGPT is also highly customizable, allowing you to fine-tune the type of content it generates based on your specific needs.

Firefly

Firefly is the new family of creative generative AI models coming to Adobe products, focusing initially on image and text effect generation. Firefly will offer new ways to ideate, create, and communicate while significantly improving creative workflows. Firefly is the natural extension of the technology Adobe has produced over the past 40 years, driven by the belief that people should be empowered to bring their ideas into the world precisely as they imagine them.

Adobe claims it's model is "trained on a dataset of Adobe Stock, along with openly licensed work and public domain content where copyright has expired." making it one
of the more ethical options for delving into generative AI.

Additional Resources:
Here is a link to VTTRPG Resources, which seems to be a how-to/useful tips and tricks guide to prompts for getting useful content for TTRPGs.

Overall, these AI tools can be a great resource for generating new content for traditional games like Dungeons and Dragons and Magic the Gathering. Whether you're looking for new spells, magic items, or quests, these tools can help you come up with fresh ideas and add some excitement to your game. This thread is for encouraging you to share your experiences with AI tools and how you've used them to generate new content for your roleplaying games. This can be a great way to inspire others and help them get the most out of these tools. For example, you might share:

1) Examples of new spells, feats, or magic items you've generated using Stable Diffusion or ChatGPT
2) Artwork or graphics you've created using Midjourney or Dall-E
3) Stories or adventures you've created using AI Dungeon or other interactive storytelling tools

However remember that there's probably no one size fits all solution to generating your content. Many outputs at first might be mediocre and require significant guidance and handholding. By sharing your experiences, you can help others see the potential of these tools and give them ideas for how they can use them in their own games; and you can learn how to further improve the output you might derive from these tools. You might also discover that others have found new and creative ways to use these tools that you hadn't thought of before.

In addition to sharing your experiences with these specific tools, we also encourage you to share any other tools you know about that might be useful for generating new content for roleplaying games. There are always new tools and resources being developed, and by sharing your knowledge, you can help others discover new ways to enhance their games.

So don't be shy! Share your experiences with AI tools and other resources for generating new content for roleplaying games. You never know who you might inspire or what new ideas you might discover.

Suggested Behaviour & Etiquette
To repeat this isn't the AI Debate Thread, and isn't for Litigating the arguments and ethics surrounding AI. Please consider some
of the other threads, either in GBS, CSPAM or D&D if that has your interest.

With that in mind, Please be respectful, not just of people interested in AI, but likewise also of people broadly who have reasonable concerns.

As such just as much this isn't the thread to criticize people for their interest in integrating AI into their hobby, it isn't the place
to voice grievances regarding posters critical of AI. Please keep those disputes to more appropriate threads.

Rules Section TBD

This post originally generated by ChatGPT, modified June 17 2023 by a human.

The following I think might be too wordy for an OP, so I'd add it to a second and/or third reserved post.

quote:

More Detailed Explanations & Guides.

Some regulars of the previous thread have helped put together additional and more indepth resources!

By Megaman's Jockstrap:

Megaman's Jockstrap posted:
Stable Diffusion

This is a huge topic and you could literally write a book about it. I won't. We'll try to make this short(ish) and sweet.

At first glance, Stable Diffusion is not as good as MidJourney. Until SDXL (an upcoming model) it had a bug that meant everything in it was literally trained wrong, so all of it's produced images will generally looked more washed out and lower-contrast than Midjourney (this is somewhat fixable). It's base models have less cohesion and more artifacting than Midjourney. Given the same prompt, it will generally produce a weaker and less interesting image.

Well, why use it then? The answer is easy: it's very customizable and you can have a lot of control over it. Forgive the facile analogy, but if Midjourney is Windows then Stable Diffusion is Unix/Linux - the analogy even holds a little because Stable Diffusion is open source and Midjourney is closed and proprietary. The basic expectation here is that you are going to take the time to learn the in and outs and dig under the hood of Stable Diffusion to learn it. If not, stop right now. Use Midjourney.

We'll go over the four things that are good about Stable Diffusion that Midjourney doesn't offer:

1) You can run it on your local box

Yes, if your local box has an NVIDIA graphics card made in the last 10 years it can run Stable Diffusion. Obviously the beefier and faster the card the better this works, but I was getting perfectly acceptable (if slow) results on a decade-old i5 with a GTX 970. You can also run it on AMD stuff but I don't know how that works at all. I know it's a bit trickier and requires some special config work but people do it.

The de-facto interface for it is a web front-end called Automatic1111. There's a ton of tutorials out there on how to install Automatic1111 and get the SD models so I'll leave that to you and Google.

Because you host it you don't have to pay credits or setup an account or anything. You have control over it. You can just generate whatever images you want. It's FREEEEE (unless you're an insane person and buy a new graphics card for SD! don't do that!)

2) Customized models

Stable Diffusion released their models completely for free. Open source and all that. Because it's open source, about 4 minutes later somebody made a way to continue to train models. We're not going to get into how models work but just know that it's entirely possible to extend these models. For example, if a model doesn't know what a Porsche is, you could give it a bunch of pictures of a Porsche and say "that's a Porsche" and then it "knows" and can gen images of that. So lots of people have been extending the two biggest Stable Diffusion models (referred to by their version numbers, 1.5 and 2.1 - but mostly 1.5) with new stuff, or refining old stuff.

Because these models use Stable Diffusion as a base (so that they can continue to "understand" styles and concepts that the creators didn't specifically train them on) they're usually a decent size, 6 gigs or more. But they CAN be highly useful. I say CAN because some of them are very good, but some are WAY overtrained. If you use only a few images for a concept, or a bunch of very similar images, the model will not have any variance and keep generating things that look the same. For example, if I made a model and the only pictures of men I uploaded were of 1973's Mr. Universe Arnold Swartzenegger, then every time I asked for a picture of a man I would get some variation of Arnie (probably flexing). Well, people do this all the time. In fact I'm sorry to say that most people just train them on big tiddy anime waifus. No, I am not joking. The model training community are a bunch of terminal horndogs.

At the time of writing, Civitai.com hosts Stable Diffusion models. Should you choose to visit this site: NSFW NSFW NSFW the front page will instantly poison your mind and make you hate AI. Push past it. Search for RPG to find some decent stuff (and a loving ton of samey anime girls). The "RPG_v4" model (soon to be v5) is pretty good and great for generating portraits and weapons for your games.

3) LORAs

So let's say you have a model that you really like except for one small problem: it doesn't know what a BattleTech Marauder 2C looks like. Criminal! But who wants to train a model and generate gigs of data just to add a Marauder? Well you don't have to, luckily.

Think of a LORA as a post-it-note that you stick to a model that contains exactly one style or concept that you want it to use. It requires incredibly less computer power than adding to a model, and less work too. It's entirely feasible for you (yes you!) to train a LORA, the amount of images needed is shockingly small (5 - 7 gets you decent results, more is better of course). You can use these to do things that would be nearly impossible for Midjourney, such as outputting art in the style of RIFTs. You can get real powerful results here that are simply not obtainable with other methods using custom models and LORA styles.

A picture from a little known RIFTS sourcebook from 1992, RIFTS: Robot Boyz

Remember earlier when I said you could mitigate some of the aspects of Stable Diffusion's bad training bug? Well that's a LORA. https://civitai.com/models/13941/epinoiseoffset

4) Controlnet

The best way to think of Controlnet is a way to apply traits from one image to another, like telling an artist "I want this person to be doing this pose" or whatever. So if you like the general composition of an image, or the pose a person is holding, but you want other things to change completely, that's Controlnet. If you want to apply some aspect of an image to another image, that's Controlnet.

You might have noticed earlier that people are still using SD 1.5 even though 2.1 is out. Well, that's because ControlNet only works with it. That fact alone has kept it going.

When it comes to RPG applications it's really good for getting the facial expressions you want. Here's a rakish rogue with the prompt wanting him smiling:

That's just not good enough! I want a real big smile and a particular pose, let's go ahead and grab this picture that's pretty close to what I want and put it into Controlnet, then tell Controlnet to apply that facial expression to my guy:

It doesn't matter that it's a cartoon. If it's got a face, Controlnet can work with it. Here's the final output:

So we didn't get the Dreamworks Smirk (and there's a small artifact and some small teeth issues that I just didn't bother to fix) but he does look a lot happier! Also note that his head position and pose is exactly like the cartoon picture (although his eyes aren't looking the same way). And if my player cared enough we could try another picture or whatever, but this is good enough as an example I did in 5 minutes.

Here's another 5 minute example. I wanted to make a flower sword where the blade is coming out of petals near the hilt. A quick sketch in GIMP (not even using a tablet) and sending it to Controlnet on the RPG_V4 model with a very basic "a photograph of a flower sword, highly detailed blade" prompt goes from this:

to this:

Again, it's in the exact orientation I wanted, proportions what I asked for, no guessing for a magic prompt. It's not perfect but for 5 minutes it's a great player handout.

It also does full body poses, fingers and hands, architecture, edge detection, A LOT. Controlnet is magic. Controlnet is life. Covering what Controlnet can do is waaaaay beyond the scope of this post so instead here's a mega big tutorial on it: https://stable-diffusion-art.com/controlnet/

Final Thoughts

Stable Diffusion is huge and I've barely scratched the surface here (I did not mention image-to-image, interrogating, inpainting, Krita integration, hypernetworks, Dreambooth, etc), but sufficed to say you're not locked in to giving money to Midjourney or trying to find the "magic phrase" if you don't want to. You can host and use your own image diffusion system and get a really cool way to make player portraits, player handouts, maps, and anything else you could want. A very useful tool for upping the professionalism of your game.

Addendum regarding Stable Diffusion by KewgiboHB

KwegiboHB posted:
Amazing writeup! Thank you for this!

I just want to add there have been advances and an NVIDIA card isn't strictly required anymore. There is AMD, Intel, CPU, hell even iPhone support now. Yeah you can run this on your phone.
Here is a list of quick installers, no messing with git or PATH required.

- Automatic1111 - NVIDIA GPU - By far the most popular webui with an incredible array of options and extensions. https://github.com/EmpireMediaScience/A1111-Web-UI-Installer
- InvokeAI - AMD or Intel GPU - This uses DirectML instead of CUDA. https://invoke-ai.github.io/InvokeAI/installation/010_INSTALL_AUTOMATED/
- NMKD - NVIDIA GPU - Executable windows GUI not webui. https://nmkd.itch.io/t2i-gui
- ComfyUI - NVIDIA GPU - Node based with downloadable premade workflows. https://github.com/comfyanonymous/ComfyUI/releases
- Stable Horde - NO GPU - Crowdsourced donated compute free for those without other means. https://stablehorde.net/ links to a client interface with no installation required https://dbzer0.itch.io/lucid-creations
- Mac and iPhones - I don't know anything about Mac, here's a link anyways https://apps.apple.com/us/app/draw-things-ai-generation/id6444050820
- Runpod - 'The Cloud' - Just means someone elses computer. Rent someone elses GPU for money. Method of last resort. https://blog.runpod.io/stable-diffusion-ui-on-runpod/
- OpenVino - CPU mode - This takes forever but is still doable on literally a toaster, hell yeah stable toast. https://github.com/bes-dev/stable_diffusion.openvino

Regarding Local LLM's by BrainDance:

BrainDance posted:
Here's a quick thing I wrote up about running local LLMs for the next OP.

Open-source local LLMs are currently going through their Stable Diffusion moment. Before March open source LLMs were much weaker than GPT3 and mostly would not run on consumer hardware. We were limited to running GPT-Neo or, at best, GPT-J slowly. Training custom models on them was slow, hard, and poorly documented.
Then in March two things happened at about the same time. Meta released (it �leaked�) their LLaMA models, trained differently from previous open-source models with more training data to compensate for a lower parameter size making a 60B parameter model perform about as well as a 120B parameter model. Then the hardware requirements for running these larger models was drastically reduced through �quantization�, shaving off some bits and leaving the model a fraction of the size. A group at Stanford then trained a model on top of LLaMA in basically the same way as ChatGPT making the first actual open-source equivalent to ChatGPT.
Soon after LLaMA was released methods to run LLaMA and other models at an actually decent speed on a CPU were released.
Just like with Stable Diffusion, LoRA (Low Rank Adaptation) finetuning allowed LLMs to be finetuned with much weaker hardware than before, people could replicate exactly what Stanford had done on their own with a gaming GPU.

Now we�re at the point where we have new models that outperform the last by a huge jump almost every week. LLaMA was released in four sizes, 7B, 13B, 33B, and 65B. The 65B models are a thing most people still cant run, but the 33B models are, in many tasks, almost indistinguishable from GPT3.5.

What do I need to run these models?
7B models with 4-bit quantization require 6GB of vram to run with a GPU, or 3.9GB ram to run with a CPU
13B 10GB vram or 7.8GB ram
30B 20GB vram or 19.5GB ram
65B 40GB vram or 38.5GB ram
The low ram requirements don�t mean you can realistically run a 30B model on any computer with enough ram. You technically can, but it will be slow. Still, running 7B and 13B models on modern CPUs is probably faster than you think it�s going to be.
There are other kinds of quantization which used to make more sense when it was more difficult to use 4-bit models but that�s not all that relevant anymore.

How do I run these models?
There are so many frontends at this point but the two big names are oobabooga�s text-generation-webui and llama.cpp

Oobabooga�s text-generation-webui - The LLM equivalent to Automatic1111�s webui for Stable Diffusion. Download the one-click installer and the rest is pretty obvious. Allows you to download models from within it by just pasting in the huggingface user and model name, has extensions, built in powerful support for finetuning, works with CPUs and GPUs. It now supports 4-bit models by default. The one thing that does require a little bit more setup is using LoRA�s with 4-bit models. For that you need the monkey patch, and to start the webui with the �monkey-patch flag (stick it after CMD_FLAGS = in webui.py, this recently changed so some documentation tells you otherwise) instructions are here

Llama.cpp - started as a way to run 4-bit models on macbooks (which works surprisingly well) and is now basically the forefront of running LLMs on your CPU. Getting a lot of development, like currently the big thing is using the CPU but offloading what you can to the GPU to get a big speedup, sometimes outperforming GPU only models.
I don�t use it though, because I spent way too much on getting 24gb of vram

Which models do I use?
There are two main quantized model formats right now. Things are a little chaotic so who knows how long this will stay true, and one of the formats has even gone through a big update making all previous models obsolete (but just needing conversion) so things can change. Generally though:

GPTQ Models � For running on a GPU
GGML Models � For running on a CPU
Generally you�re going to want to have the original LLaMA models to apply LoRAs to.
Otherwise, almost all models get quantized by one guy right now, TheBloke https://huggingface.co/TheBloke
Right now, good general models are the Vicuna models, Wizard Vicuna uncensored, and for larger models (30B) Guanaco, though I don�t think an uncensored version of this exists yet. Even censored models here though are usually a lot less censored than ChatGPT. You can find all these models in different parameter sizes on TheBloke�s huggingface.
There are a lot of models that have a more niche purpose though. Like Samantha, a model not trained to be a generally helpful model but instead trained to believe she is actually sentient.

How do I finetune models?
This is really where the local models become incredibly useful. It�s not easy, but it�s a lot easier than it was a few months ago. The most flexible and powerful way to finetune a larger model is to finetune a LoRA in the oobabooga webui. This actually has good documentation here
The trickiest thing is formatting all of your training data. Most people are using the alpaca standard right now where this is a field for an instruction and then a field for an output. The AI learns when it gets an instruction it�s then supposed to generate an output like in its training data.
The other, easier way is to just give it a kind of trigger word. You have examples of the kind of data you want it to output all following a word you made up so it learns �when I see this word, complete for it with this kind of text�

This is an area though that gets incredibly complicated, there are too many different ways to do it and too many steps to just write down here. This though, is probably where open source models become most relevant to traditional games, models trained to be aware of specific lores or to respond to questions in a way that fits that world or who knows what else (I�ve had an idea for training a model on questions and answers from DnD masters before for the longest time, just as an experiment to see how much it learns the rules and how much it just hallucinates)

Another option for the hard way to finetune a whole model that is probably going to require renting an A100 somewhere is mallorbc�s finetuning repo, which started before LLaMA as a method for finetuning GPT-J. This is where I started out, before the oobabooga webui existed.

There is a lot more. There are ways to give an AI a massive memory from a larger database, experiments in much larger or even infinite context sizes are popping up, people trying out new formats and new ways to quantize models but, this post can only be so long.

Aside from Da Actual Rules we're waiting on Leperflesh and Anti for; anyone have any thoughts or feedback or suggestions on how to maybe rewrite some sections to be shorter? One thing maybe we can do is have the OP be a brieifer summary of the options out-there; and then everyone can post their own indepth guides that the OP can link to instead of posting everything in the beginning?

# ? Jun 17, 2023 21:56

feverish and oversexed: Mar 9, 2007; I LOVE the galley!

Bucnasti posted:

None of this is stuff you need ChataGPT for, it could be done with many random generators available, but what I like is that it's really fast, I just tell it what I want, I don't have to go looking for a specific chart or table or anything and it spits out good enough content for my purposes.

yeah, for me its all about the ease of accessibility. When I'm running my "mobile" version of simplified ironsworn on the go I just have two tabs open on my phone, the character sheet and the chatgpt window. I prefer my other way of playing, sitting down at the computer with my notebook and everything else + chatgpt and other AI open.

feedmyleg posted:

Yeah, I can't believe that ChatGPT hasn't implemented the editing feature yet. It goes a LONG way in helping maintain consistency.

I really hope they get around to doing this, I am doing periodic "summarize this" but depending on how long I've been "running" a session it will often mess up details or leave some out, so I "remind" it on the next message, the edit would be a godsend.

Repaired Radio posted:

One of my favorite parts of AI is that I don't have to pay artists if I want work done in their style. What's everyone favorite artists to try and replicate the style of using AI tools?

the one where instead of going to google and finding a image that sorta matches that somebody already made, I can have a more direct result with my solo rpg that I'm not sharing with literally anyone else but other hobbiests. That one.
Just like this:

Here Chatgpt interpreted my structured roll, I thought it was really funny so I fed the entire scene to bing to get this result:

Would you like to have a discussion about how accessibility is making this process incredibly therapeutic for me, a veteran with PTSD and other issues that doesn't feel comfortable in social situations with almost anyone, except through the safety of forums and the internet. Would you like to have a conversation about poverty, privilege and accessibility? no? I wouldn't in the AI tools for traditional games thread, personally.

edit: mixed up my AIs

feverish and oversexed fucked around with this message at 00:32 on Jun 18, 2023

# ? Jun 17, 2023 23:45

sebmojo: Oct 23, 2010; Legit Cyberpunk

A strong post imo. Let's leave it there please.

# ? Jun 18, 2023 00:12

The Eyes Have It: Feb 10, 2008; Third Eye Sees All
...snookums

Doctor Zero posted:

Even with all that it�s still a very useful assistant. But it�s not able completely GM anything, at least not yet.

On this topic, one little niche I found it does very well with is little bits like this:

In D&D 5e, what would be a reasonable DC number and applicable skill for [thing a player wants to attempt]?
In D&D 5e, the players approach a scholar named Vril to learn more about how to break a curse. Vril is a curt man who values getting a job done as quickly as possible for maximum reward. Here is how Vril responds to the players, and what he is thinking:

Basically an assistant that needs no appreciable amount of time to think.

# ? Jun 18, 2023 21:15

YggdrasilTM: Nov 7, 2011

Repaired Radio posted:

One of my favorite parts of AI is that I don't have to pay artists if I want work done in their style. What's everyone favorite artists to try and replicate the style of using AI tools?

I have some PG portraits made with Alphonse Mucha Art Nouveau style. I'm sure he doesn't mind, being dead for the last 84 years.

# ? Jun 19, 2023 09:55

Communist Thoughts: Jan 7, 2008; Our war against free speech cannot end until we silence this bronze beast!

think of the copyright holders the next time you make a picture to amuse your friends

# ? Jun 19, 2023 11:35

trapstar: Jun 30, 2012; Yo tengo un par de ideas.

Megazver posted:

The first session went okay. Now I have time to write this post instead of prepping!

So in addition to the main characters, the module also a page's worth of one sentence descriptions of optional NPCs you can plug in for flavor. Obviously I generated all of them as well. With some of these, as it goes, I used a picture I really liked even if it didn't quite fit.

The Smoking Girl. In the inciting incident of the adventure, a girl rapidly burning up into corrupted Yin Qi descends from the sky and desperately tries to warn the locals, before exploding into foul smoke. (Unless she is saved.)

Auntie Wei, a middle-aged matchmaker who knows every private scandal in Gongfang. Fat, beaming, and unctuous.

Blue-Button Mei, an elderly seller of ribbons, notions, and trinkets. She has justification for getting into any house�s female quarters, and often serves as a go-between for �private meetings� at the local shrine. Cultivates an air of doddering harmlessness, but has a sharp mind.

Crooked Zhong (�chong�), a charcoal-burner, now the Golden Lord�s priest.

Feng the Invincible, a young hooligan born with a strange dislocation of destiny; he always loses every fight he ever gets in, but due to some chance turn of events he never suffers serious injury. Only an appropriate divine miracle can overcome this fated survival.

Fishmonger Chin, rancid seller of river catfish, pesters strangers incessantly to buy a fish, if only to make the reeking young man go away. Seems totally oblivious to physical threats or dangers.

Ivan Szabo, a fur trader from Raktia who privately hates all Dulimbaians for their former conquest of his homeland. Will quietly help anyone who seems to be hurting them.

Marika Toth, a half-Raktian prostitute both beautiful and bitter. Has contacts among many of Gongfang�s elite, but desperately wants to leave both Dulimbai and Raktia, as she is accepted in neither place.

Mother Hao, a plump, sweet-mannered old procuress who keeps a house with her �daughters� inside town. Actually as heartless as any street pimp, and deals information to Big Feet Zu about likely marks or trouble. Her slave girls are terrified of her wrath; she sells the troublemakers to Xiao.

Old Father Lu, the oldest man in Gongfang at 107 years of age, astonishingly spry, gives oracular warnings and sage counsel that is always plausible yet catastrophically wrong. Locals pretend to hang on his every word but then discreetly ignore his advice.

Red Cap Hong, a skinny Taoist priest who sells folk-magic cures, lucky talismans, and exorcisms. Knows what Real Man Xiao is trying to do and can guess at his powers, but is too afraid to be seen opposing him.

Venerable Wan, a middle-aged priest with a patchy beard responsible for tending the shrine to Gongfang�s ancestral founders. His daughter was taken by Xiao as a �volunteer� and he is desperate to recover her, or at least learn her fate. He is greatly respected by the common folk of the town.

Xia Ming, a beautiful and tubercular young woman of good family, morbidly obsessed with her inevitable early death. Writes bad poetry, and is eager to do something important or romantically dramatic before she dies.

Xiang the Axe, a burly woodcutter who knows the location of many camps of Golden Lord devotees. Won�t willingly talk unless he gets drunk.

These are great!

# ? Jun 19, 2023 13:40

Fuzz: Jun 2, 2003; Avatar brought to you by the TG Sanity fund

I'm still trying to get the hang of using Firefly to make actual drawing style stuff, but it's aces at modifying photos. Asked her and my friend is okay with showing off this heavily modified photo I took of her. It was originally a close up and then I used FF to outpaint in the entire rest of it. Also changed her hair and eye color.

You can slightly see the seam where it started to just make up its own poo poo, I think it needs to have a more articulated base plane to help hide that.

Edit: it also didn't know what to make of her jacket and its sleeve pocket zipper, so it just sorta went hog wild with zippers, haha.

Fuzz fucked around with this message at 16:28 on Jun 19, 2023

# ? Jun 19, 2023 16:16

feverish and oversexed: Mar 9, 2007; I LOVE the galley!

The Eyes Have It posted:

On this topic, one little niche I found it does very well with is little bits like this:

In D&D 5e, what would be a reasonable DC number and applicable skill for [thing a player wants to attempt]?

In D&D 5e, the players approach a scholar named Vril to learn more about how to break a curse. Vril is a curt man who values getting a job done as quickly as possible for maximum reward. Here is how Vril responds to the players, and what he is thinking:

Basically an assistant that needs no appreciable amount of time to think.

having discovered more games exist besides ironsworn, I'm using Chatgpt4 currently to simply narrate my current run of "thousand year old vampire" and its actually doing really well since you "lose memories". I've been plugging in my ongoing character sheet, then how I interpreted the roll (or how I want the roll to be interpreted), and it paints the scene for me.

I've been boring my partner by talking about it, its basically like an open-world interactive book and its really fun and mitigates some of the issues I was having with other games and the token memories

# ? Jun 19, 2023 19:39

The Eyes Have It: Feb 10, 2008; Third Eye Sees All
...snookums

I also thought that these tools would be great to narrate something like Ironsworn, but for me it was just an idea -- I'm glad to hear you tried it out and got good results!

# ? Jun 19, 2023 19:45

feverish and oversexed: Mar 9, 2007; I LOVE the galley!

The Eyes Have It posted:

I also thought that these tools would be great to narrate something like Ironsworn, but for me it was just an idea -- I'm glad to hear you tried it out and got good results!

I actually developed a pretty good system to use with it, but I and an interaction with the creator about Ai and I'm petty and its tainted the game for me.

Thankfully I literally had 0 ideas that there was 1) so many other types of games in this genre, and 2) this poo poo is incredibly hackable so I'm just going to make my own version that's better for me that is designed to work with Ai.

I was sitting there the other day "well, what are my favorite aspects in a game?" and then I was like... why not just put those mechanics into one game... and now I have multiple new journals an-

# ? Jun 19, 2023 19:48

Megaman's Jockstrap: Jul 16, 2000; What a horrible thread to have a post.

So I had my first truly negative experience with AI. A player asked to submit some backstory to me and I said sure. I received 3 pages which surprised me greatly as this player is not really that type of table presence. I read it and it was just bland trash about frolicking in the woods near their cabin with their uncle and gathering herbs and poo poo. It wasn't interesting at all, although it did fit the background for a primal/ranger type.

I got suspicious and asked them if an AI wrote it and the answer was yes. They gave ChatGPT some sort of "write a nature background for my character" and it came up with this nonsense. They then sent it over to me. They hadn't even done more than skim it!

So now I have a rule with my players: do not use AI for the connective tissue of your backstory. I'm fine with them using it for names, inspiration, whatever, but the actual connective tissue needs to be players because honestly this loving sucked. It took them 3 minutes to waste 10 minutes of my time.

# ? Jun 26, 2023 03:11

trapstar: Jun 30, 2012; Yo tengo un par de ideas.

Megaman's Jockstrap posted:

So I had my first truly negative experience with AI. A player asked to submit some backstory to me and I said sure. I received 3 pages which surprised me greatly as this player is not really that type of table presence. I read it and it was just bland trash about frolicking in the woods near their cabin with their uncle and gathering herbs and poo poo. It wasn't interesting at all, although it did fit the background for a primal/ranger type.

I got suspicious and asked them if an AI wrote it and the answer was yes. They gave ChatGPT some sort of "write a nature background for my character" and it came up with this nonsense. They then sent it over to me. They hadn't even done more than skim it!

So now I have a rule with my players: do not use AI for the connective tissue of your backstory. I'm fine with them using it for names, inspiration, whatever, but the actual connective tissue needs to be players because honestly this loving sucked. It took them 3 minutes to waste 10 minutes of my time.

JFC...

I like feeding it character backstories I wrote and then having it write a scene in which that character shares their life philosophy. It actually gets it quite on point for the characters I've written. Here's an example:

trapstar posted:

Chat GPT is a pretty cool/fun tool to use to analyze and present the motivations and ideals of characters you have written. Here is the backstory I wrote for a villain of mine called Kor, Taker of Widows.

Pics (Disclaimer: Not AI Art, They Are Actual Artist Commissions)

Here is my villain backstory written by me:
"Kor was the son of Gor, chieftain of the Agarians, a barbarian tribe in the region of the world known as the �Barbarian Lands�. Growing up, Kor was willing to do anything to prove his strength. By the time he was fourteen, he was already a fearsome warrior, favoring the greataxe. As a young man, Kor became the chief war leader and enforcer for his father. He led many raids on other tribes especially against their mortal enemies the Terveri.

Soon after Kor grew into manhood, Gor arranged a bride for his son. He was to marry Thesna, daughter of Eretil, chieftain of the Franes, for the price of seven cows. Shortly after making this arrangement Gor passed away peacefully in his sleep and was honored with a great feast. Kor was still mourning his father�s death when he learned that Eretil had broken his promise to marry his daughter Thesna to Kor. He was instead planning to marry her to Agaric, chieftain of the Terveri.

Kor gathered his warriors and led a raid into the great hall of the Franes, killing many as they slept. Eretil awoke in time to meet Kor in battle, but the old chieftain was brutally defeated by the powerful Kor.

Kor brought Thesna and Ilva, the widow of Eretil, out naked before his warriors. Kor boldly declared that he was �Kor, Taker of Widows''. He and his warriors plundered all the gold and treasures of the Franes, including their livestock and returned to their homes rich in the spoils of their conquest.

When the Terveri heard news of this conquest, Agaric came to confront Kor. He entered Kor�s great hall demanding that he release Thesna and give her to him as Eretil had promised. He demanded that Kor release Ilva as well. In response, Kor had his warriors butcher Agaric's men. Kor�s warriors disarmed him and restrained Agaric. Kor then told Agaric that he was going to �cut off his head and drink from his skull� as he decapitated him with one fell swing of his greataxe.

Kor rallied his warriors and plundered the Terveri of their women, treasures and livestock. The few Terveri warriors that remained to resist them were disheartened by the loss of their chieftain and surrendered. Kor had each man executed and razed the Terveri village to the ground.

Kor, now confident in his supremacy, adopted an ox skull mask as a symbol of his strength. He pillaged neighboring tribes forcing them to pay tribute to treasure, livestock and daughters.

Kor, Taker of Widows, is now one of the most feared chieftains in the Barbarian Lands known far and wide for his strength and brutality."

Here is Chat GPT after I fed it the story I wrote and writing a scene in which Kor explains his life philosophy:

In the dimly lit war tent, the air thick with the scent of sweat and the faint remnants of battle, Kor sat upon a makeshift throne, his imposing figure enhanced by the eerie glow of torchlight. His ox skull mask adorned his face, casting a haunting shadow upon his features. Surrounding him were his trusted lieutenants, battle-scarred warriors who had witnessed the extent of his ruthlessness and power.
Kor: (in a low, gravelly voice) Brothers, listen well, for I shall share with you the essence of my life's philosophy.
The warriors leaned in, their eyes fixed on Kor, their anticipation palpable.
Kor: (his gaze sweeping across his audience) We are the Agarians, forged in blood and fire. We are the embodiment of strength, unyielding in the face of adversity. We conquer and we dominate, for that is our nature. We take what is rightfully ours, and we crush all who stand in our way.
The warriors nodded in agreement, their expressions reflecting a shared understanding of their chieftain's words.
Kor: (his voice growing in intensity) I am Kor, Taker of Widows, for I have witnessed the frailty of existence, the fleeting nature of life itself. I have seen fathers fall to my blade, leaving their wives to mourn. I have heard the lamentations of widows and the wails of children left fatherless. But in their weakness lies our strength. We exploit their vulnerability, for it is the way of our people.
His words hung heavy in the air, a chilling reminder of the brutality that defined their existence.
Kor: (his tone filled with conviction) We live by the code of power, by the law of the strong. We seize what we desire, mercilessly. We take the women, the treasures, the very breath from the lungs of our enemies. In doing so, we forge our destiny, shaping the world to our liking.
The flickering torchlight danced upon Kor's mask, casting flickering shadows that mirrored the darkness in his words.
Kor: (raising his greataxe high) My brothers, we are the harbingers of chaos, the masters of destruction. Our legacy will be written in the blood spilled and the empires we crumble. We shall strike fear into the hearts of all who dare oppose us. Let our enemies cower, let them tremble before the might of the Agarians!
The warriors rose to their feet, their weapons clashing together in a cacophony of affirmation. The air filled with a resounding roar, the battle cry of a tribe consumed by a thirst for power and conquest.
Kor: (his voice thundering) We are the wolves among sheep, the untamed force that shall reign supreme. Our strength knows no bounds, our will unyielding. Together, we shall carve a path of bloodshed and dominance across these lands, leaving none to challenge our reign!
The war tent echoed with the fervor of their conviction, as the warriors bellowed in unison, their loyalty to Kor unshakable. In that moment, they were united by a shared belief in the merciless pursuit of power, pledging their lives to the legacy of Kor, Taker of Widows, the feared and revered chieftain of the Agarians.

....I think it did a pretty cool job, that is actually in line with what I wrote for the character!

# ? Jun 26, 2023 04:52

Megaman's Jockstrap: Jul 16, 2000; What a horrible thread to have a post.

Uh, who is that artist? They loving rock.

# ? Jun 26, 2023 06:09

BrainDance: May 8, 2007; Disco all night long!

My DM lets me use AI but it's because I'm usually spending a bunch of time on it and using stuff from my own models I trained so it's really not just dumps of garbage from chatgpt.

The exception being the one time the new DM in one of my games was like "hey so you guys should probably make a new character cuz you're probably gonna die tonight" about 2 hours before the game started. Even then I really tried to tell ChatGPT what I wanted from it, I had a story in my head. And it refused because it was disrespectful to the dead and death is a serious issue etc etc.

# ? Jun 26, 2023 06:14

trapstar: Jun 30, 2012; Yo tengo un par de ideas.

Megaman's Jockstrap posted:

Uh, who is that artist? They loving rock.

Pedro Silva
Here is their commission info: https://www.canva.com/design/DAEteQ_YFa0/Gh9McMg7vY1W8Gp_V4P1Kw/view
And here's there artstation: https://pedrosilvah.artstation.com/

# ? Jun 26, 2023 07:28

Bucnasti: Aug 14, 2012; I'll Fetch My Sarcasm Robes

My favorite thing to do with ChatGPT is to give it some rough text I wrote and then have it re-write it in the voice of a specific writer or character.
In my current campaign my players are exploring a wizards lair and there are multiple libraries, so I used this trick to make up a bunch of books, some are nonsense just to be flavorful but some have important hints for the dungeon and some have worldbuilding information.
The most recent one I did was about geomancy and after writing a rough bit about it including the things I wanted to convey to the players I fed it into ChatGPT and told it to re-write it in the style of Gary Gygax.

quote:

Geomancy, verily, is the esoteric art of wielding and optimizing the boundless energies that flow through the mystical Ley Lines. Its purpose lies in the creation of structures and artifacts, cleverly designed to tap into the vast reservoirs of raw power that permeate the very fabric of our world.

Yet, the rules of Geomancy, I must admit, are labyrinthine and intricate, even for the most discerning of initiates. Oftentimes, they give rise to mazes of bewildering complexity, wherein rooms and spaces intertwine in a manner that seems devoid of rhyme or reason, baffling the uninitiated observer.

Alas, Geomancy carries its share of perils. Should a geomancer's calculations falter, resulting in an imprecise construction or a structure left to wither through neglect, the currents of energy drawn from the ley lines may deviate from their intended course, manifesting as capricious and unforeseen effects. Indeed, such manifestations have been known to assume curious forms and exert their influence upon those who dwell in their proximity.

A bit flowery even for Gygax, but perfect for my needs.

# ? Jun 26, 2023 08:44

feverish and oversexed: Mar 9, 2007; I LOVE the galley!

Giving it a "outline" and "structure" is honestly the only way to go without it being generic poo. After you've given it some "coaching" or prompting or whatever it does a pretty good job at maintaining that narration as well (I'm always referring to chatgpt4, exclusively).

Also brainstorming is great with it, just as a sounding board for my own ideas has really helped.

I've used it to playtest new mechanics as well, where I am the DM and it is the "player" with middling success, needed some direction to understand my mechanics.

# ? Jun 26, 2023 17:51

Adbot: ADBOT LOVES YOU

# ? May 22, 2024 18:14

Abhorrence: Feb 5, 2010; A love that crushes like a mace.

BrainDance posted:

My DM lets me use AI but it's because I'm usually spending a bunch of time on it and using stuff from my own models I trained so it's really not just dumps of garbage from chatgpt.

The exception being the one time the new DM in one of my games was like "hey so you guys should probably make a new character cuz you're probably gonna die tonight" about 2 hours before the game started. Even then I really tried to tell ChatGPT what I wanted from it, I had a story in my head. And it refused because it was disrespectful to the dead and death is a serious issue etc etc.

The trick, I have found is to A: don't ask it to do something it has refused to do later in the thread, instead edit the question that made it refuse, and B: explicitly ask it to include a content warning. For whatever reason, it is more willing to write about stuff if it gives a content warning.

# ? Jun 26, 2023 18:27

The Something Awful Forums > Discussion > Traditional Games > AI tools for traditional games

«‹›14 »