SakeTami
manyeyedhydra
manyeyedhydra

patreon


Generative AI in House of Hellish Harlots

Given that the controversy over generative AI has ramped up, I thought I'd put out a post with my own thoughts, along with a declaration of where and why I use it for the House of Hellish Harlots game project.

Personally, I regard it as another tool in the toolkit. A human artist is nearly always better (and they will draw boobies for you without having to worry about "Content Moderated" censors), but most projects do not have infinite money, and so AI joins the other tools to bridge those gaps when the resources aren't there.

If you've been following me long enough, you'll have seen this in the evolution of my book covers. The early ones were composited from clip art and looked pretty rough. Then I used early generative AI for replacement covers back in 2022. And now I think the sales are there to justify hiring a human artist to redo all the old book covers, which is what I'm doing.

The same is also true of HoHH. It was originally text-only. AI art opened up the possibility of adding pictures, which then justified the hiring of an artist to make a better job of the harlot pinups. In this case AI art actually resulted in a lot of commissions and work being generated for a human artist!

Because I've never thought of AI as a replacement, the increasingly ill-tempered discourse about it online has caught me by surprise. My own opinion is if you can afford and justify the human artist, hire the human artist. If the resources aren't there or you're trying to do something that isn't feasible to get humans to do, use the machine.

Some people have differing and strong opinions on this, so in the interests of full disclosure, here is a full breakdown of where and why generative AI has been used within HoHH.

My overall aim has been to make HoHH be the best it can. However, it's a niche concept (fully XXX-rated adult game featuring femdom monster girls) and my budget is miniscule. Originally, I set aside about £10K and I've used about £7K of that on harlot art commissions. The subs here have helped a lot, and so I'm probably going to set aside a further £10K and give the harlots proper H-scene art as well. It's stretching my finances a little, but I think I have something good and I'd like to end up with something similar to classic monster girl games like Monmusu Quest or Shrift.

Now for that breakdown.

WRITING
I don't use AI in any of the writing. That is my primary skillset, so I don't feel I need machine help there.

CODING
I don't use AI here either.

Mainly because I haven't figured out a way to make it helpful. This is my secondary skillset, so I'm not averse to bringing a machine in, if I figure out a way to make it work. Currently, the project resources are not there to hire a specialist and I haven't figured out how to incorporate AI, so I will continue to clumsily hack the html and CSS myself.

ART – HARLOT ART (PINUPS, PORTRAITS and H-SCENES)
This is outside my skillset, but I also think this is the primary art for the project. People are here for the sexy harlots, so the limited budget should be focused on them.

Currently, all the harlot art is commissioned from human artists and I will continue to do this.

ART – OTHER NPCs
This covers the male NPCs such as the Barman, Portly Gentleman, and the car NPCs that arrive with the player. In H-RPGs the non-essential NPCs are usually depicted as empty silhouettes. This seems a good use case for AI to improve on that.

Given the choice between commissioning proper art of non-essential NPCs (i.e. ones the player can't have sex with) or harlot sex scenes, I'd rather use the limited funds to commission more sex scenes.

Non-essential NPC art will be AI-generated.

ART – CHARMS, POTIONS, AILMENT
This covers all the various icons you might see in an RPG. These fall in the middle ground of being somewhere where you can get away with generic art, but having custom original art is probably better.

For me, the budget is so tight it comes down to how serviceable the AI art is.

The charm cards and icons came out surprisingly well and consistent. I could replace them with human-drawn art, but I think that money would be better spent commissioning more H-scenes instead.

The potions and ailment icons are a little rougher. AI wasn't so useful there, so I might need to get some human help in out of necessity.

My rule for here is use AI if it looks good, but I might have to divert some money from the sexy harlot art funds to fix the things AI is not so good at (consistency, being allowed to draw nice boobies, etc).

ART – BACKGROUNDS
Priorities come into play here. Backgrounds can also be relatively more expensive to commission depending on level of detail. Most H-RPGs seem to use generic art packs here, which makes sense. People play those games for the sexy monster girl sprites, not look at the background scenery. AI is a slightly better generic background creator than re-using the same art that shows up in multiple small H-RPGs, and I think this is where it will see a lot of use for very small RPG projects.

AI also opens up additional possibilities. Using it, I can give each harlot a custom room background. This wouldn't be feasible for a project of this scale to commission from human artists. It's feasible to get the machine to do this, especially as most of the art will be obscured and not visible anyway.

Letting AI handle the backgrounds frees up more funds for sexy harlot art, so I will continue to use AI for the background art.

ANIMATION
Animation is out of scope. HoHH has over 80 harlots and the final version will have around 100+. Animation is extremely expensive and time-consuming. Hiring humans to animate 100+ characters just isn't feasible for a project of this scale.

AI might help here, depending on whether the current annoying content moderation measures are lifted.

Some of the AI animation added to marwmellow's art (with his permission) has looked promising. Not being currently able to animate the more explicit art does hamper its potential usage.

It's something I'm keeping an eye on, but for the moment adding animation is out of scope.

SOUND – MUSIC, FX
Music and voice acting are even further out of my wheelhouse. Currently there is no sound at all in HoHH. I feel if I'm going to release it as a commercial game, I will need to put something there.

Because the project funds are strictly limited and focused on the art, my options were to look at open-source files (ideally not the same music themes that seem to show up in every H-RPG!). If AI adds to those options, I might consider it, but it's not something I know much about, so I'll leave that up in the air for now.

SOUND – VA
There are going to be 100+ characters in HoHH. Voice acting for all of them simply isn’t feasible for a project of this scale.

I'm not sure AI will do anything to help there either. I did get snippets of voices and introductions when experimenting with grok, but the voices mostly still have that 'uncanny valley' problem and keeping them consistent for a particular character will probably be a nightmare.

I'm not planning for HoHH to have any form of VA at the moment.

And I think that covers all areas.

I’ll repeat my opinion that AI is best used as a fill-in tool for when you don't have the resources or it's not feasible to hire a human creator. Given infinite funds, I'd always hire humans. In the real world we don't have infinite money, so trade-offs need to be made. If I can only fund one, I'm going to fund the art of a sexy succubus harlot doing sexy things rather than nice art of her bedroom. And as I want the project to be the best it can, using AI as a fill-in tool is better than 'nothing at all'.

Those are my own personal opinions. I appreciate others are going to have differing opinions on this. In the interests of openness, I think it's correct to declare what I'm using, where, and why, so people can make up their own mind on what they choose to spend their money on.

I ask people to be civil and respect others might not have the same opinions on this. The kind of performative harassment and bullying I've seen on social media on this topic will not be tolerated here.

And with that out of the way it’s time to get back to your regularly scheduled succubus smut! There is another HoHH1.5 release on the way I’m hoping to drop tomorrow or over the weekend. This adds custom bedroom backgrounds for all the harlots as well as some minor fixes.

- manyeyedhydra

Generative AI in House of Hellish Harlots Generative AI in House of Hellish Harlots

Comments

I think the discussion surrounding AI is overblown. When used correctly, the quality of AI art is incredible now. People complain that it relies on existing art, but that is simply how AI works. Unless we had self-aware AI (which we don't), it has to learn from data. As such, the entire discussion is stupid to begin with. More importantly, good GenAI outperforms almost any artist you can hire at a reasonable rate. Top tier artists, like Shiki, are simply too expensive for most people. I would rather see decent AI art than mediocre human art. Imagine getting a character like Nurse Honey in Shiki’s style—you can even mix styles to be safer regarding copyright. Bottom line: GenAI is currently better than most human art available at a reasonable price point.

Wakozi

I'll try it and see what people think. Mainly it was because I felt the bed picture was too generic given a lot of the harlot rooms are much weirder. There is no need to worry about content creep on it. Ethics aside, this is one area where AI has a good use case (it might be the best use of it and I wouldn't be surprised if a lot of small indie RPGs start using it for this). Aside from a couple where I needed to vary the prompts, for most of the pictures I just plugged in the text description from the scenario. Maybe a day or two if that to generate 36 harlots rooms. The main content creep problem is the harlot notebook. I think it will be worth it once it's in place, but with hindsight I think that was a step too far. I'll keep that in mind for the Barman and Portly Gentleman. Maybe I'll get marwmellow to redraw them in his style when I get a spare moment.

Many-Eyed Hydra

Thank you for posting about this. I was severely torn on whether to comment about this on the room background thread. Personally, I don’t like GenAI aesthetically or ethically. In it for the long haul to keep supporting the project, though in some cases e.g. the room backgrounds, I think it might be better not to have them at all than something that risks looking too generic. Also a bit worried this is the dread “content creep.” One point I’m strongly in favour of doing otherwise is having real custom drawings for the major NPCs i.e. Angelic Barman and the Portly Gentleman (car npcs less so, unless until they start showing up in side quests). They’re so prominent it would be clashing to have them in a radically different style from the other post-room characters.

CrimsonCrown


More Creators