Getting started in the wide world of adult games.

[Julie L Griffin]

Hello all! So, I've been writing erotica under a number of pennames, but well, the market has some sales issues, and more importantly, I'm getting bored with it, so I decided to start out making ren'py games, so my family can enjoy the monkey shrieks of rage coming from me as I try to learn a language. So I was wondering if y'all would take a look at my plan.

1. start out simple. I'm learning Renpy, so I'm starting out with more or less simple visual novels. Short, not a lot of changes, mainly a story in VN sense. Why?

Because on this site and others, I've noticed an immense number of games that start out with: Multiple characters! Many ends! Flexible corruption, love, and golf skill rating stats that influence the story! Complete minigam--

last developer post: June 2017.

So rather than start complex, doing stuff that I don't know how to do, I'll stick to only releasing more complex games as I learn, and yeah, the first few games will be short, but I can always go back and redo them, and I think it's better to be able to say: yeah, it's short but it's done.

2. Art assets. I'm good with Stable diffusion and baking LORAs for various characters. I might share my method on that, but artists are expensive and so I'll keep the games Ai-art until I reach a certain level of income so I can pay the artists I know up front. Minor and transition scenes are likely to stay AI art generated, with main scenes being replaced by human-sourced art.

3. Voice and sound assets...that gets into the "learning how to do this." I might, but I need to do a lot more research on how much it costs and such.

So! for this admittedly lose "plan" does anyone have any suggestions? Just note: A few people did suggest Unity to me, but from what i've seen, Renpy may be more my speed, and I don't want to work on more than one system at a time.

 

[Vanderer]

AI art? I am not sure that you can get the AI to be consistent enough...but if you can get AI to be consistent enough to do whatever you want, props to you, but I doubt it. I would use DAZ3D or one of the ILLUSION studios if I were you.

Simple fact: making art is the more time-consuming part of making a game for the huge majority of devs. But look at it that way: none of the popular games use AI.

For the rest, you got the idea. Starting simple is the best. Be aware that the more branches your story has, the more your workload will increase. If you are writing erotica, you could simply make a kinetic novel, a visual novel with no choices. Less popular, but simpler.

Renpy is IMHO the best choice for VNs, by far. It's also quite simple to learn as long as you want a simple VN and you are not an idiot.

As for the sound...I don't have any in my VN, so no help from me there.

I hope you have a good time making your Vn.

 

[Atemsiel]

Only advice really is to start. 90% of these posts talk about the plan, but very few of them ever go beyond the planning stage. Start making something to get you familiar with the process. If you don't like it, you don't have to release it, but at very least you'll learn more from that than you will from just planning. Best of luck to you.

 

[Julie L Griffin]

Well, I've already gotten to the opening screen, the "are you 18" screen and... a blank screen. So... Progress!

 

[Glib_Gentleman]

AI art? I am not sure that you can get the AI to be consistent enough...but if you can get AI to be consistent enough to do whatever you want, props to you, but I doubt it. I would use DAZ3D or one of the ILLUSION studios if I were you.

Innovations in the past two or three months have made it much easier to get consistent characters from the AI.

Simple fact: making art is the more time-consuming part of making a game for the huge majority of devs. But look at it that way: none of the popular games use AI.

While this is true, it is also a bit . . . premature to say? The AI artwork has only gotten good enough in the past 2-3 months to be useful for making games. The truth of the matter is that any game using AI art hasn't had long enough to get popular. And as the tech improves, some game is going to be the first game to get popular using AI art.

If OP is comfortable with AI art, and I'd say if he is making LORAs, he is, that is going to be far faster to generate good art with than learning DAZ3D or Illusion from scratch.

 

[Julie L Griffin]

yeah, i mean, for those who don't know, a LORA is a way of training a specific character, style, or clothing style, far more quickly than you normally can do if you're creating an entirely new checktpoint and you can mix and match. Along with things like Controlnet, AI art has become magnitudes more doable in the last few months.

THAT BEING SAID, it still has a lot of problems when compared to an artist. Even the best LORA is going to see issues with consistancy in clothing and interactions between characters, even with control net, can still be an issue. I'll be honest, I think that a fully "done" AI drawing, one set to your own desires rather than 'eh, I've generated a thousand, this is good enough" is highly likely when you take inpainting, outpainting, etc, into account, to be about as effort intensive as a regular artist (meanwhile, an actual artist using AI art can just zip ahead).

My plan is to start with characters just generated from a base model. There probably won't be a lot of consistancy, but it'll give me an idea of how the first VN will work, then make some decisions on what the final look will be like and train really specific LORAs (probably several per character for different clothing). Dunno if I'll show the initial model off because there won't be a lot of consistancy.

Also, I just had my first fatal error in Renpy!

 

[Atemsiel]

Also, I just had my first fatal error in Renpy!

One good thing about Ren'Py is that due to how simple it is, there's plenty of people who can help. Depending on what the issue is, even someone who's only been using it for a few weeks may potentially be able tell you the issue. People help out in the programming section here, or you could post it on the Ren'Py subreddit, and there's the offical Ren'Py forums which is full of the nerdiest Ren'Py people possible. (Though you have to get a code to join, and I can't quite remember how to do that.)

Generally though, when you're first starting, fatal errors in Ren'Py are much less of an "Oh Shit" moment than with most other platforms. Most of the time at least. Reach out in the programming forums here if you find something you really can't fix, but be sure to include all the relevant information. You'd be surprised by how often people say stuff like "Hi guys, I got an error, I don't know what to do, please help", but don't give any further information.

I can't speak for everyone, but I think most people like to help out where they can. The subreddit especially since that's basically the entire point of it, and it has a good mix of experienced devs, and totally new people. (Which means the community there is quite used to answering questions of varying levels of complexity.) Same rules for actually explaining the issue apply there too though.

 

[InteriorCrocodile]

So rather than start complex, doing stuff that I don't know how to do, I'll stick to only releasing more complex games as I learn, and yeah, the first few games will be short, but I can always go back and redo them, and I think it's better to be able to say: yeah, it's short but it's done.

Congratulations, you are smarter than 80% of the people I work with!

I have been going to college to get my Bachelors of Science in Game Design for about 3 years now and I can tell you what roughly $80,000 worth of tuition has gotten me so far.
(GI Bill covering the cost because there's no way in hell I'm paying that much for art school)

40% is documentation and planning
35% is implementation
25% is putting out fires.

These aren't based on how long it takes to do each, but their importance. You essentially have to have the entire game done (or as close as you can get because it will change by the end of production) on paper before you can actually put the time into making it virtually. Then you have to get the right people to fill in the roles that are required. Once production begins, you have to manage your time and energy to make sure that you get the game done within the set amount of time you are given. Then pray all the problems that will come up are small enough to not delay the game for a significant amount of time.

This isn't how everyone does it, but this is 100% how professional teams do it.

For you that would probably translate to writing out the entire story out first and making a map of all the choices the player can make. This way when you bring in other people, all you have to do is point and say, "this is what I want" and they can tell you either yes or I can do this other thing instead.Planning can also help keep you focused and prevent some problems before they begin. Sometimes an idea is solid in your head, but you only find holes in it when you put it down on paper.

I can talk about this stuff forever so I will leave you with this. The best way to learn game development is to see other people's games and how they made it. This is a post-mortem for Deus Ex and I put a time stamp for the things that went right and wrong in the project. Notice how much of it is just planning out things ahead of time.

TLDR; Watch the video.
----

 

[crnisl]

AI art? I am not sure that you can get the AI to be consistent enough...but if you can get AI to be consistent enough to do whatever you want, props to you, but I doubt it. I would use DAZ3D or one of the ILLUSION studios if I were you.

What about combining the approaches?
Nowadays, you can generate characters and poses to your liking in 3d, then use img2img/controlnet with sd models like

You must be registered to see the links

, and turn the 3d to cool art, and so on.

 

[Julie L Griffin]

To talk about this, i'm starting to build some of the characters.



Now this isn't the finished version. What this is is the start of creating a number of images, which will then be "averaged out' to get a singular look that you can define in the game. So, you generate these, a metric shitton--usually at least 200 to start with, and then prune the ones that either don't look like your image, or say, have hands emerging from their stomach, before you use the remaining images, train them in a LORA and then use that LORa to generate your final creation which will be sort of the "averaged out" version of all the selected images. The reason for the gray background is that you don't want a lot of extraneous detail that the drawing program might latch onto.

 

[Julie L Griffin]

I've learned an important lesson, finally explaining why my college professor back in the days of failing FORTRAN would flunk anyone who didn't comment their code, even if it worked.
comments are not just your friend, they're your lifesaver, because they keep you on track. I have learned this, and not because I just ended up going down a rabbit hole and having to chuck three hours work out.

 

[M$hot]

One thing I found useful when dealing with code is sorting the code into categories, so like all code that relates to size, location, function etc. of all different buttons is under 1 header, all code relating to layout, all code relating to stats etc. Then I'd give each header a 4 digit hash (Like 4GeP - Buttons) and at the top I wrote "Button section = 4GeP" and for all sections. Once code gets long and complex, it's very handy to have it both organized, and the hash essentially lets you flip directly to the right page via Ctrl+F.

Comments are absolutely useful, but if you are talking about 20 different subjects on the same page you will still risk getting lost. But I have to admit, I don't write much in Ren'Py so I don't know how feasible it is for your desired game to sort it all out.

One last tip that you may have already thought of, and if so, I apologize for the redundancy. But you can make a dead node that you stuff with all the info (E.g: 'You are using this font, image sizes are this, my ToDo list is this, my story board is vaguely this') so you can stay within the program. Sometimes having to Alt+Tab is not a good thing, and it's handy to have it within the same program.

Either way, you seem to have a good handle on what and how, sticking to small scale etc. Good luck!

 

[Julie L Griffin]

So, I've finished enough and haven't run out into the night screaming so a bit on my first game, which will include 3d immersive graphics, a VR system, a LLM based smart chatbot...

Nah. It's "The Landlady" and is a very simple story of a guy, his blond milf landlady...and a hypnosis book he finds.

I'm doing it this way because 1. two characters is simple. very simple, but very expandable if it works, IE, if people like it, it should be easy enough to add more rooms/characters.
2. Practice with variables, in this case, time (day and date), and personality variable (hypnotized true/false, submission, lust). That is ultimately four variables two conditional and two simple increases.

I've finished the set up, but so far it looks like a move that A. challenges me, but B. doesn't do things that are completely beyond my skill set.

 

[crnisl]

a LLM based smart chatbot...

Huh, that's how you are approaching the corruption mechanic?

 

[Julie L Griffin]

Oh, that was just me being a little joke--linking an LLM to renpy is way beyond my skill set. I was just having fun, thinking of all the "first look" trailers we used to get with games that promised teh world... and then you bought the game.
No,I'm working on a fairly conventional "If corruption >n" jump to X" style mechanic.

 

[crnisl]

Actually, I had quite successful results with coercing chatgpt to produce different events and phrases from a character, based on different grades of corruption.

chatpgt resists this, but there're ways to get what you want.

And it's hilarious, for example, making 2 version oft the same girl "before" and "after" meet each other and have a dialogue about the "backdoor activities", etc.

And linking llm to renpy is not that difficult, right? You can finetune your own llama or alpaca, and the lightest versions of them can even be run on cpu. Sure, it's not as good as chatgpt, but there're some cool possibilities for gamemaking.
Another option is selling the game as a subscription, and using a proper gpu model running on replicate.com. Or better, maybe using the capabilities of botsonic?

P.S. For inspiration about the hypnosis corruption stages, you can try Hunter's Harem. It's a html game, so reading the text the sources is quite simple.

 

[Julie L Griffin]

Well, i'm close to putting up a developer thread. spent more time working, and am about to 0.1 in the game design (art assets have yet to be finished (AI)). Learning a lot about Renpy and Python, and I figured I'd share my workflow.


More seriously, I'm learning a lot (and just finished eliminating a bunch of lines by using Call instead of jump), so hopefully in a few days the first images will be up!