Ren'Py Trap Resort [v0.4] [Wet Dreams]

[Whyhellothere]

Sir, the AI did not understand traps. These people have tits. Those are not traps.

Actually, the AI understood to truly be tricked you need to see titties attached to their chest, intelligent AI.

 

[Darkyz]

Just like the anime industry really, dev's are scared to be unique half the time and just want to make an infinite money glitch patreon game that takes 4 years.
Ima give this a try and see how I like it.

Well futanari do !

 

[SynthScribe]

Stop barking with this "ai" finally, the first powerful and large fully automated factory was recently opened in China, at night not even the lights are on there to save energy, everything is done there by AI and robots and its 100% running by self without a single human being.

I have nothing against "ai" art, in fact I like it, I will even say that it is easier and faster, "ai" is already overtaking many "creators", in a few years it will have no equal, that is a fact, you will not change it, so stop crying about "ai" because it is simply pathetic.

The blonde one remaind me of one character actually
View attachment 4665935

Indeed. For those that do AI art generation correctly, it can be a great help.

Now that said, let me preface that the following in no way applies to Trap Resort. In fact one of the very first things I noticed about the game was the attention to congruity of each character's images.

Now for the general rant for most of the other projects that use AI generated resources.

I rarely see ai included where the LLM model has been additionally trained on their specific art they decide to use.
(Picture 1 barely resembles Picture 2).

What I usually see is
1) Dev generates an AI art character (picture A).
2) Dev makes a lewd scene with another image of the character.
3) Dev limits scenes for characters, as multiple pictures of the same character that have very obvious differences is extremely jarring and removes immersion, or even worse, they don't and you're either getting images that don't even remotely fit the source text or they stick to one image for an entire scene.

For dev's I can imagine (and I definitely still fall in this scenario) that it's extremely tedious work doing custom training to any LLM, (100's and 100's of the same image, labeling them all correctly with context etc), not to mention it's extremely taxing on gpu/ram (not to mention training has an insane entry requirement to even start 32Gig+ Gpu).

 

[knightblade155]

Indeed. For those that do AI art generation correctly, it can be a great help.

Now that said, let me preface that the following in no way applies to Trap Resort. In fact one of the very first things I noticed about the game was the attention to congruity of each character's images.

Now for the general rant for most of the other projects that use AI generated resources.

I rarely see ai included where the LLM model has been additionally trained on their specific art they decide to use.
(Picture 1 barely resembles Picture 2).

What I usually see is
1) Dev generates an AI art character (picture A).
2) Dev makes a lewd scene with another image of the character.
3) Dev limits scenes for characters, as multiple pictures of the same character that have very obvious differences is extremely jarring and removes immersion, or even worse, they don't and you're either getting images that don't even remotely fit the source text or they stick to one image for an entire scene.

For dev's I can imagine (and I definitely still fall in this scenario) that it's extremely tedious work doing custom training to any LLM, (100's and 100's of the same image, labeling them all correctly with context etc), not to mention it's extremely taxing on gpu/ram (not to mention training has an insane entry requirement to even start 32Gig+ Gpu).

Hi, that's not really how training works, at all.
Tagging is completely automated and takes about 30 seconds, there are dedicated models for this. You don't need hundreds of images, you can absolutely nail a character with less than 20 (style takes a bit more). You absolutely do not need a "32 gig gpu", you can easily train loras on a 3060 or better and it will take between 30minutes - a few hours.

These also aren't generated by LLMs, which are only capable of generating text, they're generated by diffusion models.

 

[Whyhellothere]

Hi, that's not really how training works, at all.
Tagging is completely automated and takes about 30 seconds, there are dedicated models for this. You don't need hundreds of images, you can absolutely nail a character with less than 20 (style takes a bit more). You absolutely do not need a "32 gig gpu", you can easily train loras on a 3060 or better and it will take between 30minutes - a few hours.

These also aren't generated by LLMs, which are only capable of generating text, they're generated by diffusion models.

I get your point about the time to an extent. Like yea being done within a day is expected, definitely not 30 minutes with good quality and efficient steps unless you're looking for more subtle drawings.
Looking more towards the 5+ hours range especially if you're making a game and not just random posts to an AI discord.
The creators I follow didn't use automatic tagging at all instead they just used manual tagging of each image to it's following category so it understood which photo was meant for which.

 

[knightblade155]

I get your point about the time to an extent. Like yea being done within a day is expected, definitely not 30 minutes with good quality and efficient steps unless you're looking for more subtle drawings.
Looking more towards the 5+ hours range especially if you're making a game and not just random posts to an AI discord.
The creators I follow didn't use automatic tagging at all instead they just used manual tagging of each image to it's following category so it understood which photo was meant for which.

The timing thing depends on card of course, so it's hard to generalize, I was mostly just pushing back on the idea that it's a massive time commitment. After all, half an hour on a rented H100 is probably equivalent to many hours on a 3060, and a 3090 or 4090 probably falls somewhere in the middle. Even a full finetune on pony only needs overnight or the better part of a workday on a 4090.

Interesting on the tagging thing, I've always gone with the approach to tag everything you don't want, which generally works pretty well with WD-14, but I know some people take the opposite approach, which probably is easier to wrangle manually.

 

[diamond123456]

hope have some update soon

 

[cache-coherent]

I get the impression that some of this game's writing may be AI-generated, too. The writing is quite good, but sometimes there's jarring inconsistencies, especially during the sex scenes. For example, during the Caitlyn one, "I push her down onto a table, pinning her beneath me" but later "For a moment, neither of us move—just panting, tangled together in the sheets". If this is human-written, it's a strange and lazy mistake to make, but it's pretty common in generated stories.

(Also the sex scene CGs usually don't match the text, but that may have a sane explanation, like "they wrote the text before generating the CGs and couldn't figure out the right tags to get the situations to match perfectly".)

 

[Tatami Visual]

I get the impression that some of this game's writing may be AI-generated, too. The writing is quite good, but sometimes there's jarring inconsistencies, especially during the sex scenes. For example, during the Caitlyn one, "I push her down onto a table, pinning her beneath me" but later "For a moment, neither of us move—just panting, tangled together in the sheets". If this is human-written, it's a strange and lazy mistake to make, but it's pretty common in generated stories.

(Also the sex scene CGs usually don't match the text, but that may have a sane explanation, like "they wrote the text before generating the CGs and couldn't figure out the right tags to get the situations to match perfectly".)

I write them before the CGs. I write a lot of sex scenes, mostly while reading Literotica, that I keep in archives. When I need a specific scene, I usually check my drafts and see if there is one close to the scenario I want in particular or just write it from scratch. Sometimes I get tidbits from multiple ones. In that case I must have copied the ending from a different draft (which makes sense, since I hate writing post-sex dialogue, that's why I make it so short).

AI doesn't even write NSFW scenes (to my knowledge).

But thanks for pointing it out, I'll fix it for the next release.

 

[Tatami Visual]

Depends on what it means to write,but I do use a site to create all types of stories for me to enjoy,either from scratch or based on a plot related to an anime or cartoon,for example.It uses AI but it goes well if you also 'help it' with instructions now and then,basically a bigger start story for the AI to create its own story,with also some lines later on,either to continue the plot or fix stuff.Most times I need to change the AI story itself,as it tends to make errors and 'forget' details or swap character A with character B,for example.But yeah,the stories tend to become generic and repetitive,at least concerning the details,and it includes both 18+ and 18- parts.

So your phrase is mostly correct,AI doesn't write NSFW scenes,not without a human touch,yes.(as far as I know).

I have never heard of such thing, can you link me by PM?

 

[Sorreah]

I kinda asked but only got a semi-"meh" answer, so there is a scene where the MC can give one of the traps a blow job, but is there plans for a route where the traps or at least one of them can be the top? or at least a switch?

 

[Tatami Visual]

I kinda asked but only got a semi-"meh" answer, so there is a scene where the MC can give one of the traps a blow job, but is there plans for a route where the traps or at least one of them can be the top? or at least a switch?

I already answered this, yes, it will be on the next release.

 

[Sorreah]

I already answered this, yes, it will be on the next release.

Sorry, I wrote my comment way early when the page first went up and the only response I got was "You give someone a blowjob" and that's all I heard. Sorry to make you have to answer it again.

 

[zufield]

This has an error when I try to run it in Joiplay, I think it isn't using an up to date version of ren'py?

 

[cache-coherent]

This has an error when I try to run it in Joiplay, I think it isn't using an up to date version of ren'py?

It uses renpy 8.2.1 it seems; I don't know how fresh a version Joiplay needs.

 

[zipcaze]

Version 0.2 already up someone have?

 

[ffive]

How on earth are those guys "traps" when they walk around pretty much waving their dicks in the open?

Like, you'd need to be literally blind not to spot what's between their legs from the get-go.

 

[Keiken96]

Does anyone has crack for patreon?

 

[SubmissivePpl]

Will be there any oral sex?( the player doing oral sex or they)

 

[gilgamesh9]

Does anyone has crack for patreon?

Yea feel like we are missing out on alot of scenes from there. I cant get many new with jess she rarely shows up compared to the other two.