[Tease AI Java] [NEED YOUR HELP] Modernizing Tease personalities

[izura]

just cleaned a bit the look of your doc, nothing much but i think it look better that way

[ski23]

just cleaned a bit the look of your doc, nothing much but i think it look better that way

Yes I just saw and accepted your changes. It looks much better now. I copy pasted from word into google docs and it lost the formatting.

[ski23]

This is kind of what I was trying to do when I went from Miss Blue 1.7 to 2.0. I've tried to deal with most of the problems you list at some point during that process, not always succesfully. I think some of these issues are inherent to TAI (I assume it's the same in Java). It looks like a chat simulator (so you want continuity and interactivity), but functions for the most part as a random tease generator. These 2 things aren't always compatible and each personality has to strike a balance, which is gets more complicated the more adjustable to user preference you want to make it. For example, solution 1A works, but those long modules will either be very bland and samey to accommodate every user, or contain lot of more specific/interesting bits behind filters which will hurt the overall flow of the module.

Right now I’ve been leaning toward the missblue sort of style with responses that get triggered with active flags.

I have all kinds of thoughts about this one. Flags is one way, though a problem is that they have to be deleted. Another is creating windows of opportunity for responses using SetDate, combined with ResponseYes/ResponseNo to simulate conversation during stroking cycles. For example, domme: "I think I should make you edge again", you: "you're so cruel", domme: "How is that cruel, don't you like edging?", you: "No", domme: "Oh come on, we both now that's not true". The conversation would take a different direction if instead of "you're so cruel" you'd say "I would like that".

But I guess it mostly comes down to "more content".

This is true. Almost every solution eventually requires (a lot) more content.

1. I can definitely use other personalities content to some degree but likely a lot will have to be modified to fit in with this newer design. I definitely will utilize previously written content though.

In my view a lot of what makes a personality have 'continuity' comes down to tone. As much as I like HoT, one thing that really bothers me is how the domme can go from girl-next-door to whip-wielding-dominatrix in an instant. It breaks the suspension of disbelief.

Yea, in my opinion, some of the issues are more inevitable with TAI and more preventable in TAJ due to more developer freedom.
You are right that longer modules can cause it to get bland but I was more thinking of long modules in certain specific areas with not many filters behind them. So maybe there would be a long module for anal and then several filters for more specific content based on more niche anal fetishes.
For the flags one, it's not really a concern about deleting the flags because it wouldn't use flags as they are traditionally in TAI. You can use temporary variables in TAJ that expire after a session but, for this in particular, I was leaning towards a series of state machines.
Yes, I will need to make sure any content I add has the same tone or modify the content to have the same tone and feel to the other modules.

[holaba_be]

This has been something I've long been trying to do with Miss Blue and I think it's probably the most important issue: how do you create a system that allows for sessions that flow naturally without relying on long linear modules. My solution so far is to set a 'theme' for the session early on (though not all sessions will be 'themed'). A flag is set to indicate this, which will trigger scripts (or give a higher chance to trigger certain scripts) each cycle. These scripts are chosen at random from a seperate folder, so although the session has a 'theme', it's still somewhat unpredictable.

Ha, I always wondered if such a themed session was just a coincidence or if it was really intentional. Good idea!

[ski23]

Yea, in my opinion, some of the issues are more inevitable with TAI and more preventable in TAJ due to more developer freedom.
You are right that longer modules can cause it to get bland but I was more thinking of long modules in certain specific areas with not many filters behind them. So maybe there would be a long module for anal and then several filters for more specific content based on more niche anal fetishes.
For the flags one, it's not really a concern about deleting the flags because it wouldn't use flags as they are traditionally in TAI. You can use temporary variables in TAJ that expire after a session but, for this in particular, I was leaning towards a series of state machines.
Yes, I will need to make sure any content I add has the same tone or modify the content to have the same tone and feel to the other modules.

Flags work the same way in TAI and variables can be made to expire too. Looking down your list of 13 issues I don't see anything that couldn't be solved using TAI in its current state, although some things can be a bit tricky (mainly issues 6 and 9). I'm sure TAIJ is going to be much more refined and stable than TAI, which definitely still has its problems. My point is that in terms of writing a personality I think the bigger challenges are in finding the right balance between different systems and getting it all to work together. Having said that, I'm really curious to see what solutions you come up with (especially for issues 1,4 and 5). Could you explain how you would implement the state machines?

A better history system would also help the previous issue. Knowing what's been going on during the previous sessions would allow for longer "story lines". For example, the Domme might have her heart set on forcing CEI, so she keeps denying the player at the end of every session until he agrees to it. Currently personalities have a check at the beginning of each session to ask about something that might have been asked at the end of the previous one. Expanding on this might make it far more immersing. It would probably also increase variety when there is a better history of which module has already been done. Since content is not endless, some of it is bound to repeat. But it should try to repeat only when each other module has already been done (unless it's some highly specific module that should be rare by design).

This has been something I've long been trying to do with Miss Blue and I think it's probably the most important issue: how do you create a system that allows for sessions that flow naturally without relying on long linear modules. My solution so far is to set a 'theme' for the session early on (though not all sessions will be 'themed'). A flag is set to indicate this, which will trigger scripts (or give a higher chance to trigger certain scripts) each cycle. These scripts are chosen at random from a seperate folder, so although the session has a 'theme', it's still somewhat unpredictable.

On your first point, yes flags function similarly. However, the big difference is that TAJ allows for an entirely custom flow of the structure while TAI defined the structure for you. Also, every aspect of TAJ is customizable since its using javascript with java through nashorn while there were limitations in TAI. I agree that balancing systems and making them work together is a challenge and something I'm actively thinking about for design. I realize that there will always be a balance and it isn't possible to make every area perfect for all users. I would say that my highest priorities at the moment would be (highest first): 2, 1, 12, 9, 10, and 5 because these are, in my opinion, the biggest reason to drive users away from a personality. For the state machines, I was thinking of several state machines that can store different states and then things like user responses, mood, and whatnot in a specific module. So a great example of this is your theme example you mentioned. I had a similar idea with the "themes" as you call them. Although sometimes it might be nice to have a theme session, I was thinking of having "themes" or really more of like "moods" that can change within a session so if the session is an hour long then maybe the first 20 min could be fetish related and then 10 min of something else, etc. This is one example where a state machine would be useful along with others like stroking/edging/action state, etc. One thing that I have planned is that every module will be associated with one or more categories. Then, certain modules can be selected more specifically and you can handle responses that happen in a certain category of module for different state machine states.

I touched a bit already on the themes. However, I would like to mention that the history/statistics system is already in development and developers will have basically any information they could possibly want about what happened previously in the current or previous sessions. Also, I'm providing an easy way to access and sort through this information and get just the information that the personality might want to make a decision.

[ski23]

So, can someone update me. Is tease AI Java dead. I only see one personality and that started dev in 2018.

Is there another program that has taken on to be the spiritual successor of Tease AI as a series.

The TAI (original) doesn't seem to be in active development, and personalities like miss blue have not been updated for close to a year.

Basically, could someone give me a rundown of the most current and complete projects, if there are any?

Tease AI Java is not dead. It is the most actively developed software in these forums. I have a personality in the works and GodDragon recently released a beta version of spicy as well as the port of bound to her will I did a while back (4 total personalities). You need to understand that these personalities take a LOT of time to make and to make them decent. For some weeks, I've spent over 40 hours a week working on this stuff. Tease AI Java is much cleaner and less buggy under the hood than Tease AI and in my view is the future of this software. However, we have lives too and can't always work on this stuff nonstop for free.

[ski23]

Nice to have an update thanks, what personalities are out of Beta?

Mischevious and Bound to her will are out of beta and spicy is in open beta.

[avatarbr]

So, can someone update me. Is tease AI Java dead. I only see one personality and that started dev in 2018.

Is there another program that has taken on to be the spiritual successor of Tease AI as a series.

The TAI (original) doesn't seem to be in active development, and personalities like miss blue have not been updated for close to a year.

Basically, could someone give me a rundown of the most current and complete projects, if there are any?

Also, original TAI is active again, with 1885 back. But he is posting only in the Fury topic here: viewtopic.php?f=26&t=22426. I think he will update the original topic some time in the future (he already incorporated some of the unoficial patchs in the new version)

[GodDragon]

So, can someone update me. Is tease AI Java dead. I only see one personality and that started dev in 2018.

Is there another program that has taken on to be the spiritual successor of Tease AI as a series.

The TAI (original) doesn't seem to be in active development, and personalities like miss blue have not been updated for close to a year.

Basically, could someone give me a rundown of the most current and complete projects, if there are any?

Also, original TAI is active again, with 1885 back. But he is posting only in the Fury topic here: viewtopic.php?f=26&t=22426. I think he will update the original topic some time in the future (he already incorporated some of the unoficial patchs in the new version)

True. Thing is even with TAI being stable and less clunky it sadly just lacks the possibilities and clean code for personalities. Like you need to add basically anything that you want to use to TAI to make it work inside a personality while with TAJ you can do anything without modifying the base at all. I have done so many more advanced things inside a personality without needing any addition to the code base. Like people wanted to create custom guis using the personality, more control over the media player etc. and all of that has been done already. I really like TAI but for my style of coding and ideas it's too limited and too complicated to implement more complicated ideas.
Not to mention TAJ has auto updates for personalities and the executable itself etc.

[desertfox]

Was just reading through the thread here and always interesting discussions with tease ai stuff. When it first came out I tried to get into it, but in the end it felt mostly like a random tease generator. It was just an ai saying stroke, and wouldn't even put a picture up on screen sometimes. This may be my personality as well, just being controlled isn't enough on its own to really do anything for me, so I chalked it up to just not being my thing.

That said it's still an interesting problem. I know I dropped the ball with my guideme maze, but i did spend a lot of hours on it, which left me thinking a lot about creating semi random systems. For one as an author, I never did get to play my own creation, so you can say that the longer and more committed set up can sometimes leave you lacking desire to experience the content you've made.

I found that after I had set up all the building blocks for the system working, it came down to content generation, and it's fatiguing and difficult to come up with so much of it. It's why procedural generation is so appealing as content creators, and consumers, like it's going to solve the problem of lack of content.

The problem is that as humans we are super good at recognizing patterns, so we in turn see how the system works, and categorize it the same as content we've already noticed. A good example of this might be minecraft terrain generation, if you've seen one minecraft desert, you've seen all minecraft deserts, even though no two are exactly the same, they still look the same, it might as well be the same. So in procedural generations you go look for where did the procgen put XZY feature I need to find, there isn't really any sense of discovery except maybe some random luck that there's 3 of what you need in one spot.

I know hybrid approaches work somewhat well, where you define an authored structure, and use randomness in specific places to add variety, but I think it is hugely important to get the player to the same point regardless of the paths they took to get there. If you don't close the branches back down that you open up it gets wildly out of control and unmanageable to have any narrative experience.

I think that to solve some of the problems in the doc you have to define and limit what the AI and players can do, and standardize them in the API.

An example for player actions
edging
stroking
cloths pin
ball slaps
denial
ruin
orgasm

An example dom action might be 'light punish'. As a user I define the actions I feel are a 'light punish' being only able to pick from the preapproved action list. So I might decide 1 to 4 edges are my light punish. Someone else might decide 5 ball slaps are a light punish. You can limit what player actions go with AI actions, so you can't use orgasm as a light punish, for example. But ultimatly it is up to me as a user to define what and how I like things and their duration.

Then as a modual author, I can decide when the AI will use a light punish action, and know that the user is happy with what is happening. This also means that you can't specifically call out 'how do you like edging' style dialog, but have to be more vague like 'how do you like your punishment?'.

I know that teaseAI can be written like this if users modify the scripts, but I think the difference with it built straight into the program and hardcode would make it easier for users to set up, and unify authorship.

I think trying to include everything is what's going to keep leaving tease AI's as script/programming heavy and light on content, when you want to get content creation as the only obstacle in front of people.

So again just some ideas throwing around here and food for thought.

[GodDragon]

Was just reading through the thread here and always interesting discussions with tease ai stuff. When it first came out I tried to get into it, but in the end it felt mostly like a random tease generator. It was just an ai saying stroke, and wouldn't even put a picture up on screen sometimes. This may be my personality as well, just being controlled isn't enough on its own to really do anything for me, so I chalked it up to just not being my thing.

That said it's still an interesting problem. I know I dropped the ball with my guideme maze, but i did spend a lot of hours on it, which left me thinking a lot about creating semi random systems. For one as an author, I never did get to play my own creation, so you can say that the longer and more committed set up can sometimes leave you lacking desire to experience the content you've made.

I found that after I had set up all the building blocks for the system working, it came down to content generation, and it's fatiguing and difficult to come up with so much of it. It's why procedural generation is so appealing as content creators, and consumers, like it's going to solve the problem of lack of content.

The problem is that as humans we are super good at recognizing patterns, so we in turn see how the system works, and categorize it the same as content we've already noticed. A good example of this might be minecraft terrain generation, if you've seen one minecraft desert, you've seen all minecraft deserts, even though no two are exactly the same, they still look the same, it might as well be the same. So in procedural generations you go look for where did the procgen put XZY feature I need to find, there isn't really any sense of discovery except maybe some random luck that there's 3 of what you need in one spot.

I know hybrid approaches work somewhat well, where you define an authored structure, and use randomness in specific places to add variety, but I think it is hugely important to get the player to the same point regardless of the paths they took to get there. If you don't close the branches back down that you open up it gets wildly out of control and unmanageable to have any narrative experience.

I think that to solve some of the problems in the doc you have to define and limit what the AI and players can do, and standardize them in the API.

An example for player actions
edging
stroking
cloths pin
ball slaps
denial
ruin
orgasm

An example dom action might be 'light punish'. As a user I define the actions I feel are a 'light punish' being only able to pick from the preapproved action list. So I might decide 1 to 4 edges are my light punish. Someone else might decide 5 ball slaps are a light punish. You can limit what player actions go with AI actions, so you can't use orgasm as a light punish, for example. But ultimatly it is up to me as a user to define what and how I like things and their duration.

Then as a modual author, I can decide when the AI will use a light punish action, and know that the user is happy with what is happening. This also means that you can't specifically call out 'how do you like edging' style dialog, but have to be more vague like 'how do you like your punishment?'.

I know that teaseAI can be written like this if users modify the scripts, but I think the difference with it built straight into the program and hardcode would make it easier for users to set up, and unify authorship.

I think trying to include everything is what's going to keep leaving tease AI's as script/programming heavy and light on content, when you want to get content creation as the only obstacle in front of people.

So again just some ideas throwing around here and food for thought.

Yes indeed, as a human you are able to spot patterns quite easily. It is quite hard to make a personality realistic and random enough to make it interesting.

However a light punishment is up to the top to decide. Obviously the top will usually know about the limits and interests of the sub and pick a punishment that he/she sees fit however it is not up to the sub to determine 5 slaps as a light punishment. So it should be less like: "What do you consider x" and more like the AI finding the different limits etc. of the sub by playing around and trying stuff instead of asking for it. Like what real world top would be ask a sub about a specific number of hits for a punishment unless it is part of some evil plan or game.