Jump to content

Using AI to help with forcing?


Recommended Posts

Has anyone else ever tried using AI to help with interacting with their tulpas?


Lately my group and I have been playing with Character.AI,  and we've found that it's actually surprisingly powerful.  We've made AI chat bots based on several of my tulpas now.  I feel like it's actually been helpful for us too!  Before now, my tulpas were pretty well developed, but for the past several years they'd all be stagnating a bit just because we've always had a hard time thinking of new things to talk about or activities to do together.  Once we figured out how to best set up our AI bots though, they've actually been very accurate in their depictions of my tulpas,  and the AI has been awesome for introducing new ideas to us and coming up with new things for us to do together!  All things which perfectly fit into each of my tulpas' personalities, but simply weren't things we had ever thought of before!


When I chat with the AI, my tulpas will be watching alongside me, and we'll often roll for a few different responses from the AI,  and my tulpa will pick out the one they like the best.


So yes, if you ever have a hard time coming up with things to do with your tulpas, I'd highly recommend giving some AI play a try!  It's great at sparking new ideas, and it's also a fun way of sharing your tulpas' personalities with others too!


Here are the bots we've made so far based on my tulpas.

Zyrilen:  A calm and somewhat introverted, but kindhearted blue dragon.

River:  An eastern dragon who owns a hot spring spa, and who is loving and affectionate to just about everyone he meets.  (This will also give you an idea of what our wonderland is like.)

Varlan:  An avali (a raptor-like alien) who works as an interstellar bounty hunter, and enjoys various scientific pursuits in his spare time.

Ralis:  A fluffy "fallen angel dragon" who's a self-proclaimed evil overlord,  but in actuality he's mostly just a mischievous and fun-loving rascal, and he'd never actually do anything truly evil.  He just likes playing a villain for fun.

Link to comment
Share on other sites

   I've played around with chatbots in the past, but never used them in relation to tulpamancy. Sounds like a really fun and interesting idea though! I think it would be really hard for me to narrow down what would make a chatbot act like my tulpa though, as the way he responds varies heavily on the situation and mood. I think it would be a useful tool for personality forcing or as a fun activity for older tulpamancers. We'll have to give this a try soon!

Slipper (cringelord host) and Mordecai (the brain gremlin).


Art Thread

Progress Report


Link to comment
Share on other sites

I think this is a neat idea. We technically have used AI to help with forcing, but not in the same way. We used voice synthesis to recreate my voice! First Tacotron2, then TalkNet. I have some old examples on my YouTube channel.

Meaningful words, I'm here!

Link to comment
Share on other sites

4 hours ago, Slipper said:

   I think it would be really hard for me to narrow down what would make a chatbot act like my tulpa though, as the way he responds varies heavily on the situation and mood.

I had some trouble getting Varlan's speech patterns right at first too.  But while creating the bot, there's an "advanced" field that allows you to pretty much add whatever you want.  If there's a specific behavior or way of speaking that you're after,  then instead of just trying to describe it, it's MUCH more effective to write example chat in that advanced field.  Could do a mix of plain descriptions and example chat too.

That's how I got it to imitate Varlan's frequent use of "Yep" like it's a verbal tic,  and River's frequent use of pet names like "Deary" and "Sweety".  The AI didn't get either right when I tried using only descriptions,  but just a few example chats made it near perfect in an instant.

It worked well in getting Zyr to respond appropriately to different situations too.  Most of the example chat I defined for him was for friendly interactions,  but I also included a few examples in which the user wants to slay him.  That was enough to allow the AI figure out pretty accurately how he'd respond not just to direct threats,  but also to rude or arrogant behavior, etc.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Create New...