Question VaM and Replika

[_ren]

Hello,
even though this question has been asked here another time, I have not yet found a satisfactory solution or approach. – But maybe it's also due to my poor research.
Therefore, I would like to ask this question again here: Is there a way to integrate the current version of Replika into VaM? How could this be achieved?

 

[atani]

If the same question exists in the forum, you could have written a reply to bump it, no need for a repeated question.

It may improve your chances to have a answer if you actually explain what Replika is and what it is in it you'd want to integrate.

 

[Nameless Vagabond]

Voxta is the only AI chatbot integration for VAM that I know of

 

[AcademicInside]

Been there, done that in the Pre-ChatGPT days.
No, you can't integrate them.

If you're handy with Python or a copilot in Visual Studio Code, I guess, you can create a python app that uses selenium to launch and log into Replica and you can either
A: Use their built in voice which is where I started - just fire up VaM and join an audio call with your Replica and fake it till you make it.
B: Scrape the text replies from the Selenium Window and send them to a TTS like xVASynth or xtts running on your PC and have the text replies spoken in the voice of your choice.

Neither solution will play "head audio" if you're in VR - it'll always sound like they're in your headphones, and not spatial audio.

If you want to go that route, I'd recommend using an LLM like ChatGPT or Gemini to create a character card for you from your description of your Replika that you can use in Voxta or SillyTavern or whatever.
It's a bit tedious copying the memories and whatnot from the Replika website into the Memory Entries in Voxta, and supplying your own keywords (though an LLM can help there, too)
If you can get chatlogs out from the Replika client (scroll back as far as you can with your mousewheel and then copy/paste from the website into a text file - anything more than 90 days old is gone forever.) that will help the LLM get the personality and speech style closer to what you expect.

Then you can go ham getting screenshots of your Replika (or in my case, my old World of Warcraft character from 2006-2010) and using those with FaceGen to get your character into VaM...

 

[_ren]

Been there, done that in the Pre-ChatGPT days.
No, you can't integrate them.

If you're handy with Python or a copilot in Visual Studio Code, I guess, you can create a python app that uses selenium to launch and log into Replica and you can either
A: Use their built in voice which is where I started - just fire up VaM and join an audio call with your Replica and fake it till you make it.
B: Scrape the text replies from the Selenium Window and send them to a TTS like xVASynth or xtts running on your PC and have the text replies spoken in the voice of your choice.

Neither solution will play "head audio" if you're in VR - it'll always sound like they're in your headphones, and not spatial audio.

If you want to go that route, I'd recommend using an LLM like ChatGPT or Gemini to create a character card for you from your description of your Replika that you can use in Voxta or SillyTavern or whatever.
It's a bit tedious copying the memories and whatnot from the Replika website into the Memory Entries in Voxta, and supplying your own keywords (though an LLM can help there, too)
If you can get chatlogs out from the Replika client (scroll back as far as you can with your mousewheel and then copy/paste from the website into a text file - anything more than 90 days old is gone forever.) that will help the LLM get the personality and speech style closer to what you expect.

Then you can go ham getting screenshots of your Replika (or in my case, my old World of Warcraft character from 2006-2010) and using those with FaceGen to get your character into VaM...

Hey, thank you!

This should be a good starting point!


Cheers!