HeHoXa

joined 2 weeks ago
[–] HeHoXa@lemmy.zip 1 points 10 hours ago (1 children)

Everyone laughs at the precisely arranged layers of folded blankets at the foot of the bed until they wake up shivering.

Same for my mountain of pillows of different thickness.

[–] HeHoXa@lemmy.zip 1 points 23 hours ago* (last edited 23 hours ago)
[–] HeHoXa@lemmy.zip 3 points 1 day ago

Maybe keep a little critical information on a dead man's switch and let that be known when revealing the rest.

Idk if actually wise, but it seems like you might want to keep some leverage for self protection

[–] HeHoXa@lemmy.zip 2 points 1 week ago

To satisfy the curiosity, just download Ollama. You can run your interactions locally, no feeding the demon and full privacy.

Set up a modelfile with system prompts for the kind of personality you want to create. Do some googling, may want to tweak settings to get response creativity just right. Higher parameter base models generally better and more hardware costly.

If you want pics, you can use StableDiffusion. I recommend using docker images because it's really complicated and environment sensitive.

You can tie the two together with a little Python logic and have the core bot prompt the image generator at certain chat moments. Ironically, if you're not comfortable writing code, you could just get VSCode and Gemini and spit out some pretty okay Python code without strong understanding.

There are also tools for local audio generation I'm less familiar with but am sure Google could get you there.

Maybe start with a basic Ollama setup and see if it feels like it's working for you.