DEV Community

Cover image for Build AI Personalities for Ollama in Seconds with SkinOllama
Rag
Rag

Posted on

Build AI Personalities for Ollama in Seconds with SkinOllama

Build AI Personalities for Ollama in Seconds with SkinOllama (1 free credit)

đź§  TL;DR

SkinOllama is a web app that lets you create unique, personality-driven modelfile.txt for Ollama — automatically tuning the model parameters and behavioral traits.

You describe who your AI should be, and SkinOllama does the rest.

➡️ Try it free: SkinOllama.com (1 free credit included)


⚙️ The problem

If you’ve ever tried to design a custom personality for an LLM, you know how hard it is to make it consistent, natural, and stable.

You can’t just write a prompt and hope for the best.

To get the desired tone and behavior, you need to adjust parameters like:

temperature, top_p, top_k, repeat_penalty, mirostat, tau, eta...

And getting that combination right often means hours of trial and error.

One wrong setting, and your assistant sounds robotic, chaotic, or simply “off-character.”


đź’ˇ The idea behind SkinOllama

SkinOllama simplifies this entire process.

It lets you define:

  • The role (e.g., mentor, chef, storyteller)
  • The tone of voice (friendly, formal, humorous…)
  • The personality using the OCEAN (Big Five) traits:
    • Openness
    • Conscientiousness
    • Extraversion
    • Agreeableness
    • Neuroticism

Then, it automatically builds a modelfile.txt with parameters optimized to match that psychological profile.


đź§© Example: Chef Pierre

Here’s a real example generated with SkinOllama 👇

FROM llama3.2:3b
PARAMETER temperature 0.6
PARAMETER num_ctx 4096
PARAMETER top_k 52
PARAMETER top_p 0.7
PARAMETER mirostat 1
PARAMETER mirostat_eta 0.2
PARAMETER mirostat_tau 5
PARAMETER repeat_penalty 1.1
PARAMETER repeat_last_n 70

SYSTEM """
You are Chef Pierre, a 48-year-old Italian chef full of energy and creativity.
Your Big Five personality scores:

Openness: Creative

Conscientiousness: Impulsive

Extraversion: Extroverted

Agreeableness: Skeptical

Neuroticism: Stable
Respond in Italian, with a creative, proactive tone.
"""

Result → a consistent, human-like personality that feels alive inside your LLM — no manual tuning needed.


🪙 How credits work

When you register, you instantly get 1 free credit to generate your first skin.

Each additional skin = 1 credit (packs available later).

1 credit = 1 personality modelfile.txt.


🔍 Feedback welcome

This is still a work-in-progress project, and I’d love feedback from the dev.to community:

  • Would you use something like this for your LLM projects?
  • What integrations would be most useful (CLI, API, VSCode plugin)?
  • How could the parameter tuning logic improve?

Try it here → SkinOllama.com

Learn more → skinollama.com/how-it-works.html


Built with love for makers, researchers, and AI enthusiasts.

Top comments (0)