AI-Uni & Ani

• 

  Alien (Uni): grows into a child → regresses back to alien

• 

  Dinosaur: healthy → thrives → overuse → fossil

• 

  Fish: healthy → vibrant colors → overuse → pale skin

• 

  Parrot: healthy → fluent dialogue → overuse → short replies

Inspiration

We wanted to explore how AI could help address the challenge of overuse and unhealthy attachment to digital platforms. Instead of showing users warnings or pop-ups, we asked: what if the AI itself reflected the user’s state?

That idea became AI-Uni — a world where characters grow, thrive, or regress depending on how you engage with them.

What it does

• Characters evolve or regress based on user engagement patterns:
• Balanced engagement → Characters grow and flourish
• Excessive use → Characters regress, providing natural emotional feedback
• Multiple character types:
• 👽 Alien (Uni) → grows into a child → regresses back to alien
• 🦖 Dinosaur → thrives → turns into fossil
• 🐠 Fish → vibrant colors → fades pale
• 🦜 Parrot → fluent dialogue → short replies
• GPT-OSS powers the conversational logic:
• Adjusts tone, verbosity, and dialogue style dynamically
• Links text responses to visual character states

How we built it

• GPT-OSS: Core AI for analyzing input and generating character dialogue in different styles
• Prompt-based character design: Prototyping unique “alien speech” and “parrot-style responses” without fine-tuning (fine-tuning planned later)
• Frontend prototype: Visual assets + state changes tied to usage patterns
• Domain: AI-Uni.ai (secured for future expansion)

Challenges we ran into

• Designing dialogue shifts that feel natural, not forced
• Balancing “fun character interaction” with “healthy usage feedback”
• Limited time to build multiple characters with distinct states

Accomplishments that we're proud of

• Created a novel non-verbal feedback loop for digital wellness
• Prototyped multiple characters with unique growth/regression mechanics
• Showcased how GPT-OSS can extend beyond text into visual emotional design

What we learned

• Emotional feedback works best when it’s subtle and playful, not judgmental
• Visual + conversational change is far more impactful than text alone
• Users quickly empathize with characters, making state changes meaningful

What's next for AI-Uni & Ani

• Implement fine-tuning for consistent “alien” and “parrot” speaking styles
• Launch as a character-based AI wellness app under MASAL (MINAMI AI Safety & Alignment Lab)
• Expand with more characters, states, and personalization
• Public beta release on AI-Uni.ai

Built With

• adobe-creative-sdk
• github
• python
• react-native