Ai-Vandilizer

Inspiration

The internet is drowning in clean, perfect, polished AI-generated content — but that perfection often lacks personality, emotion, and chaos. We wanted to fight back against the sterile and robotic feel of most AI outputs by adding a human layer of raw, unfiltered expression.

We asked ourselves: What if we could vandalize AI’s perfection with the flaws that make us human?

Thus, AI Vandalizer was born — a rebellion against robotic perfection and a celebration of emotional noise, typos, sarcasm, and all the weirdness that makes us, us.

What it does

AI Vandalizer is a Flask-based web app that takes any AI-generated content (text like essays, poems, or even code) and "vandalizes" it by adding:

Typos, slang, sarcasm, and chaotic annotations
Emotional commentary like "This hit hard 💔" or "This part reminds me of my ex 😩"
Visual elements like strikethroughs, scratch-outs, emojis, and sticky notes
Optional personalities: e.g., burnt-out artist, melodramatic teen, sarcastic coder
Adjustable chaos level to control how "vandalized" the output becomes

How I built it

We used the following stack:

Frontend: Tailwind CSS + basic HTML/JS for a clean but expressive UI
Backend: Python Flask
LLM API: Groq’s implementation of Meta’s LLaMA model
Text transformation: Prompt engineering with fine-tuned system prompts to inject emotion, tone, and personality overlays
Rendering: Added custom annotations, emojis, and markdown-style edits for a vandalized visual style

The key was crafting effective prompts to turn bland AI responses into vibrant human-like text with attitude.

Challenges I ran into

Tone control: Getting the LLM to consistently follow a chosen personality while still maintaining the context of the original text
Chaos balancing: Making sure the vandalization was funny and emotional — not just messy or broken
User experience: Designing a frontend that looked like vandalism but still felt functional and clean to use
Latency issues with the Groq API at high chaos levels due to prompt complexity

Accomplishments that I'm proud of

Created a unique and fun experience that makes people feel something when they read AI content
Built a working MVP with chaotic personalities, dynamic tone shifting, and real-time processing
Created an actual personality layer on top of AI — something that feels like a co-writer with mood swings 💀

What I learned

How powerful tone and prompt engineering can be when building expressive AI apps
The delicate balance between being funny and being annoying — especially with slang or sarcasm
How to build for vibe and emotion, not just utility
Learned how to intentionally break AI perfection and make it more human

What's next for AI Vandalizer

Add support for image vandalism — drawing over AI-generated images with doodles or annotations
Let users create and save their own personalities
Build a Chrome extension that vandalizes content across the web in real-time
Launch on Product Hunt, GitHub, and open it up to the public for chaos-creation
Maybe even build a collaborative vandalizer, where multiple users can vandalize together in real time