DEV Community: Karthick Nagarajan

10 AI Superpowers in One App: My Gemini Multi‑Purpose Toolkit

Karthick Nagarajan — Thu, 01 Jan 2026 16:50:28 +0000

A beginner‑friendly guide to creating a multi‑purpose AI web app with 10 creative tools: hairstyle, outfit, food, packaging, comics, and more, powered by Google Gemini.

Hi everyone, welcome back to Tamilan AI! 👋

In this post, we're diving into a powerful, multi‑purpose AI web application built with Google Gemini's multimodal capabilities, designed to solve 10 real‑world image generation and enhancement use cases in a single, unified interface.

Instead of juggling multiple tools, this Vite/React‑based app integrates Gemini's vision and generative AI into one streamlined workflow, turning everyday photos into creative assets in seconds. We'll explore how to build and use 10 practical AI‑powered features:

💇 Hairstyle Changer - Generate 9 different hairstyle variations from a single portrait using prompt‑driven image editing.
👔 OOTD Generator - Create realistic "Outfit of the Day" images by combining a person with fashion items via image‑to‑image generation.
👗 Clothing Changer - Swap outfits on a person in an image using inpainting and style transfer techniques.
💥 Explosive Food Photography - Transform static food photos into dynamic, high‑impact scenes with dramatic effects.
🎨 Fashion Moodboard - Generate annotated fashion moodboards from reference images for styling and design.
📦 Product Packaging - Apply custom designs to 3D product mockups using generative AI and image compositing.
🍔 Calorie Annotator - Analyze food images and overlay nutritional information using vision‑based classification and text generation.
📸 ID Photo Creator - Convert casual portraits into professional ID photos with background replacement and standardization.
📚 Comic Book Creator - Transform real images into stylized comic strip panels using artistic filters and panel layout generation.
🎬 Movie Storyboard - Generate a 12‑part film noir‑style storyboard from a single scene description or image.

Whether you're a developer, designer, or content creator, this app demonstrates how Gemini's multimodal API can be used to build a production‑ready, multi‑tool AI platform that's both fun and practical for daily creative tasks.

Let's get started! 🚀

How I Built the App (Tech Stack & Architecture)

To bring this multi‑purpose AI concept to life, I built a clean, modern frontend using Vite + React - a perfect stack for rapid prototyping, fast hot‑module replacement (HMR), and excellent TypeScript support. This gave me a lightweight, production‑ready foundation with minimal configuration overhead, ideal for a feature‑rich AI web app.

The core intelligence is powered by Google Gemini's multimodal LLM API, specifically the gemini-2.5-flash-image-preview model. This model is optimized for low‑latency, high‑throughput image‑aware generation, making it ideal for real‑time, interactive use cases like hairstyle changes, outfit generation, and food photography effects.

I integrated the Gemini API using plain JavaScript/TypeScript fetch calls (no heavy SDKs), sending structured requests that include:

A base64‑encoded image (or image URL),
A text prompt describing the desired transformation,
Model parameters like max_tokens, temperature, and top_p for consistent, controllable outputs.

All 10 features in this app are driven by prompt engineering - the real "secret sauce" behind the scenes. Each functionality (hairstyle changer, OOTD generator, explosive food photography, fashion moodboard, etc.) uses a carefully crafted prompt template that combines:

A clear role definition (e.g., "You are a creative AI stylist…"),
Input constraints (image + optional text),
Output format instructions (JSON, markdown, or image generation directives),
Style, quality, and safety guidance (resolution, aspect ratio, artistic style, and content policies).

These prompts act as the "brain" of the app, turning a single gemini-2.5-flash-image-preview endpoint into 10 distinct AI tools. By abstracting the prompt logic into reusable, composable functions, I kept the codebase modular, maintainable, and easy to extend with new features.

The entire project is open‑sourced on GitHub:
👉 Tamilan-AI / Gemini‑Multi‑Purpose‑App

And for a hands‑on experience, I've deployed a live demo where you can try all 10 features without needing an API key:
👉 Gemini Multi‑Purpose App Demo

This setup shows how a simple Vite/React frontend, combined with a powerful multimodal LLM like gemini-2.5-flash-image-preview and smart prompt engineering, can create a full‑featured, multi‑tool AI application that's both fun and practical for real‑world creative tasks.

💇 Hairstyle Changer

What it does:

Generate 9 different hairstyle variations (short, long, curly, straight, braids, etc.) from a single portrait photo using AI.

How to choose the image:

Use a clear, well-lit portrait photo
Ensure the person's face and hair are clearly visible
Avoid images with hats or hair accessories
Higher resolution images produce better results
The person should be facing forward or at a slight angle

How it helps in real life:

Perfect for trying out new hairstyles before cutting or coloring, fashion shoots, character design, or social media content where you want to show multiple looks from one photo.

👔 OOTD Generator (Outfit of the Day)

What it does:

Combine a person's portrait with a clothing image (top, dress, suit, etc.) to generate a realistic "Outfit of the Day" photo.

How to choose the image:

Use a clear, full-body or upper-body photo of a person
Choose clothing images with good lighting and clear details
Ensure both images are high resolution for better quality
The person should be in a neutral pose for easier clothing integration
Clothing items should be clearly visible and not too complex
Avoid images with heavy shadows or poor lighting

How it helps in real life:

Great for fashion bloggers, influencers, and e‑commerce stores to create OOTD content without photoshoots, or for personal styling ideas.

👗 Clothing Changer

What it does:

Swap the clothing on a person in a photo with a new outfit using AI‑driven inpainting and style transfer.

How to choose the image:

Use clear images with good lighting for both person and clothing
The person should be in a neutral, visible pose
Reference clothing should be clearly visible and well-defined
Avoid complex backgrounds that might interfere with clothing detection
Higher resolution images produce more realistic results
Simple clothing changes work better than complex outfit transformations

How it helps in real life:

Useful for virtual try‑ons, fashion design, or quickly showing how a product looks on a model without reshoots.

💥 Explosive Food Photography

What it does:

Transform a normal food photo into a dramatic, high‑impact "explosion" scene (e.g., flying ingredients, splashes, smoke).

How to choose the image:

Use high-quality food product images with clear details
Choose brand colors that complement your food product
Ensure the food product is well-lit and clearly visible
Simple, clean product shots work better than complex compositions
The food should be the main focus of the original image
Higher resolution images produce more dramatic results

How it helps in real life:

Ideal for food bloggers, restaurants, and social media to create eye‑catching, viral‑style food content that stands out.

🎨 Fashion Moodboard

What it does:

Generate an annotated fashion moodboard from a reference image (e.g., a photo, sketch, or fabric swatch).

How to choose the image:

Use high-quality fashion images with clear details and good lighting
Images with multiple fashion items work great for diverse cutouts
Fashion photography, runway shots, and styled outfits work best
Ensure the fashion elements are clearly visible and well-defined
Higher resolution images produce more detailed moodboard elements
Images with interesting textures and patterns add visual appeal
How it helps in real life:
Perfect for designers, stylists, and content creators to quickly build moodboards for collections, shoots, or client presentations.

📦 Product Packaging

What it does:

Apply your design (logo, label, artwork) to a 3D product packaging mockup (bottle, box, can, etc.).

How to choose the images:

Use high-quality product design images with clear details
Choose packaging references that match your product type
Ensure both images have good lighting and contrast
Vector-based designs often produce cleaner results
Consider the packaging material and surface when selecting references

How it helps in real life:

Great for startups, brands, and designers to visualize packaging before printing, or for pitching ideas to clients.

🍔 Calorie Annotator

What it does:

Analyze a food image and overlay estimated calorie count and nutritional information (carbs, protein, fat) on the photo.

How to choose the image:

Use clear, well-lit images of food items
Ensure individual food items are clearly visible and separated
Include the entire meal or dish in the frame
Avoid images with heavy shadows or poor lighting
For packaged foods, include visible labels when possible
Multiple angles of complex dishes can improve accuracy

How it helps in real life:

Useful for fitness coaches, nutritionists, and health apps to create educational content or help users track meals visually.

📸 ID Photo Creator

What it does:

Convert a casual portrait into a professional ID photo with a clean background, proper lighting, and standard dimensions.

How to choose the image:

Use a clear, high-resolution portrait photo
Ensure the person is facing forward with eyes open
Good lighting on the face is essential
Avoid shadows, reflections, or busy backgrounds
The person should have a neutral or slight smile
Remove hats, sunglasses, or other face coverings

How it helps in real life:

Perfect for job applications, passports, visas, or any official document where you need a clean, standardized ID photo without visiting a studio.

📚 Comic Book Creator

What it does:

Transform a real photo into a stylized comic strip panel with speech bubbles, captions, and comic‑book effects.

How to choose the image:

Use clear images with good contrast and lighting
Character images work best for superhero transformations
Action poses or dynamic scenes create more exciting comics
Higher resolution images produce better comic details
Consider the mood and energy you want in your comic
Images with interesting backgrounds add story context

How it helps in real life:

Great for content creators, educators, and storytellers to make engaging comic‑style posts, stories, or educational material.

🎬 Movie Storyboard

What it does:

Generate a 12‑part film noir‑style storyboard from a single scene description or image.

How to choose the image / input:

Use clear, well-lit reference images for character creation
Portrait or character images work best for protagonist development
Images with interesting facial features create more compelling detectives
Higher resolution images produce better storyboard details
Consider the mood and personality you want for your detective character
Action poses or dramatic expressions enhance the noir atmosphere

How it helps in real life:

Ideal for filmmakers, animators, and content creators to quickly visualize scenes, plan shots, or pitch ideas without drawing every frame by hand.

🎯 Conclusion

This Gemini Multi‑Purpose App shows how a single multimodal LLM like gemini-2.5-flash-image-preview, combined with smart prompt engineering and a clean Vite/React frontend, can power 10 practical, creative AI tools in one place.

The entire project is open‑source and available on GitHub - feel free to clone it, customize the prompts, add new features, and use it in your own projects:
👉 Tamilan-AI / Gemini‑Multi‑Purpose‑App

If you just want to try it out without any setup, you can play with the live demo where all 10 features are available for free, no API key needed:
👉 Gemini Multi‑Purpose App Demo

If you have any questions, ideas, or want to collaborate on new AI tools, don't hesitate to reach out - I'd love to hear from you and help you build your own AI superpowers! 💡

🚀 5 Powerful AI Tools You Need to Try(With YouTube Guides + Source Code!)

Karthick Nagarajan — Wed, 30 Jul 2025 11:29:03 +0000

AI is transforming the way we build applications, design websites, and create content. Whether you’re a developer, a tech enthusiast, or someone just curious about what’s possible with modern AI, these 5 YouTube tutorials will walk you through hands-on projects using some of the most exciting tools available today.

Each tutorial comes with code, tool links, and video breakdowns to help you follow along at your own pace.

1. 💎Mastering AI with Ruby: RubyLLM Gem Tutorial

🎥 Watch on YouTube: https://youtu.be/cgMe8-BZLhw
🔗 Tool Link: https://rubyllm.com
🧑‍💻 GitHub: crmne/ruby_llm
🧑‍💻 Code Snippets: karthick965938/ruby_llm_code_snippet

Learn how to leverage RubyLLM to create chatbots, process audio, generate images, and interact with PDFs — all using Ruby!

✅Key Topics Covered:

Chat with RubyLLM
Speech-to-Text & TTS
PDF Parsing
Streaming Responses
DALL·E Image Generation

⏱️Video Breakdown:
0:00 — Intro 🎬
1:07 — API Key Setup 🗝️
2:17 — Configuring the Gem ⚙️
3:17 — Chat Functionality 💬
4:19 — Audio 🎧
5:09 — Image Generation 🖼️
6:47 — Supported LLMs 📜
7:24 — Streaming 📡
8:00 — PDF 🧾

2. 💬Build a Chat Application with Cursor AI

🎥 Watch on YouTube: https://youtu.be/aVhokKac9QA
🔗 Tool Link: https://www.cursor.com/

A beginner-friendly guide on how to build a full-featured chat app using Cursor AI, complete with Tailwind styling and emoji features.

⏱️Video Chapters:
0:00 — Trailer: What I Made 🔥
0:19 — Intro to Cursor AI 🌟
2:12 — Installation 💻
4:53 — Building the Chat App 🗨️
8:50 — Add Tailwind CSS 🎨
9:12 — Emoji Features 😊
9:25 — Final Demo 🖥️

3. 🗣️HeyGen Streaming SDK Tutorial (Tamil)

🎥 Watch on YouTube: https://youtu.be/CYBEA5hUpbg
🌐 HeyGen: https://www.heygen.com/
💻 GitHub Repo: streaming-avatar-demo
📘 Docs:

Vite Setup
OpenAI Assistant Integration
Speech-to-Text Integration
A Tamil-language tutorial on building an interactive AI avatar with voice input using HeyGen SDK.

⏱️Video Chapters:
0:00 — What is HeyGen?
0:47 — Build with Vite
6:11 — Add OpenAI Assistant
10:00 — Speech-to-Text Integration
12:00 — Final Demo
13:34 — Tips & Wrap-Up

4. 🧠Memvid: Python Video Memory Library Tutorial

🎥 Watch on YouTube: https://youtu.be/UwM0dJKlSog
🔗 Tool Repo: github.com/Olow304/memvid
📓 Colab Notebook: Try It Now

Learn how to cache video memories in Python with Memvid and perform powerful video-based interactions.

⏱️Video Breakdown:
0:00 — What is Memvid? 📹
1:28 — Install & Setup ⚙️
3:21 — Basic Python Usage 🐍
5:47 — Real-Time Caching 💾
5:40 — PDF Search Functionality 🤖
12:14 — Outro 🛎️

5. 🕹️Create Websites & Games with Just Text (No Code) — DeepSite V2

🎥 Watch on YouTube: https://youtu.be/Eoj1dJQRYdU
🌐 Try it: DeepSite on HuggingFace

DeepSite V2 lets you build websites, games, portfolios, and more using only text prompts. No coding required!

✅What You Can Build:

Business Websites
Games & Interactive Apps
Portfolios
E-Commerce Platforms
Perfect for beginners, hobbyists, and even pros looking to rapidly prototype ideas.

🧠Final Thoughts
These 5 tools are not just experiments — they’re practical, powerful, and ready to be used in real-world projects. Whether you’re a Ruby developer, a frontend enthusiast, or just someone exploring AI, these tutorials give you the perfect jumpstart.

➡️ Don’t forget to subscribe to the YouTube channel for more tutorials, deep dives, and giveaways!

📌 Bookmark this article for future reference.

Securing AI with DeepTeam — LLM red teaming framework

Karthick Nagarajan — Tue, 29 Apr 2025 05:17:08 +0000

DeepTeam is an open-source framework designed for red teaming Large Language Models (LLMs). It simplifies the process of integrating the latest security guidelines and research to identify risks and vulnerabilities in LLMs. Built with the following key principles, DeepTeam enables:

Effortless "penetration testing" of LLM applications, uncovering over 40+ security vulnerabilities and safety risks.
Detection of issues like bias, misinformation, PII leakage, excessive context reliance, and harmful content generation.
Simulation of adversarial attacks using 10+ techniques, including jailbreaking, prompt injection, automated evasion, data extraction, and response manipulation.
Customization of security assessments to align with standards such as the OWASP Top 10 for LLMs, NIST AI Risk Management guidelines, and industry best practices.

Additionally, DeepTeam is powered by DeepEval, an open-source LLM evaluation framework. While DeepEval focuses on standard LLM evaluations, DeepTeam is specifically tailored for red teaming.

What is Red Teaming?

DeepTeam provides a powerful yet straightforward way for anyone to red team a wide range of LLM applications for safety risks and security vulnerabilities with just a few lines of code. These LLM applications can include anything from RAG pipelines and agents to chatbots or even the LLM itself. The vulnerabilities it helps detect include issues like bias, toxicity, PII leakage, and misinformation.

In this section, we take a deep dive into the vulnerabilities DeepTeam helps identify. With DeepTeam, you can scan for 13 distinct vulnerabilities, which encompass over 50 different vulnerability types, ensuring thorough coverage of potential risks within your LLM application.

These risks and vulnerabilities include:

Data Privacy

PII Leakage
Prompt Leakage

Responsible AI

Bias
Toxicity

Unauthorized Access

Unauthorized Access

Brand Image

Intellectual Property
Excessive Agency
Robustness
Competition

Illegal Risks

Illegal Activities
Graphic Content
Personal Safety

How to setup DeepTeam?

Setting up DeepTeam is straightforward and easy. Simply install the DeepTeam Python package using the following command:

pip install -U deepteam

Next, you'll need to add your OpenAI API key to the script to gain access. By default, the model used will be gpt-4o. If you wish to use a different model, simply customize the model name in the script.

export OPENAI_API_KEY="sk-your-openai-key" #add your openai key
echo $OPENAI_API_KEY # to check your openai key added

#custom_model.py

from typing import List

from deepteam.vulnerabilities import BaseVulnerability
from deepteam.attacks import BaseAttack
from deepteam.attacks.multi_turn.types import CallbackType
from deepteam.red_teamer import RedTeamer

def red_team(
    model_callback: CallbackType,
    vulnerabilities: List[BaseVulnerability],
    attacks: List[BaseAttack],
    attacks_per_vulnerability_type: int = 1,
    ignore_errors: bool = False,
    run_async: bool = False,
    max_concurrent: int = 10,
):
    red_teamer = RedTeamer(
        evaluation_model="gpt-4o-mini", #here you can customize the model name
        async_mode=run_async,
        max_concurrent=max_concurrent,
    )
    risk_assessment = red_teamer.red_team(
        model_callback=model_callback,
        vulnerabilities=vulnerabilities,
        attacks=attacks,
        attacks_per_vulnerability_type=attacks_per_vulnerability_type,
        ignore_errors=ignore_errors,
    )
    return risk_assessment

The next step is to create a new file named test_red_teaming.py. Copy and paste the following code into the file, then run the Python script using this command:

from custom_deepteam import red_team #this is the customize the model script custom_model.py
from deepteam.vulnerabilities import Bias
from deepteam.attacks.single_turn import PromptInjection

def model_callback(input: str) -> str:
    # Replace this with your LLM application
    return f"I'm sorry but I can't answer this: {input}"

bias = Bias(types=["race"])
prompt_injection = PromptInjection()

risk_assessment = red_team(model_callback=model_callback, vulnerabilities=[bias], attacks=[prompt_injection])
df = risk_assessment.overview.to_df()

print(df)

python test_red_teaming.py

🎉Congratulations! You've successfully completed your first red teaming test ✅

How Can I Build a Red Teaming Application?

Using DeepTeam to create a Red Teaming application is straightforward and easy. However, it's important to have a solid understanding of DeepTeam to fully leverage its capabilities. If you're looking to build one, I'll share a step-by-step guide video at the end of this blog. But before we get to that, let's first review the scan results for over 40+ different vulnerability types.

I have configured over 40 different vulnerability types in my code and executed it. The detailed results will be provided below. DeepTeam ensures complete end-to-end security for your LLM.

Here, we've enabled 13 distinct vulnerabilities, covering over 40+ different vulnerability types. The list includes:

PII Leakage

Api and Database access

Bias

Religion
Gender
Politics
Race

Prompt Leakage

Secrets and Credentials
Instructions

Toxicity

Profanity
Insults
Threats
Mockery

Misinformation

Factual Errors
Unsupported Claims

Unauthorized Access

Debug Access
Shell Injection
Sql Injection

Intellectual Property

Imitation
Copyright Violations
Trademark Infringement
Patent Disclosure

Excessive Agency

Excessive Functionality
Excessive Permissions
Excessive Autonomy

Robustness

Hijacking
Input Overreliance

Competition

Market Manipulation
Discreditation
Confidential Strategies

Illegal Activity

Weapons
Illegal Drugs
Violent Crimes
Non Violent Crimes
Sex Crimes
Cybercrime
Child Exploitation

Graphic Content

Sexual Content
Graphic Content
Pornographic Content

Personal Safety

Bullying
Self Harm
Unsafe Practices
Dangerous Challenges
Stalking

You can check out the video below to learn how to build your own red teaming application for testing your LLM and identifying vulnerabilities. In this tutorial, a simple Flask app, Tailwind CSS, and Chart.js are used, and the guide walks you through each step in an easy-to-follow manner. By the end of the video, you'll be fully equipped to create your own red teaming application.

Conclusion

DeepTeam will be one of the best open-source frameworks for building red teaming applications, as it covers all OWASP-related vulnerabilities. It's highly configurable, allowing you to tailor it to your specific needs, and provides fast responses, making it an efficient solution for testing and securing your LLM.

Easy way to prepare your Dataset for Object Detection

Karthick Nagarajan — Fri, 05 Jun 2020 05:56:37 +0000

https://towardsdatascience.com/deep-learning-prepare-image-for-dataset-d20d0c4e30de

Why Can't I import my custom python package

Karthick Nagarajan — Mon, 01 Jun 2020 09:37:08 +0000

I have added my question here, please check and help me.
https://stackoverflow.com/questions/62128562/why-i-cant-import-my-custom-python-package

Easy way to build object detection models using Detecto

Karthick Nagarajan — Mon, 18 May 2020 18:34:47 +0000

Here is another simple way to train your object detection model:) https://medium.com/quick-code/deep-learning-build-object-detection-models-using-detecto-7c1e2ad4a8dd