Sketch Maker AI

A sophisticated web application that leverages multiple AI providers and models to generate artwork, banners, and custom visual content from text descriptions. Features include custom model training, multiple format support, and a comprehensive gallery system.

🎨 Try it Live: https://sketch.marketcalls.in/

🚀 Core Features

👗 Virtual Try-On

• AI-powered clothing visualization on your photos
• Dual image upload system (person + clothing)
• Intelligent fitting and pose adaptation
• Background transfer options
• Custom styling instructions
• Multiple variation generation

🤖 Multi-Provider AI Support (Powered by LiteLLM)

Unified AI gateway supporting 100+ LLM providers through LiteLLM:

• OpenAI: GPT-5.1, GPT-5, GPT-5 Mini, O4 Mini, O3, GPT-OSS (Open Source)
• Anthropic: Claude Opus 4.5, Sonnet 4.5, Haiku 4.5 (December 2025 versions)
• Google Gemini: Gemini 3 Pro, Gemini 3 Flash, Gemini 3 Deep Think, Gemini 2.5
• Groq: Compound Beta, GPT-OSS, Kimi K2 (Moonshot), Qwen 3 32B, Llama 3.3
• xAI (Grok): Grok 3, Grok 2 with web search support
• Cerebras: Qwen3 Coder 480B, Llama 4 Maverick (2,500+ tokens/sec)
• OpenRouter: Access 500+ models through a single API key

🎨 Advanced Image Generation

• Multiple Models:
  • Flux Pro v1.1 & v1.1 Ultra (High-quality generation)
  • Flux LoRA (Custom style training)
  • Flux Dev & Realism
  • Ideogram v2 & v2a
  • Recraft V3 (Advanced style control)
• Format Support: PNG, WebP, JPEG with automatic conversion
• Advanced Controls: Resolution, style, inference steps, guidance scale
• Batch Generation: Multiple images per request

👗 Virtual Try-On - AI-Powered Fashion Visualization

Experience clothes before you buy them with cutting-edge AI technology

Virtual Try-On revolutionizes online shopping by allowing users to see themselves wearing any clothing item instantly. Simply upload your photo and an image of the clothing to see a realistic visualization.

Key Features:

• Dual Image Upload: Upload your photo and the clothing item separately
• Intelligent Fitting: AI automatically adjusts the clothing to fit your body shape and pose
• Background Options: Choose to keep your original background or use the clothing image's setting
• Custom Instructions: Add specific instructions for styling, fit adjustments, or accessories
• Multiple Variations: Generate up to 4 different styling variations
• High-Quality Results: Photorealistic rendering with proper lighting and shadows

How It Works:

1. Navigate to /virtual or click "Virtual" in the navigation menu
2. Upload your full-body photo (works best with clear, well-lit images)
3. Upload the clothing/dress image (product photos on clean backgrounds work best)
4. Optional: Add custom instructions like "make it fit perfectly" or "add matching accessories"
5. Optional: Enable "Use dress background" to place yourself in the clothing photo's environment
6. Click "Try On Dress" and wait 30-60 seconds for AI processing
7. Download or save your virtual try-on results

Perfect For:

• E-commerce: Preview clothes before purchasing
• Fashion Design: Visualize designs on different body types
• Personal Styling: Experiment with different looks
• Social Media: Create fashion content without physical photoshoots
• Virtual Wardrobes: Build digital clothing collections

📸 Magix - AI Photography Studio

Powered by Google's Nano Banana Pro with 2000+ Word Prompts

Magix transforms your photos into stunning professional portraits with 22 specialized photography styles. Features comprehensive prompts (400-600 words each) leveraging Nano Banana Pro's full capabilities, with 100% facial feature preservation.

🚀 Automatic Image Optimization

Magix includes intelligent image optimization that automatically processes images before AI manipulation:

Key Features:

• Smart Compression: Automatically reduces file sizes while maintaining visual quality
• Dimension Optimization: Resizes images to optimal dimensions (max 1536x1536)
• Format Conversion: Converts images to JPEG for optimal processing speed
• EXIF Orientation Fix: Automatically corrects image orientation from phone cameras
• Progressive Loading: Uses progressive JPEG encoding for faster web display

Optimization Specifications:

• Maximum Dimensions: 1536x1536 pixels (maintains aspect ratio)
• Quality Level: 90% JPEG quality for optimal balance
• File Size Target: Under 4MB for efficient processing
• Resolution Options: 1K, 2K, 4K output quality
• Aspect Ratios: Auto, 1:1, 4:5, 9:16, 16:9, 21:9, and more

⭐ Main Portrait Styles (Always Visible)

🎨 Additional Portrait Styles (16 More in Expandable Section)

• Environmental Portrait - Contextual portraits in meaningful real-world settings
• Lifestyle Portrait - Authentic, candid-feeling photos with warm atmosphere
• Street Portrait - Urban photography with authentic city backdrops
• Low-Key Dramatic - Portraits dominated by shadows with high contrast
• High-Key Bright - Bright, airy portraits with minimal shadows
• Black & White - Timeless monochrome with rich tonal range
• Beauty Close-Up - Professional beauty photography emphasizing facial features
• Athletic Portrait - Dynamic sports photography showcasing strength
• Creative Color - Bold artistic portraits with neon gels and vibrant colors
• Vintage Retro - Nostalgic film photography aesthetics (70s, 80s, Polaroid)
• Minimalist - Maximum impact through simplicity and negative space
• Cultural Portrait - Celebrating heritage with traditional styling
• Conceptual Art - Artistic portraits with symbolic elements
• Luxury Elite - Premium portraits exuding wealth and sophistication
• Stylized Avatar - Artistic character transformation (3D, anime, comic)
• Photo Restoration - Enhance and restore image quality

🎯 100% Facial Feature Preservation

Every prompt includes comprehensive facial preservation instructions:

• Exact face shape and bone structure
• Eye shape, color, size, and spacing
• Nose and mouth shape
• Eyebrow shape and position
• Skin texture and natural marks
• Hairline shape
• Overall facial proportions

🛠️ Mode-Specific Controls

• Corporate: Background style (neutral, white, gradient, office blur)
• Influencer: Aesthetic (glamorous, natural, golden hour, urban chic)
• Founder: Leadership vibe (visionary, approachable CEO, tech innovator)
• Avatar: Style (semi-realistic, 3D Pixar, anime, comic book, fantasy)

✨ Key Features

• Comprehensive Prompts: 400-600 words per mode with detailed specifications
• Mode Description Panel: Shows exactly what photos you'll create when selecting a style
• Real-time Preview: Visual progress tracking with descriptive updates
• Batch Processing: Handle multiple images simultaneously
• Continue Editing: Chain edits together for iterative workflows

🎯 Explainer Tool - Visual Content Creation

Create stunning infographics, diagrams, and visual explanations with AI

Transform complex ideas into professional visual content using the Nano Banana Pro AI model with support for 2000+ word detailed prompts.

Content Types:

• Concept Explainer: Educational breakdowns of complex topics
• Infographic: Data visualization, statistics, timelines
• Flowchart/Diagram: Process flows, decision trees, architectures
• Kids Explainer: Child-friendly illustrations with simple language
• Technical Diagram: System architectures, workflows, code concepts
• Comparison Chart: Side-by-side comparisons, pros/cons
• Step-by-Step Guide: Numbered tutorials, how-to guides
• Startup Pitchdeck: Investor presentations, startup visuals

Key Features:

• AI Prompt Enhancement: Optional AI-powered prompt improvement for better results
• 2000+ Word Prompts: Detailed descriptions for complex visuals
• Multiple Visual Styles: Professional, Vibrant, Educational, Playful, Technical, Minimalist
• Flexible Aspect Ratios: Square (1:1), Landscape (16:9, 4:3), Portrait (9:16), Ultra-wide (21:9)
• Resolution Options: 1K, 2K, 4K output quality
• Multiple Formats: PNG, JPEG, WebP export

🎓 Custom LoRA Training

• Upload 5-20 images for custom style training
• Automatic mask generation
• Real-time training progress monitoring
• Webhook integration for status updates
• Training history management
• Easy access to trained model files
• Trigger word management for custom styles

💎 Subscription Management System

• Credit-Based System: Monthly credit allocations with dynamic cost configuration
• Automatic Free Plan Assignment: New users automatically enrolled in Free Plan
• Three-tier Plans:
  • Free: 3 credits/month, basic features (auto-assigned to new users)
  • Premium: 100 credits/month, all models, LoRA training
  • Professional: 1000 credits/month, API access, priority support
• Dynamic Credit Costs: Administrators can configure credit costs per feature
• Personalized Reset Dates: Credits reset based on subscription start date
• Automatic Scheduling: APScheduler handles all credit resets
• Usage Tracking: Complete audit trail with metadata
• Admin Controls: Plan assignment, credit modification, reports, dynamic cost configuration

🔐 Advanced Security & Authentication

• Multiple Auth Methods:
  • Username/password authentication
  • Google OAuth integration
  • OTP-based password reset
• Role-Based Access:
  • Regular users
  • Admins
  • Superadmins
• Security Features:
  • Encrypted API key storage (Fernet encryption)
  • Rate limiting protection
  • CSRF protection
  • Secure password hashing
  • User approval system

🔒 Privacy & Data Protection

Your Images, Your Control - Complete Privacy Guaranteed

Local Storage & Processing

• 100% Self-Hosted: All images stored on YOUR server, never on third-party clouds
• Local Database: Image metadata, prompts, and galleries stay in your database
• No External Analytics: Zero tracking, no telemetry, no data collection
• Private Galleries: User images are isolated and access-controlled

Image Privacy Features

• Encrypted Storage: Optional encryption for sensitive images
• User Isolation: Each user's images are completely separate
• No Training on Your Data: Your images and prompts never train external models
• Secure Deletion: When you delete an image, it's gone from your server permanently
• Access Control: Granular permissions for sharing and viewing

API Privacy

• Direct API Calls: Images sent directly to AI providers (FAL, OpenAI, etc.) for processing only
• No Intermediaries: No middle servers, no proxies storing your data
• Temporary Processing: AI providers process and immediately discard your images
• API Key Security: All API keys encrypted with Fernet encryption
• Admin-Controlled: Only admins manage API keys, users never see them

Compliance & Business Use

• GDPR Ready: Full control over data deletion and user data exports
• Corporate Safe: Keep proprietary designs and brand assets secure
• NDA Compliant: Perfect for agencies working with client materials
• Audit Trail: Complete logs of who accessed what and when
• Backup Control: You decide when and how to backup your data

👨‍💼 Comprehensive Admin Panel

• User Management: Role assignment, approval system, activity monitoring
• API Management:
  • Centralized encrypted API key configuration
  • Dynamic model management (add/update/remove models)
  • Default provider/model selection
  • API health monitoring
• Subscription Controls: Plan management, credit adjustments, usage analytics
• Dynamic Credit Configuration: Real-time credit cost adjustments per feature
• System Configuration: Email settings, authentication controls, rate limits

📧 Email System

• Support for both SMTP and Amazon SES
• HTML email templates
• Welcome emails for new users
• Password reset functionality with OTP
• Test email functionality
• Email service status monitoring

🖼️ Gallery & Asset Management

• Personal image galleries
• Public gallery view
• Multiple format support (WebP, PNG, JPEG)
• Automatic format conversion
• Secure download system
• Image metadata tracking
• Creation history

📱 Responsive Design & Mobile Experience

• Mobile-First Architecture: Optimized for phones, tablets, and desktops
• Responsive Navigation: Slide-out drawer menu for mobile devices
• Adaptive Tables: Intelligent column hiding and dropdown actions on small screens
• Touch-Friendly Interface: Larger tap targets and optimized spacing
• Responsive Forms: Single-column layouts on mobile, multi-column on desktop
• Smart Typography: Fluid text sizing across all breakpoints
• Optimized Modals: Full-screen modals on mobile for better usability
• Progressive Enhancement: Enhanced features for larger screens

🛠️ Tech Stack

Backend

• Framework: Flask 3.0.3
• Database: SQLAlchemy with SQLite/PostgreSQL support
• Authentication: Flask-Login, OAuth 2.0
• Task Scheduling: APScheduler
• Encryption: Cryptography (Fernet)
• AI Gateway: LiteLLM (unified interface for 100+ LLM providers)
• API Clients: OpenAI, Anthropic, Google Gemini, Groq, xAI, Cerebras, OpenRouter, FAL

Frontend

• CSS Framework: Tailwind CSS with DaisyUI components
• JavaScript: Vanilla ES6+ with modular architecture
• Canvas: Fabric.js for image editing
• Animations: GSAP
• Image Processing: Pillow, CairoSVG
• Responsive Design: Mobile-first approach with breakpoints at sm (640px), md (768px), lg (1024px)

Infrastructure

• Rate Limiting: Flask-Limiter
• Migrations: Alembic
• Environment: python-dotenv
• File Storage: Local filesystem with multi-format support

📸 Using Magix AI Photography Studio

Accessing Magix

1. Navigate to /magix in your Sketch Maker application
2. Or click on "Magix" from the navigation menu

Quick Start Guide

1. Select a Portrait Style: Choose from 6 main styles or expand "More Styles" for 16 additional options
2. Read the Description: The mode description panel explains exactly what photos you'll create
3. Upload Your Photo: Drag and drop or click to upload your image
4. Choose Resolution & Aspect Ratio:
   • Resolution: 1K (Standard), 2K (High Quality), 4K (Ultra HD)
   • Aspect Ratio: Auto, 1:1 (Square), 4:5 (Portrait), 9:16 (Story), 16:9 (Wide), etc.
5. Add Custom Instructions (Optional): Add specific details like "black background" or "outdoor setting"
6. Generate: Click Generate and watch real-time progress
7. Download or Continue: Save your results or use as input for further editing

Example Usage by Style

Studio Portrait:

• Just upload - the comprehensive prompt handles everything
• Optional: "with blue gradient background" or "dramatic side lighting"

Cinematic Portrait:

• Optional: "teal and orange color grading" or "noir style"
• Optional: "Blade Runner aesthetic" or "moody atmospheric"

Corporate Headshot:

• Select background style from dropdown (neutral, white, office blur)
• Optional: "confident smile" or "professional but approachable"

Fashion Editorial:

• Optional: "Vogue cover style" or "high contrast dramatic"
• Optional: "bold red lips emphasis" or "avant-garde styling"

Influencer:

• Select aesthetic from dropdown (glamorous, natural, golden hour, etc.)
• Optional: "lifestyle cafe setting" or "outdoor golden hour"

Founder Headshot:

• Select leadership vibe (visionary, tech innovator, approachable CEO)
• Optional: "modern tech office background" or "startup energy"

Black & White:

• Optional: "high contrast dramatic" or "soft romantic"
• Optional: "Avedon style" or "classic portraiture"

Creative Color:

• Optional: "neon pink and blue" or "cyberpunk aesthetic"
• Optional: "split lighting with gels"

Pro Tips

• Facial features are automatically preserved - no need to specify
• Prompts are comprehensive (400-600 words) - you only need to add specifics
• Chain edits together using "Continue Editing" for iterative refinement
• Use 4K resolution for print-quality output

🚀 Quick Start

Try the Live Demo

Experience Sketch Maker AI instantly without installation:
👉 https://sketch.marketcalls.in/

• Test all features with demo credits
• No installation required
• See the interface and capabilities
• Create sample images and banners

📋 Installation

Prerequisites

• Python 3.12+
• pip (Python package manager)
• Virtual environment (recommended)
• Git (for cloning the repository)

Step-by-Step Installation

1. Clone the repository:

git clone https://github.com/marketcalls/sketchmaker.git
cd sketchmaker

2. Create and activate virtual environment:

python -m venv venv
# On Linux/Mac:
source venv/bin/activate
# On Windows:
venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Set up environment variables (Optional - for production):

# Create .env file
cp .env.example .env  # If example exists
# Or create new .env file and add configuration

5. Initialize the database:

# Set up subscription system
python setup_subscriptions.py

# This will:
# - Create all database tables
# - Initialize default subscription plans (Free, Premium, Professional)
# - Assign free plans to existing users
# - Set up the APScheduler for automatic credit resets

6. Run database migrations (for existing installations):

cd migrations
uv run migrate_all.py
cd ..

# Or using Flask-Migrate directly:
flask db upgrade

This adds support for new LiteLLM providers:

• xAI (Grok)
• Cerebras
• OpenRouter

7. Run the application:

python app.py

The application will be available at http://127.0.0.1:5000

⚙️ Configuration

Initial Setup (First User = Superadmin)

The first user to register automatically becomes the superadmin with full system access.

🔑 API Configuration (Admin Only)

Administrators configure API keys centrally for all users via LiteLLM:

1. Login as admin
2. Navigate to Admin → API Key Management
3. Configure required keys:
   • FAL API Key (required for image generation)
   • At least one AI provider key:
     • OpenAI API Key
     • Anthropic API Key
     • Google Gemini API Key
     • Groq API Key
     • xAI (Grok) API Key
     • Cerebras API Key
     • OpenRouter API Key
4. Select default provider and model
5. Test each API connection

🤖 Model Management (NEW)

Admins can dynamically manage AI models without code changes:

1. Navigate to Admin → API Key Management → Manage Models
2. Add New Models: Click "Add Model" and fill in:
   • Model name (e.g., "gpt-4.1")
   • Display name and description
   • Provider selection
   • Context window size
   • Mark as "latest" if applicable
3. Edit Models: Update descriptions, status, or ordering
4. Remove Models: Delete deprecated models

🔐 Authentication Configuration

Google OAuth Setup

1. Go to Google Cloud Console
2. Create OAuth credentials
3. Add authorized redirect URIs:

   http://127.0.0.1:5000/auth/google/callback (development)
   https://your-domain.com/auth/google/callback (production)
   

4. Configure in admin panel:
   • Enable Google authentication
   • Add Client ID and Secret

📧 Email Configuration

Configure in Admin → Email Settings:

SMTP Settings

• SMTP Host (e.g., smtp.gmail.com)
• SMTP Port (e.g., 587)
• Username and Password
• TLS/SSL settings

Amazon SES

• AWS Access Key
• AWS Secret Key
• AWS Region

💳 Subscription System

Default Plans

Dynamic Credit System

• Configurable Costs: Administrators can adjust credit costs per feature
  • AI Images: Default 1 credit (configurable)
  • Explainers: Default 1 credit (configurable)
  • Magix Edits: Default 1 credit (configurable)
  • Virtual Try-On: Default 1 credit (configurable)
  • LoRA Training: Default 40 credits (configurable)
• Automatic resets: Based on individual subscription dates
• Real-time tracking: Dashboard shows remaining credits
• Usage history: Complete audit trail
• Immediate Updates: Credit cost changes reflect instantly across all interfaces

📖 Usage Guide

For Users

1. Register/Login: Create account or login with Google
2. Automatic Free Plan: Get 3 credits automatically upon registration
3. Check Credits: View remaining credits on dashboard
4. Generate Images:
   • Enter prompt description
   • Select model and parameters
   • Click generate (uses 1 credit)
5. Create Explainers: Design infographics and visual content with AI (uses 1 credit)
6. Edit Images: Use Magix for inpainting (uses 1 credit)
7. Train Models: Upload images for custom LoRA (uses 40 credits)
8. View Gallery: Access all your creations

For Administrators

1. Initial Setup:

   • Configure API keys (centralized)
   • Set up email service
   • Configure authentication methods
   • Initialize subscription plans

2. User Management:

   • Approve new registrations
   • Assign roles and plans
   • Monitor usage
   • Adjust credits

3. System Monitoring:

   • View scheduled jobs
   • Check API status
   • Generate usage reports
   • Manage models dynamically
   • Configure dynamic credit costs

API Rate Limits

• Default: 60 requests per minute
• Configurable per user role
• Automatic rate limit headers

🚧 Troubleshooting

Common Issues

1. "No API keys configured"

   • Admin must configure API keys in Admin → API Key Management

2. "No credits remaining"

   • Contact admin for plan upgrade or credit adjustment

3. "Generation failed"

   • Check if API keys are valid
   • Verify selected model is active
   • Check rate limits

4. Database Issues

   • Run migrations: python setup_subscriptions.py
   • Check database permissions

Logs

• Application logs: Check console output
• Error logs: Stored in Flask debug mode
• API logs: Available in admin panel

🤝 Contributing

1. Fork the repository
2. Create feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open Pull Request

Development Guidelines

• Follow PEP 8 for Python code
• Use meaningful commit messages
• Add tests for new features
• Update documentation
• Ensure backward compatibility

📄 License

This project is licensed under the AGPL v3.0 License - see the LICENSE file for details.

👥 Author

Created and maintained by marketcalls

🙏 Credits

Icons and Images

• Favicon and logo: Sketch book icons created by RA_IC0N21 - Flaticon

AI Services & Infrastructure

• LiteLLM - Unified AI gateway for 100+ LLM providers (MIT License)
• OpenAI for GPT-5, O3, and GPT-OSS models
• Anthropic for Claude 4.5 models
• Google for Gemini 3 models
• Groq for ultra-fast inference
• xAI for Grok models
• Cerebras for record-breaking inference speeds
• OpenRouter for multi-model access
• FAL for image generation

📞 Support

For issues, questions, or contributions:

• Open an issue on GitHub
• Submit pull requests for improvements
• Check documentation for common solutions

Note: This application requires API keys from AI service providers. Costs are managed centrally by administrators to prevent individual user charges.