Train image, sound, and pose recognition models directly in your browser. Intuitive interface for anyone to create custom ML models.

What You Can Do:

• Image Classification: Teach computer to recognize objects, gestures, facial expressions
• Audio Recognition: Create sound detectors, voice command systems
• Pose Detection: Build body movement recognition applications
• Export Models: Use in websites, apps, or Arduino projects

Perfect For:

Quick prototypes, educational demos, testing ML concepts before investing in custom development.

Try Teachable Machine →

Creative AI toolkit for video, image, and generative content. No-code interface for powerful AI models.

Capabilities:

• Video Editing: Remove backgrounds, inpaint, motion tracking with AI
• Image Generation: Text-to-image, style transfer, super-resolution
• Audio: Text-to-speech, voice cloning, music generation
• 3D: Generate 3D models from images

Use Cases:

Content creation, video production, creative experimentation, rapid prototyping visual concepts.

Explore Runway →

No-code platform for automating business workflows with custom AI models. Visual interface for training classifiers.

Business Applications:

• Document Processing: Extract data from invoices, receipts, forms
• Email Triage: Auto-categorize and route customer emails
• Image Moderation: Filter inappropriate content automatically
• Sentiment Analysis: Analyze customer feedback at scale

Integrates with Zapier, Make, and other automation platforms.

Visit Levity →

Free desktop app for training custom image classification models. Simple drag-and-drop interface.

Features:

• Easy Training: Drop images, label categories, train model
• Real-time Preview: Test with webcam immediately
• Export Formats: TensorFlow, TensorFlow Lite, ONNX, CoreML
• 100% Free: No cloud costs, runs locally

Download Lobe →

Run large language models (like ChatGPT) on your own computer. Simple installation, completely private, works offline.

What Is Ollama?

Think of it like iTunes for AI models—download, manage, and run LLMs on your Mac, Windows, or Linux machine.

Popular Models You Can Run:

• LLaMA 3: Meta's open-source model, ChatGPT-quality for many tasks
• Mistral: Excellent for coding, analysis, and general chat
• Phi-3: Microsoft's small but powerful model (runs on laptops)
• CodeLlama: Specialized for programming assistance

Simple Setup (Non-Technical):

1. Download: Visit ollama.com, download installer for your OS
2. Install: Double-click and follow prompts (like any app)
3. Open Terminal: On Mac: Applications → Terminal. On Windows: Search "cmd"
4. Run Model: Type ollama run llama3 and press Enter
5. Chat: Wait for download (first time only), then start chatting!

System Requirements:

• Minimum: 8GB RAM, can run small models (Phi-3)
• Recommended: 16GB RAM for better models (LLaMA 3)
• Best: 32GB+ RAM and GPU for fastest responses

Why Use Ollama:

• ✓ Complete privacy—your conversations never leave your computer
• ✓ No internet needed after initial download
• ✓ No subscription fees or API costs
• ✓ Unlimited usage

Get Ollama →

Generate images from text descriptions on your own computer. Open-source alternative to DALL-E and Midjourney.

What You Can Create:

• Art & Illustrations: Any style, any subject
• Product Mockups: Visualize ideas before building
• Photo Editing: Inpaint, outpaint, upscale images
• Concept Art: Characters, environments, objects

Ways to Run It:

Option 1: Automatic1111 (Popular, Beginner-Friendly)

1. Download from GitHub
2. Run the installer script (double-click .bat on Windows, .sh on Mac/Linux)
3. Wait for automatic setup (downloads everything needed)
4. Open browser to localhost:7860
5. Type prompt, click Generate!

Option 2: ComfyUI (Covered below—more powerful)

System Requirements:

• GPU: NVIDIA RTX 2060 or better (6GB+ VRAM)
• RAM: 16GB minimum
• Storage: 20GB+ for models
• Mac: M1/M2/M3 works but slower than NVIDIA

Where to Get Models:

• Civitai.com: 100,000+ community models (realistic, anime, artistic)
• Hugging Face: Official Stable Diffusion versions
• Built-in: Automatic1111 downloads default model on first run

Get Stable Diffusion →

Node-based interface for Stable Diffusion. More powerful than Automatic1111 but still accessible to non-coders.

Why ComfyUI?

• Visual Workflow: Connect nodes like puzzle pieces—no coding required
• More Control: Fine-tune every aspect of generation
• Faster: More efficient than Automatic1111
• Workflows: Save and share complete pipelines

Learning Curve:

Day 1: Confusing—lots of nodes and connections.

Week 1: Start to understand the workflow logic.

Month 1: Create complex multi-step generations easily.

Setup for Beginners:

1. Download: Get ComfyUI from GitHub
2. Portable Version: Easier for Windows—comes with everything bundled
3. Run: Double-click run_nvidia_gpu.bat (or cpu version if no GPU)
4. Load Workflow: Use example workflows to start
5. Download Models: Place in models/checkpoints folder

Beginner-Friendly Resources:

• Civitai Workflows: Download pre-made workflows with instructions
• YouTube Tutorials: "ComfyUI for beginners" has many guides
• ComfyUI Manager: Extension to install custom nodes easily

When to Use ComfyUI vs Automatic1111:

• Use Automatic1111 if: You want simple, click-and-generate experience
• Use ComfyUI if: You want maximum control and don't mind learning

Get ComfyUI →

Desktop app for running local LLMs with beautiful UI. Easier than Ollama for complete beginners.

Why LM Studio:

• No Terminal: Everything through friendly graphical interface
• Model Browser: Search and download models with one click
• ChatGPT-Like UI: Familiar chat interface
• Easy Comparison: Try different models side-by-side

Setup (Zero Technical Knowledge):

1. Download: Get LM Studio from lmstudio.ai
2. Install: Drag to Applications (Mac) or run installer (Windows)
3. Browse Models: Click "Discover" tab, see available models
4. Download Model: Click download button next to model you want
5. Start Chatting: Select model from dropdown, start conversation

Recommended First Models:

• Phi-3-mini (8GB RAM): Great for laptops
• LLaMA-3-8B (16GB RAM): Excellent general model
• Mistral-7B (16GB RAM): Good at reasoning tasks

Get LM Studio →

Desktop AI assistant that runs 100% offline. Clean interface, easy setup, works on Mac/Windows/Linux.

Features:

• Chat Interface: Just like ChatGPT but private
• Model Library: Download models from built-in hub
• Extensions: Add capabilities like web search
• Import/Export: Save conversations locally

Advantages Over Cloud AI:

• ✓ No data sent to servers—completely private
• ✓ Works on planes, trains, anywhere without internet
• ✓ No usage limits or rate limiting
• ✓ Free forever—just hardware costs

Download Jan →

Install Stable Diffusion, LLMs, and other AI tools with one click. Manages all dependencies automatically.

What Pinokio Does:

Automates the complex setup process for AI tools. Instead of following 20-step installation guides, just click "Install".

Available Apps (200+):

• Image Generation: Stable Diffusion, DALL-E mini
• LLMs: Ollama, LM Studio, text-generation-webui
• Voice: Whisper (speech-to-text), Coqui TTS
• Video: AnimateDiff, video generation models

Setup Process:

1. Download Pinokio from pinokio.computer
2. Browse app library
3. Click "Install" on any app
4. Wait for automatic setup
5. Click "Run" to launch

Perfect For:

Non-technical users who want to try various AI tools without dealing with installation complexity.

Get Pinokio →

Google's open-source ML framework. Keras provides high-level API for rapid development.

Strengths:

• Production-Ready: Deploy to web, mobile, edge, cloud
• TensorFlow.js: Run models in browser with JavaScript
• TensorFlow Lite: Mobile and embedded deployment
• Large Ecosystem: Massive community, tutorials, pre-trained models

Best For:

Production deployments, mobile/web apps, researchers, teams needing Google ecosystem integration.

TensorFlow Docs →

Facebook's (Meta) ML framework. Dominant in research, increasingly popular in production.

Strengths:

• Pythonic: Feels natural for Python developers
• Dynamic Graphs: Easier debugging, more flexible models
• Research Favorite: Most cutting-edge papers use PyTorch
• Lightning: High-level wrapper for organized training

Best For:

Research, experimentation, computer vision, NLP, custom architecture development.

PyTorch Docs →

Python library for traditional machine learning. Simple, efficient, and perfect for classical ML tasks.

Algorithms Included:

• Classification: SVM, Random Forest, Logistic Regression
• Regression: Linear, Polynomial, Ridge, Lasso
• Clustering: K-Means, DBSCAN, Hierarchical
• Preprocessing: Scalers, encoders, feature selection

Best For:

Tabular data, traditional ML, quick prototypes, data analysis, when deep learning is overkill.

Scikit-learn Docs →

High-level PyTorch library making deep learning accessible. Opinionated best practices built-in.

Philosophy:

• High-Level API: Train state-of-the-art models in few lines
• Best Practices: Learning rate finder, progressive resizing, etc.
• Excellent Course: Free "Practical Deep Learning for Coders"
• Rapid Prototyping: From idea to working model quickly

Best For:

Learning deep learning, rapid experimentation, developers wanting quick results.

Fast.ai Docs →

The GitHub of machine learning. 200,000+ pre-trained models for NLP, computer vision, audio, and more.

What's Available:

• Transformers: BERT, GPT, T5, LLaMA, Mistral, and thousands more
• Diffusion Models: Stable Diffusion, ControlNet, image generation
• Datasets: 100,000+ ready-to-use datasets
• Spaces: Host ML demos and applications for free
• Inference API: Use any model via simple API

Why It's Essential:

Don't train from scratch when world-class models are free. Fine-tune for your specific needs.

Browse Hugging Face →

GPT-4, DALL-E, Whisper, and embeddings via simple API. Most advanced AI models available.

Models & Capabilities:

• GPT-4: Advanced language understanding and generation
• GPT-3.5-turbo: Fast, cost-effective, ChatGPT-quality
• DALL-E 3: State-of-the-art image generation
• Whisper: Best-in-class speech recognition
• Embeddings: Semantic search, recommendations, clustering

Pricing:

Pay-per-use. GPT-3.5-turbo: ~$0.001-0.002 per 1K tokens. GPT-4: ~$0.03-0.06 per 1K tokens.

OpenAI Platform →

Advanced language model with 200K context window. Excellent for analysis, summarization, coding.

Key Features:

• Massive Context: 200,000 tokens (~150,000 words)
• Constitutional AI: Safer, more aligned responses
• Longer Reasoning: Better at complex, multi-step tasks
• Document Analysis: Process entire books, codebases, reports

Claude API →

Run and deploy machine learning models in the cloud. Pay-per-use for thousands of open-source models.

Popular Models:

• Stable Diffusion: Text-to-image generation
• LLaMA: Open-source language models
• Whisper: Speech recognition
• CLIP: Image and text embeddings

Advantages:

No infrastructure management. Scale automatically. Use any model via API without setup.

Browse Replicate →

Models built specifically for Indian languages and contexts. Critical for building apps for Bharat.

Top Models:

• Sarvam-1: 2B parameter model optimized for Hindi, outperforms larger models.
• OpenHathi: First Hindi LLM built on Llama-2.
• IndicTrans2: State-of-the-art translation for 22 Indian languages.

Where to Find:

Available on Hugging Face and AI4Bharat Models.

50,000+ datasets covering every domain. Active community, competitions, and notebooks.

Categories:

• Computer Vision: ImageNet, COCO, OpenImages
• NLP: Reviews, social media, news articles
• Tabular: Finance, healthcare, e-commerce
• Time Series: Stock prices, sensor data, weather

Browse Kaggle →

100,000+ datasets optimized for ML. Easy to load with one line of Python.

Popular Datasets:

• GLUE/SuperGLUE: NLP benchmarks
• Common Voice: Multilingual speech
• LAION: Billions of image-text pairs
• The Stack: Code from GitHub

Datasets Hub →

Search engine for datasets across the web. Find academic, government, and open data.

Sources Include:

• Government: data.gov, census data, climate data
• Research: Academic datasets from papers
• Organizations: WHO, World Bank, UNESCO
• Companies: Open datasets from tech companies

Search Datasets →

Computer vision datasets with annotations. Over 200,000 datasets for object detection, segmentation, classification.

Features:

• Pre-Annotated: Bounding boxes, polygons, keypoints
• Multiple Formats: YOLO, COCO, Pascal VOC
• Community Datasets: Upload and discover projects
• Training Tools: Train models directly on platform

Roboflow Universe →

Jupyter notebooks in the cloud with free GPU/TPU access. No setup required—start coding immediately.

Features:

• Free GPUs: NVIDIA T4, up to 12 hours per session
• Pre-installed: TensorFlow, PyTorch, scikit-learn ready
• Google Drive Integration: Save notebooks, datasets
• Collaboration: Share notebooks like Google Docs

Perfect For:

Learning, prototyping, tutorials, training small-to-medium models without local GPU.

Launch Colab →

Cloud notebooks with free GPU/TPU. Access Kaggle datasets directly, participate in competitions.

Advantages:

• 30 hours/week GPU: More generous than Colab free tier
• Dataset Integration: Instant access to 50,000+ datasets
• Community: Learn from public notebooks
• Competitions: Practice ML on real problems

Kaggle Notebooks →

Local development environment for data science and ML. Industry standard for interactive Python development.

Why Use Local Jupyter:

• Full Control: Use your own hardware and environment
• No Time Limits: Run training as long as needed
• Privacy: Data never leaves your machine
• Extensions: Customize with plugins and themes

pip install jupyterlab to get started

Jupyter Docs →

Experiment tracking, model versioning, and collaboration for ML teams.

Capabilities:

• Experiment Tracking: Log metrics, hyperparameters automatically
• Visualizations: Real-time training curves, comparisons
• Model Registry: Version and deploy models
• Sweeps: Automated hyperparameter tuning

Free for individuals and small teams.

W&B Platform →