Quick Start: From Zero to Your First AI Image

💡 Goal: Get ComfyUI installed within 60 minutes, put models in the right folders, run the default text-to-image workflow, and get your first AI image (model download time extra, depending on your network speed). This is the first stop of the entire ComfyUI tutorial — even complete beginners can follow along to the end.

🎯 Stage Goals

• Get ComfyUI running on your own computer: open http://127.0.0.1:8188 in your browser and see the node canvas
• Know which subdirectory each model file belongs in: checkpoints, VAEs, and LoRAs each go to their own home, so you'll never stare blankly at a "model not found" error again
• Complete your first text-to-image run: write your prompt, press Ctrl + Enter (on macOS it's Cmd + Enter), and get the finished image in ComfyUI/output
• Be able to help yourself if installation fails: startup failures, port conflicts, driver issues — you'll know which step to check first

⏳ Estimated Time

• Just want an image ASAP: head straight to 1.2 Install → 1.4 Models → 1.5 Generate, three lessons totaling about 35 minutes; want to understand the concepts first, add 1.1 for about 41 minutes total
• Complete this whole stage (including alternative installation 1.2b and troubleshooting 1.3): six lessons totaling about 63 minutes, roughly 45–80 minutes with buffer time
• The above excludes model download time: the first starter model (SD1.5) is about 4GB, extra depending on your network speed

🚀 Before You Start

• A Windows / Linux / macOS computer (Apple Silicon M-series Macs are officially supported)
• GPU: NVIDIA / AMD / Intel Arc / Apple Silicon are all on the official support list; without a discrete GPU you can still run in --cpu mode, just slower
• No need to worry about VRAM: officially, thanks to asynchronous weight streaming, 4GB of VRAM + 8GB of RAM can run the largest open-source models
• Disk space: the official desktop version requires at least 4.85GB of disk; add the first model at about 4GB, and having some spare space never hurts
• For the browser, Google Chrome 143 or higher is recommended (earlier versions have known rendering issues)
• A cup of coffee ☕ — from download to image, it's faster than you think

💡 The interface has an official Chinese option

Open Settings → Comfy → Locale and select Chinese to switch the interface language. Already installed ComfyUI? Jump straight to 1.5 First Generation; stuck on an error, check Troubleshooting Overview first.

📚 Lessons in This Stage

Lesson	Content	Duration
1.1 ComfyUI What Is It	Node-based workflows, what it can do, is it free, which models are supported	6 minutes
1.2 Install: Windows Portable Package	Download the 7z package, extract, double-click run_nvidia_gpu.bat to launch, open 127.0.0.1:8188 in your browser	10 minutes
↳ 1.2b Alternative Installation	Comfy Desktop; manual installation on macOS and Linux	12 minutes
1.3 What If Installation Fails	Startup failures, 127.0.0.1:8188 won't open, black screen, driver checks	10 minutes
1.4 Where to Put Models	ComfyUI/models/ directory structure, extra_model_paths.yaml sharing models with other tools	10 minutes
1.5 First Generation	The default text-to-image workflow explained node by node, KSampler parameters, how to write prompts	15 minutes

🎒 What You'll Gain

• A working ComfyUI environment that produces images, which all later model tutorials can reuse directly
• A hands-on feel for the full "load model → write prompt → sample → decode and save" pipeline, so other people's workflows won't look confusing anymore
• A mental map of the model directories: which subdirectory each checkpoint, VAE, and LoRA goes in, crystal clear
• A sense of direction after your first image: go to Models × Tasks Overview to pick models by task; e-commerce folks can head straight to E-Commerce Images in Practice

---

Ready? Start the first lesson: What is ComfyUI →