# How to generate AI images and videos for free without internet (ComfyUI Tutorial)

Source: https://www.youtube.com/watch?v=xtwQWnIobTU
Recap page: https://rapidrecap.app/video/xtwQWnIobTU
Generated: 2025-08-07T05:31:05.515+00:00

---
## Quick Overview

This ComfyUI tutorial demonstrates how to install and use the software locally to generate AI images and videos, highlighting its flexibility and ease of use for various AI models and tasks, all without requiring an internet connection after setup.

**Key Points:**
- ComfyUI allows users to generate AI images and videos locally, without needing an internet connection after installation.
- The software is free, open-source, and highly customizable with custom nodes.
- Installation is straightforward on Windows and Mac, with options to download directly or install from GitHub.
- The tutorial covers loading diffusion models, CLIP text encoders, and VAEs, essential components for AI generation.
- Users can generate images from text prompts and also create videos by specifying start and end frames.
- The "missing models" error is addressed by guiding users to download necessary files from Hugging Face.
- The tutorial demonstrates the effectiveness of different models, including WAN 2.2, for both image and video generation tasks.

![Screenshot at 01:20: Screenshot of the ComfyUI node-based interface showing workflow connections, illustrating the visual programming approach used by the software.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-01-20.png)

**Context:** This video tutorial focuses on ComfyUI, a free and open-source application for generative AI that runs locally on your computer, eliminating the need for an internet connection after the initial setup. The presenter guides viewers through the process of installing and using ComfyUI, highlighting its node-based workflow and its versatility in generating both images and videos.

## Detailed Analysis

This video provides a comprehensive guide on installing and utilizing ComfyUI, a powerful open-source node-based application for generative AI, on your local machine. The tutorial covers the initial setup, including downloading the software from their website (comfy.org) or GitHub, and explains the installation process for both Windows and Mac. It emphasizes that ComfyUI runs locally, offering greater control, faster iteration, and lower costs compared to cloud-based solutions. The presenter walks through the ComfyUI interface, explaining the node-based workflow for generating images and videos. They demonstrate how to load diffusion models, CLIP text encoders, and VAEs, as well as how to set image sizes and configure various sampler settings. The tutorial highlights the "missing models" error that can occur and shows how to download the necessary models from Hugging Face. Finally, it showcases the generation of sample images and videos, including a "New York City skyline at night" image, a "White House in winter" image, an "Eiffel Tower at sunset" image, and a "Bigfoot taking a selfie" image, illustrating the software's capabilities and the quality of results achievable with local processing.

### Installation

- Download from comfy.org or GitHub
- Choose installation location
- Migrate existing files (optional)
- Configure desktop app settings (automatic updates, usage metrics)

### Model Loading

- Load diffusion models (e.g., FLUX)
- Load CLIP text encoders
- Load VAE

### Image Generation

- Set image size (e.g., 1024x1024)
- Input positive and negative prompts
- Use KSampler settings (steps, cfg, sampler, scheduler)
- Save image

### Video Generation

- Utilize Wan 2.2 models (Text-to-Video, Image-to-Video, First-Last Frame-to-Video, 5B Video Generation)
- Select start and end frames
- Adjust video size and length (width, height, length, batch size)
- Save video

### Troubleshooting

- Address "Missing Models" error by downloading necessary files
- Resolve "Float8_e4m3fn to MPS backend" error on Mac by using compatible models (e.g., FP16)
- Note performance differences between Mac (MPS) and PC (NVIDIA GPU) setups

### Key Features

- Runs locally, fast, and free
- Open-source and customizable with custom nodes
- Generates diverse AI outputs (images, videos)
- Offers pre-built templates for quick start

![Screenshot at 00:00: Demonstration of Donald Trump interacting with people at an event.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-00-00.png)
![Screenshot at 00:13: On-screen graphic showing a dinosaur running from cacti, mimicking an internet outage game.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-00-13.png)
![Screenshot at 00:33: Logos of Google, Microsoft, and OpenAI appear on screen, symbolizing major tech companies.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-00-33.png)
![Screenshot at 01:08: Logo for ComfyUI appears on screen.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-01-08.png)
![Screenshot at 01:20: Screenshot of the ComfyUI node-based interface showing workflow connections.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-01-20.png)
![Screenshot at 01:26: Detailed view of a ComfyUI workflow graph with multiple connected nodes.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-01-26.png)
![Screenshot at 02:07: Screenshot of the ComfyUI website homepage with the title 'ComfyUI' and a download button.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-02-07.png)
![Screenshot at 02:15: ComfyUI website showcasing features like 'Full Control with Nodes' and 'Reusable Workflows'.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-02-15.png)
![Screenshot at 02:26: ComfyUI website highlighting key features: 'It's Free and Open Source', 'Custom Nodes', and 'Runs Locally, Fast'.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-02-26.png)
![Screenshot at 02:41: ComfyUI website download section showing options for Windows and Mac, and 'Install from Github'. The presenter's Mac download is highlighted implicitly by the subsequent steps shown on a Mac interface.](https://ss.rapidrecap.app/screens/xtwQWnIobTU/00-02-41.png)