This is how I generate AI videos locally using ComfyUI
The article describes a workflow for generating AI videos locally using open-source tools and ComfyUI, without relying on cloud services.
Why it matters
This workflow demonstrates how users can leverage open-source AI tools to generate high-quality videos locally, without the need for cloud infrastructure or paid services.
Key Points
- 1Initial image generation using Z-Image Turbo
- 2High-quality upscaling with SeedVR2
- 3Video generation with Wan 2.2 FLF at lower resolution
- 4Final upscaling and interpolation for smoother motion
Details
The author uses a laptop with an RTX 4090 GPU (16 GB VRAM) and 32 GB of system RAM to generate AI videos locally. The workflow involves four main steps: 1) Generating a base image using Z-Image Turbo, 2) Upscaling the image to 2048 x 3840 resolution with SeedVR2, 3) Creating the video animation using Wan 2.2 FLF at a lower 816 x 1088 resolution to fit within the 16 GB VRAM, and 4) Upscaling and applying frame interpolation to the final video for smoother motion. The entire process is done locally within the ComfyUI tool, without relying on any cloud services.
No comments yet
Be the first to comment