If you’re here to make your images faster without breaking a sweat, you’re in the right place! ComfyUI Flux Accelerator is a nifty tool designed for all of us who love ComfyUI but wouldn’t mind shaving a little time off our workflow. Here’s the quick and easy lowdown on what it does and how to get it running.
What is the ComfyUI Flux Accelerator?
The Flux Accelerator is a special add-on (we call it a “node”) for ComfyUI. Basically, it makes image generation up to 37.25% faster by using TAEF1, a speedy autoencoder. Now, TAEF1 does cut a tiny bit of quality for speed, but for many users, the trade-off is worth it—especially when you’re after fast results.
What Makes It Tick?
Behind the scenes, the accelerator is pulling some neat tricks to cut down time:
- Torch Optimization: Using tools like
andtorchao
(don’t worry, they’re just fancy speed-boosting techniques).torch.compile() - Skipping DiT Blocks: You have the option to skip certain parts of the generation process (DiT blocks) that aren’t always necessary. Just one click, and boom—less time on the clock.
Getting Started
- Installation: First, clone the ComfyUI Flux Accelerator repository and place it in ComfyUI’s
folder.custom_nodes - Requirements: Make sure you’ve got PyTorch and xFormers installed. You’ll also need to grab the TAEF1 file for the accelerator to work.
- Start Generating: Once everything’s set up, simply use the FluxAccelerator node in your ComfyUI projects and watch the magic happen!
A Note: Linux-only for now, and the first time setup (compiling the model) can take about 30-60 seconds. Just enough time to grab a coffee, right?
Why Use It?
If you’re looking for faster image outputs with ComfyUI, the Flux Accelerator is a great addition. It’s super easy to set up, and while you might notice a slight dip in quality, the time saved makes up for it. Try it out and see if it’s the game-changer you’ve been waiting for!
