Flux LoRA

​

1. Download Flux LoRA Model

There are many Flux LoRA models, most of which are shared on Hugging Face. You can find them through the search function on Hugging Face.

However, the search function on Hugging Face is not very good, you can search and filter through the following methods:

1. Click the Models button at the top.
2. Click Text to Image on the left.
3. Then, enter flux lora in the search box on the right. This will give you all Flux LoRA models.
4. Finally, you can click the Sort button on the right to switch the sorting method, such as Most Downloads.

Choose a Flux LoRA model you like, click into it, click the Files and versions button (① in the figure), then download the LoRA file (② in the figure). Then put the file in the /models/loras directory.

​

2. Flux LoRA Workflow

After downloading the model, let’s build the Flux LoRA ComfyUI workflow. First, you can load the full Flux workflow I recommended in Flux ComfyUI Workflow, and then modify it based on this workflow. Or go to Comflowy to download this workflow and use it.

The changes are relatively small, you just need to add a LoraLoaderModelOnly node (① in the figure), note that it is LoraLoaderModelOnly rather than LoraLoader. Because we only need to load the LoRA model, not the Clip model.

Then, select the Flux LoRA model you downloaded, you can adjust the strength of the LoRA model on the image generation by setting the strength_model parameter (② in the figure), the higher the value, the greater the impact. You don’t need to add any parameters related to LoRA in the Prompt, doing this has almost no effect.

If you don’t want to manually connect, you can go to Comflowy to directly use our Flux Dev LoRA app, which takes around 8 seconds to generate a single image. Or download this workflow template and import it into your local ComfyUI. (If you don’t know how to download Comflowy templates, you can refer to this tutorial)

​

3. Chain Multiple LoRA Models

Besides chaining a LoRA model, you can also chain multiple LoRA models. This way, the two LoRA models will stack up to affect the image generation, thus generating images that are more aligned with your expectations.

Just add multiple LoraLoaderModelOnly nodes, then select the Flux LoRA models you downloaded. For example, in the workflow below, I added two Flux LoRA models, one is Flux-Super-Realism-LoRA, the other is Flux-Midjourney-Mix2-LoRA. The former makes the generated images more realistic, and the latter makes the generated images more similar to Midjourney. The image will look more exquisite, and the visual style will also be better.

Another small trick is that you can bypass a LoRA model by right-clicking on the LoraLoaderModelOnly node and selecting Bypass. When you don’t need a LoRA model, doing this is much more convenient. No need to delete the node and reconnect: