How to run FLUX.1 Dev in Forge WebUI on Runpod?

Flux is not just another AI image generator; it's a groundbreaking text-to-image model developed by the innovative minds at Black Forest Labs. This model family has rapidly gained traction among digital artists and AI enthusiasts alike, thanks to its astounding capability of converting simple text prompts into high-quality images. Luxuriating in its features is the Flux 1 family, which includes three distinct models:

• Flux 1 Pro: The powerhouse of the group, offering top-notch image generation through APIs. While direct downloads aren't possible, its performance is unrivaled.
• Flux 1 Dev: This version comes with open weights, allowing developers to explore and experiment freely, although it's not designed for commercial endeavors.
• Flux 1 Schnell: The most accessible model, open-source and available on platforms like Hugging Face. It's perfect for personal projects and integration with tools such as Diffusers or Comfy UI. Try out Flux Schnell for free now!

In essence, Flux stands as a formidable tool in the world of AI-driven creativity, inviting everyone—from digital artists to tech enthusiasts—to explore its vast potential.

Ready to dive into the world of Flux? Let’s get you started! To run the Flux.1 Dev model inside Forge Web UI on RunPod, follow these simple steps:

1. Create Your RunPod Account:

   Ensure you load at least $10 into your account to kick things off. This can be done from the account billing page section.

2. Choose the Forge Flux Template:

   We’ve created a Forge Flux Template designed specifically for running the FLUX.1 Dev model within the Forge Web UI. Clicking the link above will take you to the "Deploy GPU Pod" page.

   There are two types of Pods available: Secure Cloud and Community Cloud. Secure Cloud operates within T3/T4 data centers, ensuring high reliability and security. On the other hand, Community Cloud connects individual compute providers and consumers via a secure, vetted peer-to-peer network. For a more affordable option to start, I recommend selecting Community Cloud from the top dropdown menu.

3. Select Your GPU Configuration:

   For the best performance with Flux, select a GPU configuration with at least 24GB VRAM, but preferably 48GB VRAM (e.g., RTX A6000 or A40).

4. Configure Template & Settings
   If you've clicked the Forge Flux Template link from step 2, you should see the "Next Diffusion - Forge Flux" Pod Template being selected:

   You have the option to click Edit Template if you need to adjust settings such as Container Disk Size or Volume Disk. If you plan on running long sessions or generating large batches of images, you might want to increase the Volume Disk for additional storage.

   

5. Deploy GPU POD

   • Once you're happy with your configuration, scroll down and click Deploy.
   • After deployment, head to the My Pods section to track the initialization of your Flux Forge Template pod. This will only take a few minutes. This process essentially sets up the Forge WebUI skeleton, preparing everything for use. Once the setup is complete, you’ll be able to click the "Connect" button.

   

6. Downloading FLUX.1 Dev Model
   By default, no models are included. You need to download them manually. Once your pod is running, you’ll need to download the Flux.1 Dev model. Here’s how to do it:

   • Click on the "Connect" button. This will open up a dialog, from here make sure to click Start and Open Web Terminal:

   

7. Web Terminal Commands to Download FLUX.1 dev Model

   To download the Flux.1 dev model via the Web Terminal, follow these steps:

   • Navigate to the models folder by pasting the following command into the terminal, then pressing Enter:
     cd /workspace/stable-diffusion-webui-forge/models/Stable-diffusion/

   • Download the model by pasting and entering this command:
     wget https://huggingface.co/lllyasviel/flux1_dev/resolve/main/flux1-dev-fp8.safetensors

   ⚠️ Note: Using Ctrl + V to paste does not work inside the Web Terminal. Use right-click > Paste instead.

   The Flux 1 Dev model (17.2GB) will start downloading, and the time it takes can vary. For me, it took about 8 minutes, but the download speed may differ depending on the pod configuration you select.

8. Start Forge WebUI (Port 7860)
   Once the model has finished downloading, you can close the Web Terminal and click the "Connect" button again. Then, select the HTTP Service --> :7860. This will open the Forge Web UI. Ensure that "flux" is selected as the UI option in the top left, and the flux1-dev-fp8.safetensors model should be available in the dropdown checkpoint options.

9. Start Generating Images
   With everything set up, try inputting a detailed prompt and click Generate. For example:

   Prompt: detailed, high quality, realistic, 4k, Egyptian woman, looking at viewer, upper body shot, brunette, cleavage, hazel light eyes, ancient artifacts , inside pyramid, swimming pool, waist and up, cleavage, looking at viewer, warrior Egyptian outfit, sensual, sensual outfit, mysterious
   Dimensions: 1344 x 768 (16:9 ratio)
   
   This generation finished in just 18 seconds—how awesome is that?! 🚀 Keep in mind, though, that the very first render might take a few extra seconds.

You might be wondering where your images are stored. Similar to the Forge UI, they are saved in the outputs folder. To access them, click the Connect button again and open Port 7777 (VS Code). From there, you can find the outputs/txt2img-images folder in the left sidebar.

Here, you can view all of your generated images and manage them as needed. You have the option to download your creations, delete unwanted files, or organize them into folders for better accessibility. Additionally, if you have models that you've already downloaded, you can easily drop them into the appropriate folders. Keep in mind that the speed of this process depends on your own upload speed. For larger files, it's recommended to use the Web Terminal Approach to download models directly, as it can be much more efficient. This ensures smooth management of your content and models within the Forge WebUI environment.

Once you’ve finished generating images, it's a good idea to stop the pod to avoid incurring hourly charges. To do this, simply return to the "Pods" page, select the pod you're currently running, and click the "STOP" button. This will prevent additional hourly charges for the GPU instance.

However, keep in mind that even after stopping the GPU instance, you will still be charged for your Pod Volume Storage (in this case, 25GB). To completely terminate the pod this make sure the click the "More Actions" button and then on "Terminate Pod":

Stopping vs. Terminating a Pod – What You Need to Know

Terminating the Pod

Terminating the pod will fully shut down your running GPU instances and erase everything associated with it. This means:

• The Flux Dev Model/Checkpoint will be deleted.
• All your output images will be lost.
• You'll need to redo the entire setup process the next time you start the pod.

However, if you stop the pod instead of terminating it, you can avoid all of that.

Stopping the Pod

When you click "Stop", here’s what happens:

• Your GPU instance shuts down, but the pod stays active.
• Your downloaded models, extensions, and images are saved.
• Next time you start the pod, you can skip the major setup/booting and get right back to work.

Key Differences

If you want to save time and keep your files, stopping is the better option. If you prefer a fresh start and don’t need to keep anything, terminate the pod.

In conclusion, our Flux Forge template on Forge WebUI is an incredibly powerful template for generating high-quality AI images. Whether you're an artist, developer, or content creator, Flux unlocks endless creative possibilities, from personal projects to commercial applications. By using the Flux 1 Dev model on RunPod’s Forge WebUI, you can easily tap into the full potential of this groundbreaking technology. Ready to start generating stunning AI images? Head over to RunPod, follow this guide, and experience the creative power of Flux!