ad
ad

Flux.1 Dev Controlnet Upscaler Workflow Update | ft. Flux Lora and Florence2

Howto & Style

Introduction

Welcome back to the channel, AI Motion Studio! Today, I’m thrilled to introduce the latest update from Jasper AI: the brand new Flux.1 Dev Controlnet for upscaling images, which was just released a couple of days ago. This cutting-edge control net is unlike anything we've seen because it was trained with a synthetic complex data degradation scheme.

What Does This Mean?

Instead of relying solely on real-life images, the Flux.1 model was trained with artificially degraded versions that combine several forms of degradation, including Gaussian noise, image blurring, and JPEG compression. This advanced model is capable of transforming blurry, noisy, or compressed images into high-resolution perfection.

In this article, we'll take a closer look at the Flux Controlnet upscaler workflow, guide you through the setup process, and share tips to maximize your results. Whether you’re interested in restoring old photographs or upscaling AI-generated images, this tool is essential.

Updated Workflow

If you've been following our channel, you may already know that we previously introduced the Flux Dev 1 Lora upscaler. In our latest release, we have updated that workflow to integrate the new Flux Controlnet upscaler.

Where to Download

To download the latest workflow, head over to cit. and search for AI Motion Studio. There, you’ll find the Flux.1 Dev Controlnet upscaler workflow update. This update also features Flux NF4, Flux Lora, and Florence 2. Be sure to grab version 1.5 for access to the new control net version alongside the Lora version 1.0.

Setting Up the Workflow

Before diving into image upscaling, let’s quickly go through the setup process:

  1. Download the necessary models: Check the notes section for all links required for downloading.

  2. Model Storage:

    • Save the Lora model in the Lora folder under models.
    • Place the Controlnet model in the Controlnet folder.
    • The upscaler model should go into the upscaling models folder.
    • Make sure your base model, Flux NF4, is saved in the checkpoints folder.

Confirm that all files are correctly placed before proceeding with the workflow.

Image Upscaling Demo

Now, let's explore some practical examples of image upscaling.

Example 1: Enhancing a Blurry Image

We started with a low-quality blurry image and checked its quality, particularly noting the lack of detail around the eyes. After uploading and processing the image, the results were astounding! The face became sharp, the eyes improved in resolution, and overall quality significantly enhanced.

Example 2: Upscaling a Compressed Image

Next, we took a mid-journey generated image and compressed it down to 80%, significantly reducing its quality. After running it through the upscaling process, the results were impressive; the image appeared sharp with realistic eyes and textured fabric—an outcome one would not expect from an 80% compressed image.

Additional Tips

For those who prefer to maintain a grainy, natural look in their images, you can enable the pro post-film grain node. Conversely, if you dislike the grain, this node can also be bypassed.

If you encounter out-of-memory errors due to a low-end GPU with limited RAM, it's advisable to bypass the apply Controlnet and load Controlnet model nodes. This still leverages Lora for upscaling while delivering excellent results.

Lastly, if you face any issues with the workflow or specific nodes aren’t functioning correctly, simply navigate to the Comfy UI manager, click on “install missing custom nodes,” and hit the install button. Remember to update everything and restart Comfy UI afterward.

Conclusion

That wraps up our comprehensive walkthrough of the Flux.1 Dev Controlnet upscaler workflow. If you have any questions, feel free to leave them in the comments below. If this article was helpful, don’t forget to like and subscribe for more content. Thanks for reading, and we look forward to seeing you in the next one!

Keywords

FAQ

What is the Flux.1 Dev Controlnet?

The Flux.1 Dev Controlnet is a new image upscaling model trained with synthetic complex data degradation to enhance low-quality images to high-resolution.

How do I download the Flux.1 Dev Controlnet workflow?

You can download it from cit. by searching for AI Motion Studio. Make sure to grab version 1.5 for the newest features.

What should I do if I encounter an out-of-memory error?

If you are using a low-end GPU with limited RAM, it’s recommended to bypass certain nodes like the apply Controlnet and load Controlnet model nodes.

Can I keep grain in my images?

Yes, you can enable the pro post-film grain node for a more natural look, or bypass it if you prefer a smoother image.

What should I do if some nodes aren't working?

Head to the Comfy UI manager, click on 'install missing custom nodes,' and then hit install. Update everything and restart Comfy UI to resolve issues.