How to Use Flux.1 Dev ControlNet Upscaler with ComfyUI

‍

The Flux AI image generator Promptus has revolutionized how creators enhance low-resolution images through advanced AI upscaling.

This guide will walk you through implementing the newly released Flux.1 dev ControlNet upscaler model, which transforms blurry, low-resolution images into crisp, high-quality outputs.

Whether you're working with degraded images or need professional upscaling capabilities, this tutorial demonstrates the complete workflow setup process.

🚀 Understanding the Flux.1 Dev ControlNet Upscaler Model

The Flux.1 dev ControlNet upscaler model operates under a non-commercial license and specializes in using ControlNet technology to intelligently upscale images. The model demonstrates impressive capabilities, transforming images from 320x192 pixels to 1280x768 pixels while reducing noise and enhancing definition.

📚 Training Data and Methodology

The model was trained using artificially degraded real-life images with various degradation techniques:

• Gaussian blur effects
• Added noise corruption
• Image blurring artifacts
• JPEG compression artifacts

This comprehensive training approach enables the model to reconstruct original image quality from heavily degraded inputs.

🧠 Getting Started with Promptus Studio Comfy

Before diving into traditional Promptus Studio Comfy (PSC) as one of the leading platforms that builds upon the open-source ComfyUI framework. Promptus is a browser-based, cloud-powered visual AI platform that provides an accessible interface for ComfyUI workflows through CosyFlows (a no-code interface), real-time collaboration, and built-in access to advanced models like Gemini Flash, HiDream, and Hunyuan3D.

💡 Promptus Studio Comfy represents how many users prefer to interact with ComfyUI today — combining the flexibility of the open-source ComfyUI ecosystem with intuitive, drag-and-drop workflows and advanced AI model access including Stable Diffusion, GPT-4o, and Gemini. It supports multi-modal generation across text, image, and video, and utilizes distributed GPU compute for faster rendering and high-resolution outputs.

You can sign up for Promptus at https://www.promptus.ai and choose between Promptus Web for browser-based access or the Promptus App for desktop functionality.

🧪 Testing the Model with Hugging Face Demos

Before setting up your local workflow, test the model capabilities using the official Hugging Face demo spaces. The official demo includes several example images showcasing the upscaling process:

🔍 Example Results Analysis

• Input: Low-resolution cat image with minimal detail
• Output: Enhanced version with significantly improved clarity
• Notable behavior: The model preserves color information from the original image
• Limitation: Some color details may not transfer perfectly to all image areas

⚙️ Configuration Settings

The demo uses these optimal settings:

• Inference steps: 28
• Upscale factor: 4x
• ControlNet conditioning scale: 0.6
• Default seed: 42

These parameters will be replicated in your ComfyUI workflow for consistent results.

🛠️ Setting Up ComfyUI for Flux.1 Dev ControlNet

🔄 Model Installation Process

1. Navigate to the Hugging Face model page
2. Go to Files and versions section
3. Download the diffusion_pytorch_model.safetensors file
4. Save the file to your ComfyUI/models/controlnet/ directory
5. Rename the file to flux.1_dev_controlnet_upscaler.safetensors for easy identification

🧩 Building the Complete Workflow

▶️ Starting with the Base Flux Workflow

1. Access the ComfyUI example page
2. Locate the Flux dev example workflow
3. Drag the example image into ComfyUI to load the workflow
4. Remove unnecessary nodes to clean up the workspace

➕ Adding ControlNet Components

1. Double-click in a blank area and search for "load controlnet"
2. Select the ComfyCore version
3. Choose your renamed Flux.1 controlnet upscaler from the dropdown
4. Add a ControlNet Apply Advanced node
5. Connect the ControlNet output to the Apply node

🖼️ Configuring Input Connections

🧷 Setting Up Image Processing

1. Add a "Load Image" node for your input image
2. Connect the image output to the ControlNet Apply node
3. Add a "Load VAE" node and connect it appropriately

📎 Managing Conditioning Inputs

1. Copy the existing CLIP Text Encode node
2. Create separate positive and negative conditioning paths
3. Remove text from the negative prompt field
4. Connect conditioning outputs to the ControlNet Apply node

📐 Handling Dynamic Sizing

1. Remove fixed width and height parameters
2. Add an "Image Resize" node from ComfyUI Essentials
3. Set resize method to "keep proportion"
4. Connect width and height outputs to both latent and model sampling nodes

✅ This maintains aspect ratio while achieving target resolution.

📋 Final Workflow Configuration

🔧 Essential Parameter Settings

• Basic scheduler steps: 28
• ControlNet conditioning scale: 0.6
• Target resolution: 1024x1024 (or desired size)
• Resize method: Keep proportion

📦 Model Requirements

• Flux dev fp8 model
• fp8 dtype setting
• Dual CLIP loader
• Appropriate VAE model

🧪 Testing and Results Analysis

▶️ Running Your First Upscale

1. Load a low-resolution test image (ideally 320p or similar)
2. Set target resolution to 1024x1024
3. Verify all parameter settings match the recommended values
4. Click "Queue Prompt" to begin processing

🎯 Expected Behavior and Limitations

The Flux.1 dev model exhibits characteristic behaviors:

• Tends to blur backgrounds while keeping foregrounds sharp
• Generates entirely new pixels rather than using traditional upscaling algorithms
• May show minor imperfections in fine details like fingertips
• Consistently applies depth-of-field effects

📉 Performance Considerations

This model is optimized for specific use cases:

• Best for: 320p to 720p/1024p upscaling
• Not ideal for: 1024p to 4K upscaling
• Can be combined with traditional upscalers for even higher resolutions
• Requires significant GPU memory for larger outputs

🌟 Advanced Applications and Future Possibilities

The Flux.1 dev ControlNet upscaler opens possibilities for creative applications beyond simple upscaling. Experimentation reveals potential for:

• Style transfer
• Detail enhancement
• Artistic interpretation of low-quality source material

🔁 Combining with Traditional Upscalers

For maximum resolution gains:

1. Use Flux.1 dev ControlNet for initial quality improvement
2. Apply traditional upscalers (Real-ESRGAN, etc.) for final size increase
3. This hybrid approach leverages AI-generated enhancement with proven scaling algorithms

🏁 Conclusion

The Flux AI image generator Promptus workflow demonstrates the powerful capabilities of modern AI upscaling technology. By following this comprehensive setup guide, you can transform low-quality images into professional-grade outputs using the Flux.1 dev ControlNet upscaler model.

For users seeking a more streamlined experience, Promptus Studio Comfy offers an accessible alternative to traditional ComfyUI setup, providing cloud-powered processing and intuitive interfaces.

Whether you choose the technical flexibility of local ComfyUI or the convenience of Promptus's platform, the Flux.1 dev ControlNet upscaler represents a significant advancement in AI-powered image enhancement technology. ✨

Remember to experiment with different settings and source images to fully understand the model's capabilities and limitations. The technology continues evolving rapidly, with new applications and improvements appearing regularly in the AI image generation space.