Local image generation using ComfyUI

Last week, I started experimenting with AI image generation locally using ComfyUI - and honestly, it’s incredible how accessible high-quality results have become:

✅ Runs entirely on my local workstation

✅ Web interface makes workflow building straightforward

✅ Full control over models, prompts, and outputs

My main workstation is about 4 years old:

• AMD 5900X CPU
• RTX 3080 12GB GPU
• 64 GB DDR4 3600 MHz RAM

However, high-quality image generation only takes about 3 minutes!

The setup I’m experimenting with:

• ComfyUI for node-based workflow building
• Flux-dev-1 (quantized to FP8) as the main image generation model
• Dual CLIP models (T5XXL + CLIP-L) for stronger prompt understanding and guidance
• Realistic LoRA extension stacked on Flux for enhanced realism
• Latent upscaling combined with 4x-ESRGAN for high-resolution outputs (1024x1024 ➡️ 6144x6144)

Building the workflow in ComfyUI is very intuitive (see attached workflow): You can visually chain nodes to encode prompts, generate images, apply guidance, upscale, and save - all without touching Python scripts.

I attached a few example images I generated.

👉 Next step: Train my own LoRA to get more personalized outputs!