🚀 Flux.1 Dev Hybrid-Workstation (GGUF)

Optimized for 12GB - 16GB VRAM Cards (e.g., RTX 3080/4080)

This is not just a simple generation workflow. It is a production pipeline designed for content creators who need to iterate fast but render in high quality.

💡 The Core Concept: "Draft vs. Render"

I designed this workflow with a Boolean Switch to solve the VRAM bottleneck. You don't need to load heavy models just to test a prompt.

🔴 Mode A: DRAFT (False)

- Loads a lightweight Q5 (or Q4) GGUF Model.

- Bypasses all Upscalers and Detailers.

- Goal: Rapid prompt engineering and composition testing.

🟢 Mode B: RENDER (True)

- Loads the heavy Q8 GGUF Model.

- Activates Upscale (x2).

- Activates FaceDetailer & Structure Detailers.

- Goal: Final output for publishing (Civitai/Socials).

---

🛠 Features

* Modular Bus System: Clean wiring using Set/Get nodes.

* Visual Stack: Real-time preview comparison between Raw, Upscaled, and Refined versions.

* Smart Save: Automatically organizes your files into folders based on project names.

* 16GB VRAM Safe: Uses GGUF quantification to fit Flux.1 Dev into consumer cards depending on the chosen quantization.

---

📦 Requirements (Files needed)

Please read the "Red Note" inside the workflow!

1. Models: You need both a Q5/Q4 and a Q8 GGUF version of your Flux Checkpoint (e.g., CyberRealistic Flux).

2. Clip/VAE: Standard Flux T5xxl (fp8), Clip_L, and AE.

3. Custom Nodes: If nodes are red, simply use ComfyUI Manager -> Install Missing Custom Nodes.

---

Happy Creating!

If you like this workflow, please leave a review and share your images!