imageBeginnerv1.0

FLUX Dev Text-to-Image

Maximum quality text-to-image with FLUX.1 Dev using separate model loaders for 16GB VRAM cards.

About this workflow

The definitive FLUX Dev workflow. Uses UNETLoader + DualCLIPLoader (T5-XXL + CLIP-L) for the best possible image quality. Tuned for RTX 5080 with fp16 precision. Perfect starting point for portrait, landscape, and concept generation.

FLUXtext-to-imageportraitphotorealistic16GB

Required Model

flux1-dev.safetensors

Place in: ComfyUI/models/checkpoints/

Benchmark (RTX 5080)

Gen Time

~22s on RTX 5080

Resolution

1024×1024

Hardware Profile

Settings adjust to your GPU tier

16GB Recommended — Specs

Resolution1024×1024

Steps25

CFG7

Samplerdpmpp_2m

Schedulerkarras

Flags--gpu-only

Full quality. Recommended tier for this workflow.

Export Package

✓ Real ComfyUI node graph JSON

✓ Hardware settings baked in

✓ Launch flags in metadata note

✓ Load directly in ComfyUI

// Hardware

Not sure your rig can handle this? Plan your upgrade on ComputeAtlas.

Hardware Estimator →