Back to Workflows
FLUXKleinPortraitCharacterConsistent

FLUX Klein v1 Portrait

Optimized for consistent character portraits. Uses FLUX-Klein UNET with Qwen CLIP at 8 steps CFG 1.0 for high-quality 832x1216 outputs.

Download JSON
Download this workflow JSON to import into ComfyUI
Browse All Workflows
ComfyUI node graph preview for FLUX Klein v1 Portrait workflow — Optimized for consistent character portraits. Uses FLUX-Klein UNET with Qwen CLIP at 8 steps CFG 1.0 for high-quality 832x1216 outputs.
Model
FLUX Klein v1
Type
Portrait
Difficulty
Beginner
Resolution
832x1216

Settings

sampler euler_cfg_pp
steps 8
cfg 1
resolution 832x1216
seed 42
model FLUX1/flux-2-klein-9b.safetensors

Prompts

portrait of a young woman with blue eyes, natural lighting, detailed skin texture, professional photography
text, watermark, blurry, ugly, deformed, noisy, distorted

Sample Outputs

Real outputs generated with this workflow — click to enlarge.

Sample output from FLUX Klein v1 Portrait workflow Sample output from FLUX Klein v1 Portrait workflow Sample output from FLUX Klein v1 Portrait workflow Sample output from FLUX Klein v1 Portrait workflow Sample output from FLUX Klein v1 Portrait workflow Sample output from FLUX Klein v1 Portrait workflow Sample output from FLUX Klein v1 Portrait workflow Sample output from FLUX Klein v1 Portrait workflow

What's Included

  • Consistent character portraits across multiple poses
  • Natural lighting and skin texture
  • Professional photography style
  • Optimized for 24GB VRAM

How to Use

  1. 1
    Download the JSON file using the button above
  2. 2
    Open ComfyUI in your browser
  3. 3
    Drag the JSON file into the ComfyUI canvas, or click "Load" and select the file
  4. 4
    Queue Prompt and watch the generation happen

Ready to Use This Workflow?

Download the JSON and import it into ComfyUI. All settings are pre-configured for consistent results.

Download flux-klein-v1-portrait.json

Related Workflows

FLUX Klein Identity Transfer Edit

Klein IdentityFeatureTransfer edit pipeline using ReferenceLatent conditioning and Flux2Scheduler. Preserves scene structure while applying text-guided edits — ideal for chair fabrics, background changes, and adding/removing subjects.

Z-Image Turbo Official

Official Z-Image-Turbo workflow using AuraFlow sampling, 8 steps at CFG 1.0. Uses qwen_3_4b text encoder with lumina2 type.

SDXL 6-Sampler Comparison

Side-by-side comparison of 6 samplers (euler_cfg_pp, dpmpp_2m, DDIM, LMS, dpmpp_2m_karras, dpmpp_2m_sde_karras). Same seed, same prompt — see which sampler wins.