<p>This workflow is designed to push FLUX.2 Klein text-to-image generations through a clean 4-stage quality refinement process. It’s built to be easy to follow, easy to tweak, and focused on producing polished final images without turning the graph into a total nightmare.</p><p>The workflow starts with a base FLUX.2 Klein generation, then moves into additional sampling refinement using res_2s, Detail Daemon, and a controlled Phase 2 denoise slider. From there it runs FaceDetailer for facial cleanup, then finishes with color balancing, color adjustment, Foolhardy Remacri upscaling, CAS sharpening, and a final resize to 1920px on the longest side.</p><p>Step-by-step previews are included for Phase 1, Phase 2, Phase 3, and the final Phase 4 output, along with an image comparer so you can directly compare the first pass against the finished render.</p><p>FLUX.2 Klein is already fast and flexible, but this workflow is built to give it a more polished final look with better detail, cleaner faces, sharper output, and a more finished editorial-style render.<br /></p><h3 id="node-packs-used-in-this-workflow">Node packs used in this workflow</h3><pre><code>rgthree-comfy

    Power Lora Loader (rgthree)

    Image Comparer (rgthree)

ComfyUI Impact Pack

    FaceDetailer

    SAMLoader

ComfyUI Impact Subpack

    UltralyticsDetectorProvider

ComfyUI KJNodes

    Patch Sage Attention KJ

    ModelPatchTorchSettings

    GetNode

    SetNode

Detail Daemon

    DetailDaemonSamplerNode

    DetailDaemonGraphSigmasNode

ComfyUI Easy Use

    easy seed

    easy int

    easy cleanGpuUsed

ComfyUI LogicUtils

    StaticNumberInt

    ResizeLongestToNode

MixLab Nodes

    FloatSlider

ComfyUI Essentials

    ImageCASharpening+

WAS Node Suite

    Upscale Model Loader

Core ComfyUI nodes

    UNETLoader

    CLIPLoader

    VAELoader

    CLIPTextEncode

    EmptyLatentImage

    KSamplerAdvanced

    KSamplerSelect

    BasicScheduler

    CFGGuider

    RandomNoise

    SamplerCustomAdvanced

    KSampler

    VAEDecode

    ImageUpscaleWithModel

    PreviewImage

    SaveImage</code></pre><h3 id="models-used">Models  Used</h3><pre><code>flux2-vae.safetensors

qwen_3_8b.safetensors

4x_foolhardy_Remacri.pth

face_yolov8m.pt

sam_vit_b_01ec64.pth

SageAttention 
</code></pre><p></p>

Files with SHA256: 2eb0f965208f6ca5b703430bcf54003130694846fc1d1647e7ea807ce2fe0a3b