Skip to content
WorkflowListedExternal
Open source

HR Beautify ComfyUI workflow (SDXL): refine after upscale and go as high as 24K — v1.0

HR Beautify (‘HR’ stands for Hires, or High resolution) is an advanced ComfyUI workflow to refine raw upscaled images. It doesn’t include the usual upscale with AI model component, concentrating instead on the best...

Sourcecivitai.com/models/1608027
TaskImage to imageFamilyStable DiffusionEstimated VRAMNot yet characterizedBgrade

Runtime profile

Access
External workflow
Status
Listed
Source
Civitai
Updated
Oct 2, 2026

Source description

HR Beautify (‘HR’ stands for Hires, or High resolution) is an advanced ComfyUI workflow to refine raw upscaled images. It doesn’t include the usual upscale with AI model component, concentrating instead on the best refinement of upscaled hires images possible, optionally beautifying them with a style transfer.

The workflow includes two ControlNets, for the most powerful and flexible img2img guidance, and two IP Adapters, to enhance images with style transfer and composition or reference guidance. For extra refinement, the Power Lora Loader node is included, supporting a potentially unlimited number of detail or style LoRas. Also included in the workflow are three post processing components: enhancing image’s HDR range, advanced image sharpening and color-matching.

The tiling component that enables hires img2img refining is the widely used Tiled Diffusion (TD), together with Tiled VAE Encode & Decode. TD is compatible with all ControlNet models and can assist in processing images of up to 24K resolution, optionally adding creative detail in the output along the way.

The workflow includes a note with description of the user-controlled parameters. Full description of the HR Beautify workflow and custom nodes it uses can be found on the github wiki pages of the project. To view more examples of hires images refined with the HR Beautify workflow and download full-size png files embedding it, click here .

What you need

Hardware and download-size estimates are not available yet.

Requirements

Models

8

  • CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors

    Not resolved

    Vision model · Unknown

  • dreamshaperXL_lightningDPMSDE.safetensors

    Not resolved

    Checkpoint · Unknown

  • ip-adapter_sdxl_vit-h-plus.safetensors

    Not resolved

    Checkpoint · Unknown

  • ip-adapter-plus_sdxl_vit-h.safetensors

    Not resolved

    Checkpoint · Unknown

  • Kataragi_inpaintXL-fp16.safetensors

    Not resolved

    ControlNet · Unknown

  • VAEDecodeTiled

    Possible

    VAE · Unknown

  • VAEEncodeTiled

    Possible

    VAE · Unknown

  • xinsirtile-sdxl-1.0.safetensors

    Not resolved

    ControlNet · Unknown

Node packs

37

  • ACN_AdvancedControlNetApply_v2

    Possible

    Node pack · Unknown

  • ACN_ScaledSoftControlNetWeights

    Found

    Node pack · Unknown

  • CheckpointLoaderSimple

    Possible

    Node pack · Unknown

  • CLIPTextEncode

    Possible

    Node pack · Unknown

  • CLIPVisionLoader

    Possible

    Node pack · Unknown

  • ControlNetApplyAdvanced

    Possible

    Node pack · Unknown

  • ControlNetLoader

    Possible

    Node pack · Unknown

  • CR Clamp Value

    Found

    Node pack · Unknown

  • Fast Bypasser (rgthree)

    Not resolved

    Node pack · Unknown

  • Fast Groups Bypasser (rgthree)

    Not resolved

    Node pack · Unknown

  • FromBasicPipe

    Found

    Node pack · Unknown

  • GetImageSize+

    Found

    Node pack · Unknown

  • Image Comparer (rgthree)

    Not resolved

    Node pack · Unknown

  • ImageBlender

    Found

    Node pack · Unknown

  • ImageResize

    Possible

    Node pack · Unknown

  • ImageScale

    Possible

    Node pack · Unknown

  • ImageSmartSharpen+

    Found

    Node pack · Unknown

  • ImpactBoolean

    Found

    Node pack · Unknown

  • ImpactConditionalBranch

    Found

    Node pack · Unknown

  • IPAdapterAdvanced

    Possible

    Node pack · Unknown

  • IPAdapterEmbeds

    Possible

    Node pack · Unknown

  • IPAdapterEncoder

    Possible

    Node pack · Unknown

  • IPAdapterModelLoader

    Possible

    Node pack · Unknown

  • LayerFilter: HDREffects

    Found

    Node pack · Unknown

  • LoadImage

    Possible

    Node pack · Unknown

  • Power Lora Loader (rgthree)

    Not resolved

    Node pack · Unknown

  • SaveImage

    Possible

    Node pack · Unknown

  • SetUnionControlNetType

    Not resolved

    Node pack · Unknown

  • SimpleMath+

    Possible

    Node pack · Unknown

  • TiledDiffusion

    Possible

    Node pack · Unknown

  • ToBasicPipe

    Found

    Node pack · Unknown

  • workflow>Control image group

    Not resolved

    Node pack · Unknown

  • workflow>Control mask image

    Not resolved

    Node pack · Unknown

  • workflow>KSampler (Advanced) group node

    Not resolved

    Node pack · Unknown

  • workflow>Reference image

    Not resolved

    Node pack · Unknown

  • workflow>Resize back logic

    Not resolved

    Node pack · Unknown

  • workflow>Style image

    Not resolved

    Node pack · Unknown

AI-generated commentary

AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.

Purpose

  • HR Beautify is a ComfyUI workflow for refining already-upscaled images with hires img2img processing. It takes an upscaled image, with optional control, reference, or style images, and produces a refined image; AI upscaling is not included.

Flow

  • The workflow combines Tiled Diffusion and tiled VAE encode/decode with two ControlNets and two IP Adapters for img2img guidance.

  • Tiled Diffusion can assist with images up to 24K resolution and can optionally add creative detail.

Inputs and outputs

  • The main input is a raw upscaled image; optional inputs include control images or masks, a reference image, and a style image.

  • The output is a refined image that can include style transfer, HDR enhancement, sharpening, color matching, or creative detail.

Setup

  • The workflow uses SDXL 1.0 as its base model and is published by Minsky91.

Dependencies

  • The listed model files include CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors, dreamshaperXL_lightningDPMSDE.safetensors, ip-adapter_sdxl_vit-h-plus.safetensors, and ip-adapter-plus_sdxl_vit-h.safetensors.

  • It also lists Kataragi_inpaintXL-fp16.safetensors and xinsirtile-sdxl-1.0.safetensors as ControlNet files.

  • Listed node components include TiledDiffusion, VAEEncodeTiled, VAEDecodeTiled, and Power Lora Loader (rgthree).

Configuration

  • You can use optional style transfer, composition or reference guidance, and detail or style LoRAs.

  • Tiled Diffusion is compatible with all ControlNet models, and the workflow includes a note describing user-controlled parameters.

Cautions

  • Because AI upscaling is not included, start with an already-upscaled image.

    Suggestion · not verified

  • Before use, check that ComfyUI has TiledDiffusion, VAEEncodeTiled, VAEDecodeTiled, and Power Lora Loader (rgthree).

    Suggestion · not verified

Does this need to be edited?

Sign in to send an edit request.

Sources

1 source
CivitaiPrimary
civitai.com/models/1608027

Verified Jul 19

Source excerpts

1 excerpt
NoteJul 19civitai.com/models/1608027

Source context: 397 downloads · Type Workflows · Base model SDXL 1.0