Skip to content
WorkflowListedExternal
Open source

DMD2 essential upscaler/refiner workflow — v1.0

There's few DMD2 upscale/refine workflows on this site but all are overcomplicated. Here's the essential simplest workflow presenting how powerful it is. I'm using this stuff at end of all image generation because...

Sourcecivitai.com/models/1278122
TaskMultiCapabilitiesImage to image · UpscalingFamilyStable DiffusionEstimated VRAMNot yet characterizedBgrade

Runtime profile

Access
External workflow
Status
Listed
Source
Civitai
Updated
Aug 19, 2026

Source description

There's few DMD2 upscale/refine workflows on this site but all are overcomplicated. Here's the essential simplest workflow presenting how powerful it is. I'm using this stuff at end of all image generation because its simply free 2-5sec, insane, not only upscale, but also subtle fix for every little thing.

DMD2 lora - https://huggingface.co/tianweiy/DMD2/blob/main/dmd2_sdxl_4step_lora_fp16.safetensors

At KSampler test with "start_at_step" in 60-68 range. Closer you get to 70 the less impact will it have on photo (less steps). Smaller number - more fixes (more steps) but also more intervention from the checkpoint that you potentially would not like. Input image should be in 768-1280 size. Bigger images can produce mutations like double knees.

To avoid that replace "upscale image by" node with forced resize where longest side is not bigger than 2048.

If you want be more aggressive with sharpness - resize image, pass it trough 1xReFocus upscaler and then do ksampler. https://openmodeldb.info/models/1x-ReFocus-V3

What you need

Model files
0.1 GiB known subtotal

Remaining hardware or download-size estimates are not available yet.

Requirements

Models

7

  • 4x_foolhardy_Remacri.pth

    Found

    Upscaler · 63.9 MB · PTH · Unknown

  • bigLove_xl1.safetensors

    Not resolved

    Checkpoint · Unknown

  • dmd2_sdxl_4step_lora_fp16.safetensors

    Not resolved

    LoRA · Unknown

  • ImageUpscaleWithModel

    Possible

    Upscaler · Unknown

  • UpscaleModelLoader

    Possible

    Upscaler · Unknown

  • VAEDecode

    Possible

    VAE · Unknown

  • VAEEncode

    Possible

    VAE · Unknown

Node packs

8

  • CheckpointLoaderSimple

    Possible

    Node pack · Unknown

  • CLIPTextEncode

    Possible

    Node pack · Unknown

  • ImageCASharpening+

    Found

    Node pack · Unknown

  • ImageScaleBy

    Possible

    Node pack · Unknown

  • KSamplerAdvanced

    Possible

    Node pack · Unknown

  • LoadImage

    Possible

    Node pack · Unknown

  • LoraLoader

    Possible

    Node pack · Unknown

  • PreviewImage

    Possible

    Node pack · Unknown

AI-generated commentary

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

Purpose

  • This DMD2 LoRA workflow takes an image through an upscale and refine pass, aiming to produce an enlarged image with subtle fixes.

Flow

  • ImageCASharpening+ is included from the ComfyUI_essentials node pack.

Inputs and outputs

  • The creator describes using it at the end of image generation; LoadImage and PreviewImage are listed among its nodes.

Dependencies

  • Named model files: 4x_foolhardy_Remacri.pth, bigLove_xl1.safetensors, and dmd2_sdxl_4step_lora_fp16.safetensors.

  • Listed node names: CheckpointLoaderSimple, CLIPTextEncode, ImageCASharpening+, ImageScaleBy, ImageUpscaleWithModel, and KSamplerAdvanced.

  • Also listed: LoadImage, LoraLoader, PreviewImage, UpscaleModelLoader, VAEDecode, and VAEEncode.

Configuration

  • Try KSampler start_at_step values from 60 to 68; the creator says values nearer 70 have less effect, while lower values bring more fixes and more checkpoint influence.

    Suggestion · not verified

  • To reduce the described risk of mutations, the creator suggests replacing “upscale image by” with a forced resize capped at 2048 on the longest side.

    Suggestion · not verified

  • For stronger sharpness, resize the image, pass it through the 1xReFocus upscaler, then run KSampler.

    Suggestion · not verified

Cautions

  • The creator recommends 768–1280 input images and warns that larger images can produce mutations such as double knees.

  • No license information is listed.

Does this need to be edited?

Sign in to send an edit request.

Sources

1 source
CivitaiPrimary
civitai.com/models/1278122

Verified Jul 20

Source excerpts

1 excerpt
NoteJul 20civitai.com/models/1278122

Source context: 1101 downloads · Type Workflows · Base model Other