ILLUMINATE AI — DMD2 CREATOR WORKFLOW
A clean and modular ComfyUI workflow built around fast SDXL DMD2 generation, controlled detail enhancement and a high-quality final upscale.
The workflow follows a clear processing order:
Base Generation → Base Preview → Face Detailer → Eyes Detailer → DMD Refinement → SeedVR2 Final Upscale
The main goal is to keep the workflow fast and easy to use while still providing enough control for higher-quality final images.
Main Features
Fast SDXL DMD2-based image generation
Dedicated positive and negative prompt sections
Free LoRA Loader for styles, characters and concepts
Base Preview before running the full processing chain
Optional Face Detailer
Optional Eyes Detailer
Mild full-image DMD Refinement Pass
SeedVR2 final upscale
Fast Group Bypasser for optional modules
Only the final processed image is saved
Built-in setup, download and usage guide
Base Generation
The base image is generated with a fast DMD-oriented setup.
Recommended starting values:
Steps: 10–15
CFG: 1.5
Sampler: LCM
Scheduler: Karras or Beta
Denoise: 1.0
The workflow includes a dedicated Base Preview, allowing you to check the composition before running the Detailers and final upscale.
If the base image is fundamentally wrong, changing the prompt or seed is usually better than expecting the Detailers to rebuild the entire image.
Face Detailer
The optional Face Detailer is designed to improve facial structure, skin detail and overall clarity without heavily changing the original expression or identity.
Recommended denoise range:
0.24–0.28
Higher denoise values can alter facial proportions or create an overly processed appearance.
Eyes Detailer
The optional Eyes Detailer focuses specifically on eye clarity and structure.
Recommended denoise range:
0.16–0.20
Lower values usually produce more natural results, while stronger values can make the eyes overly sharp or artificial.
DMD Refinement Pass
The second sampling stage is intended as a mild full-image refinement rather than a complete regeneration.
Recommended values:
Steps: 4–5
CFG: 1.0
Denoise: 0.10–0.15
A value around 0.12 is recommended as a balanced starting point.
Higher values can exaggerate skin texture, pores, muscles, veins or other small details.
SeedVR2 Final Upscale
SeedVR2 is used as the final processing stage.
Recommended output size:
2800–3200 px for general use
4000 px for an optional Ultra output
Very high resolutions can increase render time, VRAM usage and invented micro-details.
Recommended Quality Preset
Base Generation
10–12 steps
CFG 1.5
Face Detailer
Denoise 0.25–0.28
Eyes Detailer
Denoise 0.16–0.20
DMD Refinement
Denoise 0.12–0.14
SeedVR2
2800–3200 px
This preset prioritizes natural detail, stable skin texture and expression preservation.
Required Custom Nodes
Install through ComfyUI Manager:
ComfyUI Impact Pack
ComfyUI Impact Subpack
rgthree-comfy
ComfyUI Essentials
ComfyUI SeedVR2 Video Upscaler
The workflow structure and processing order match the original pre-OpenPose version.
Description
Version Changes / Notes
Initial Release
Added fast SDXL DMD2-based image generation
Added dedicated positive and negative prompt sections
Added a free LoRA Loader for styles, characters and concepts
Added a Base Preview before running the full processing chain
Added optional Face Detailer and Eyes Detailer modules
Added a mild full-image DMD Refinement Pass
Added SeedVR2 as the final upscale stage
Added Fast Group Bypasser controls for optional modules
Configured the workflow to save only the final processed image
Added built-in setup, model download and usage notes
Optimized the default values for speed, natural detail and stable skin texture
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.





