My default workflow for SDXL Lustify Endgame DMD2 checkpoint.
Controlnet: xinsir controlnet-union-sdxl-1.0 or replace with Xinsir Сontrolnet Tile SDXL 1.0
SDUpscale: 4x NMKD Superscale
Description
FAQ
Comments (19)
High quality output, definately one of my fav SDXL worklow around, well done, IT of course also work well with other great models such as BigLove XL1, BigASP, just add the DMD2 something lora to achieve same speed
First generation is slow as f* after that same speed as a random simple upscaler but with great results, now it rly looks like a 4K +++++ image
hey thanks for this i am testing it to see what happens, nice workflow btw
Having a hard time getting a LoRA to work with this workflow. I just run all the connections that would go to the checkpoint to the lora and then connect the lora to the checkpoint?
Add LoRA node after loading checkpoint, also check the Checkpoint page, there's workflow in recommended resources with some useful comments from the community.
I can't find some of the nodes anywhere. Looked on the manager, civitai, reddit, google, etc. .
when you zoom in on the screenshot of the workflow you can see the names of the nodes, maybe this will help. Also update your comfyui.
Is there any way to add image to image on this?
i can this
Im a little confused, 4 steps, LCM?
You need to use the dmd2 lightning lora. This specific one uses lcm + any scheduler to generate an image with just 8 steps. It basically makes a lightning checkpoint out of any SDXL checkpoint.
For the checkpoint he linked it apparently is baked in-- but I still dont see the 8 step https://civitai.com/models/1608870/dmd2-speed-lora-sdxl-pony-illustrious or https://civitai.com/models/1786270/dmd2-lighting8stepcfg15? even https://huggingface.co/tianweiy/DMD2/tree/main only has 4 step i can see.. which do you recommend
LoRA is baked in this model. This workflow is based on the example image (with embedded workflow) shared by the author. 4 step LCM gives good results.
@Adahm I have the same settings as your picture. I attached my LoRA after the checkpoint, but the workflow doesn't work for me. For me its literally backwards. The Initial Gen is super realistic, then the next series of images (Highres, Tile, Fix Face) get less and less detailed with each pass. What am I missing?
how to add lora stacker in this workflow?
was wondering this as well theres so many wires lol. i might mess around and see if I can make it work
Yea u can add a load lora node or similar here. Got it working. just plop it right after the checkpoint node. Grab the model and clip links and route them through it, there are many (hold shift and left click to grab them off of nodes and drag them over to the lora node), and route to left side of the node, then reconnect from the outgoing(right side of lora node) to the left side of all the nodes it originally linked too. This is an amazing workflow!
I don't know why my images are blurry. I'm using Listify Endgame as the image shows, but I don't know what I'm doing wrong. It's like DMD2 isn't showing up and there are no images, just a white blur. I need help
Figured out my other issues, did get this working. I had the same blurry image thing too.. i already had a version of this checkpoint and was selecting it but was getting blurry outputs. Go to the link exactly above in the description and use that one https://civitai.red/models/573152/lustify-sdxl-nsfw-checkpoint?modelVersionId=1099200 its like a few characters different than my other one, must be a different version. Once I did that the blurriness went away.