In Norse mythology, Mjölnir (Old Norse: Mjǫllnir - “Lightning” or “That which smashes”) is the hammer of Thor, the god of thunder and lightning...........
This model can do:
Photorealistic
Digital Art
Anime
Comic
Vector Graphics
Line art
Stickers
can write short text
etc
This is my first Lighting SDXL.
its an merge model based on Yggdrasil , Ratatoskr and elements of the original Lightning model. Please check also out there epic work ;-)
Changelog:
V1: First Version
Recommended settings: for the model
The model don't need the lightning Lora
dpmpp_sde or Euler a / Euler at least 3-4 steps / recommended 8-9 steps cfg 2-2.5 / cfg 2-2.5 / sgm_ uniform scheduler, also dpmpp_3m_sde_gpu runs fine on at least 10 steps will provide an xyz grid ASAP . i will place the pictures in the workflow file
btw the model works also with other samplers and higher step counts up to 50 steps (dpmpp_sde ) and tested till 7 cfg but this is experimental and up to you ;-)
V1.01
VAE Fixed
V2.2
more detailed
more creative
better with art
Note: In exchange of quality this model is a bit slower than V1.01 you need at least 8 Steps and 12-16 for best quality with for example dpm++ sde at cfg 3.
in this model i did merged in following models
Midgard , Midgard Pony , Valhalla, and Demoncore " Midgard Beast"
have made an comfy ui workflow: Mjölnir workflow
Made by Freek22
Please check out my other models too and give like ;-)
Description
V1.01 VAE Fixed : for Settings read instructions ;-)
made an comfy workflow for the model can found here ; Mjölnir Workflow
FAQ
Comments (12)
This model is making good images of people, except for always having white "artifacts" in every image, no matter what negative prompts I use. I was wondering what you do to get rid of those.
Those are usually a VAE issue
yes its the vae you need the standard sdxl vae i have made a workflow for this model for comfy ui it also contains a list of 4 xyz grid of 4 samplers diffrent settings https://civitai.com/models/321437/mjolnir-workflow?modelVersionId=360428
@Zaranna another thing is this model needs a higher cfg than any other lighting model between 2-3 cfg i have mead my test in the meantime the internal vae is a bit broken will fix that on v1.1
for anyone wondering, comfyui makes it really easy to swap out the VAE by using "LoadCheckpoint+LoadVAE->CheckpointSave" nodes. "model" and "clip" both come from loadcheckpoint, "vae" comes from your desired vae.
i hate having to fuss around with VAEs, so i swap 'the right ones' permanently into the checkpoints they work best with, as soon as i realize i like them enough to keep them around. i can always override it later
great model, btw! cheers!
@blncd thanks and I will fix this issue with the next update this is just the first version I plan to improve it even more 👍😉 maybe I do a HotFix even before the actual update
@Zaranna @blncd VAE Fixed ;-)
The Model is now also available on Tensor for online generation 😉👍
why no sharing the example runs prompts? T^T
I have that but Civit can't recognise my workflow at all it's the same workflow I published with the model and a lot of pictures below the main gallery has prompts you can also download the pictures and put them into comfy ui the metadata are inside the pictures'
@freek22 There is a Civitai metadata node "Save Image w/metadata" used in "Jadewinds Magic Moments" Workflow. I do not know from wich package... Would be a huge benefit if you use it!
Frankly speaking, a "hit" in SD XL.
Fantastic job.