Type | |
Stats | 502 3,260 789 |
Reviews | (63) |
Published | Mar 26, 2024 |
Base Model | |
Training | Steps: 10,000 Epochs: 25 |
Usage Tips | Clip Skip: 1 Strength: 0.8 |
Trigger Words | de_portrait de_skill de_scenery de_thought |
Hash | AutoV2 7DB652A568 |
This LoRA tries to replicate the art styles present in Disco Elysium. See the "Usage" section and preview images for available styles.
This LoRA is quite experimental and focuses on expressionism and abstract art. If that's not your thing, you will probably not like this LoRA.
If you like the LoRA, please leave a reaction, review and/or tip. Enjoy!
Usage
Use activation trigger, according to the desired style:
de_portrait
: For character portraits
Support tags (optional):oil painting
de_skill
: For "skill" portraits
Support tags (optional):black background
, one ofred theme
orblue theme
oryellow theme
orpurple theme
de_scenery
: For wide shots of landscapes or cityscapesde_thought
: Replicates the "thought" images from Disco Elysium. Hard to describe. Very experimental and inconsistent results. Does well with low-strength ControlNet guidance.
Recommended to use at strength 0.6 - 1.0 (start at 0.8)
Use Clip skip=1 or Clip skip=2 (other values may also work, use at own risk).
Works best at 1024x1024.
No VAE needed. Use one according to your preferences and image contents.
If you are using an AutismMix based model, use the autism version for best results.
Example
See preview images (generated with pruned version).
Unpruned vs. Pruned vs. Lite
I offer three variations of the model, which only differ in the file size and quality. Pick the one that best suits your preferences. If you cannot decide, simply pick the "pruned" version. Note that the quality is relative, even the lite version usually produces images which closely resemble the intended art style.
unpruned: The originally trained model. Largest file size, best quality. Pick if you want the best possible quality, at the cost of disk space.
pruned: Downsized model. Almost same quality as the unpruned version, with noticably smaller file size.
lite: Drastically downsized model. Smallest file size. Still has good quality, but noticeably worse than unpruned and pruned.
Notes
This was mostly an experiment with different training settings and combining multiple art styles in a single LoRA. I think it turned out well, though the art styles do clash a little. But it's still more efficient than training 4 separate LoRAs.
I'm especially happy about the "skill" style. It's able to create some really nice images.
Disclaimer
This LoRA is not affiliated with the artist/game it is based on. If you like the art style, please support the studio behind the game by buying a copy of Disco Elysium (and optionally also playing it. It's a good game).