This LoRA is both a style and a concept. It's quite experimental, allowing you to present your characters as holograms or digital versions with creative 'glitch' effects. Plus, it can even help you craft virtual worlds and glitchy environments with vivid colors.
The model can be unpredictable sometimes but generally provides interesting and creative results. Be aware this is not the easiest LoRA to use.
I'm still trying to figure out the recommended parameters lol (are there any?).
Nevertheless, here are a few things I've noticed:
- It works 100% with the base Pony XL v6. You'll probably get most creative results using this checkpoint.
- It's at 70 - 80% of its capabilities with derivatives (AutismMix, js2Prony, etc).
1 - Sampler: Euler a
Steps: >30
CFG Scale: ~6.0 - 7.0.
2 - Sampler: DPM++ 2s a (slow but good results)
Steps: ~25
CFG Scale: ~5.5
LoRA strength: it's up to you. Start with 1.0 and adjust to your liking. You may need to increase the strength a little if you're using a Pony XL derivative.
Trigger keywords: [color] hologram, holographic
Optional keywords: glitch, glowing, dark background
Examples: blue hologram, holographic, glowing
Be aware using colors for hair, eyes, etc may interfere with the LoRA.
Without keywords, it can produce good content too with dark background.
Description
FAQ
Comments (6)
Don't know why, but your model make buildings in glitch style, and it's cool.
Among other things, the model has been trained on images featuring glitchy sceneries and environments, which may explains why :)
I'm wondering whether I should train a more stable v2.0 that will always generate holograms in a particular style, sacrificing the creativity of the results. Although the current model is quite difficult to use (using Block Weight may allow better results), it's still quite flexible as you can see from my examples. Would you prefer a stable, concept-oriented version or a wilder, style-oriented one?
why not both (2 different models)?
@reindeersfallingfall Yeah that's probably the right thing to do
I use this model because it allows me to render mostly arbitrary characters with this effect. While it would be nice if it was easier to keep the concept from messing up the backgrounds, I think the versatility is more important to me.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.