3xperiment0
Chaos, Art, Light Experiments
3xperiment0 is an experimental and evolving LoRA focused on artistic chaos and expressive lighting behavior. It is designed for moments where an image already feels good, but something subtle or unexpected is still missing. This LoRA introduces small shifts in lighting, contrast and visual balance that can push an image into a more artistic and emotionally charged direction. You can try (nearly) any style you want.
It works especially well for art driven styles, but also blends nicely into anime, semi realistic and even photorealistic compositions. Rather than enforcing a strict look, 3xperiment0 acts as a creative catalyst, nudging the image toward more interesting light interactions and visual tension.
This LoRA is currently in absolute beta and still under active development. Results can vary depending on prompt structure and model combination. Some effects remain subtle, others may feel more experimental by design.
Known limitations (for flux) at the moment are hands and feet, especially fingers and toes, which may require additional guidance or cleanup.
Recommended starting strength is 0.3 to 0.5
Maximum recommended strength is 1.2
From around 0.8 and above, lighting behavior may noticeably shift or drift in unexpected ways
Anything you find of mine outside of Civitai is shared without my permission.
Please do not reupload my files. Link people to my Civitai profile instead. Thank you 🙂
Description
Go around 0.3-0.5 for the start. 0.4 seems to be fine in most cases.
Colors and sharpness tend to get stronger the higher you go.
Euler_a is nice for most artistic.
dpmpp_2m_sde works well for a bit sharper and realistic stuff
Please do not reupload my files. Just link people to my Civitai profile instead. Thank you.
https://civitai.com/user/tehalex86
FAQ
Comments (20)
very cool
Huge thanks. Still way to go. Z-Image feels kinda complicated sometimes :D
Amazing lora =)
Huge thanks. Can be handy on some themes :)
Nice Lora, thx! I used it for a YT clip about a Hamster warrior https://youtu.be/bUNj9NvoOAg (WAN Vace FLF, the snail riding on the dragonfly was QWEN, couldn't get this concept working with Z-Image)
Huge thanks for the feedback. Wow awesome. Well done. Now I'm even more afraid of the "The Snail"
What a great clip. Starting to train some Qwen LoRAs soon. Right now I'm missing the time to finish my training setup :/
@tehalex86 thank you for your generous tips as always, Love your models, I am bit adicted to this Lora models, thank you for share it with the community, I am yet to try out all your latest models, the delay is because of skill issue... lol still learning...
You're welcome. Thank you for playing around with my stuff and posting so much cool artwork here. Really enjoy it. Trying too let that tips roll ;)
No pressure with testing anything. Always happy, when someone tries my stuff. It is and will stay a pure hobby for me ;)
great lora, it manages to hit its target exactly. as much as any model has yet managed to bottle the magic of artistic composition and lighting, this one has it
Can't thank you enough for those kind (magical) words. Means a lot to me <3
I have doubted your LoRA at first sight BUT downloaded it anyway and its truly a great work, thank you. Very impressive!
A version trained with z image base would be awesome! Thanks
As soon as I find time <3
Gave it a first shot. Training is a bit tricky at the moment. Guess I need to learn to use ZIB in a good way. Photoreal seems to be working (sometimes :D), but with artistic images....ZIB kinda freaks me out at the moment.
Thinking about uploading my first alpha here, but I'm not sure yet tbh ^^
I'm gonna build you a wildcard randomizer dataset creator that plugs into Ostris 🤔
Only tried wildcards for 1-2 times, but maybe should give it another shot :) So why not