No Buzz? No VRAM? No problem.
Drawing from a dataset of over 1200 Flux images, I've been aiming to create a set of functional "Flux Style" LoRAs for SDXL (and Pony!). HUGE thanks to @Valstrix for helping me with the initial sort.
As a proof of concept, I started with a dataset of 90 landscapes, ranging from the real to the surreal, from fantasy to sci-fi, and everything in between.
Trigger: flux-scenery
There are a lot of helpful tags to discover; nature, outdoor, scenic, landscape are some of them.
I encourage you to give feedback - this is new territory for me, and I'm hoping to get better with each concept I'll extract from the dataset.
For a generalist model by someone far more talented than I, check out @AIDigitalMediaAgency and his excellent Flux Lora XL!
Description
FAQ
Comments (4)
Don't. Never use Flux images in training your stuff. If you want to know why, read the Flux license.
I did, and read the part about using its outputs to fine tune "products that compete with Flux.1"
I didn't think that applied to (specific) LoRAs, only to full model fine-tunes. They also say that you own the outputs, so that's kind of in opposition.
Unless there's a part that I missed that covers LoRAs beyond a shadow of a doubt?
Did you make them with Flux dev, pro or schnell?
@mnemic Dev
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.