About DigitalAFlux v1
DigitalAFlux is my first-ever finetuning project, developed using Dreambooth within the Kohya_ss framework. It represents over 120 hours of training on the Flux.1 Dev model so far, with plenty more to come. As a solo hobbyist, I’m well aware the AI landscape evolves faster than I can follow, but this model is a step toward keeping pace.
While not explicitly NSFW, DigitalAFlux is capable of generating convincing anatomy, particularly in the "bobs and vegana" department. "Franks and beans," however, remain a bit more elusive. That said, it integrates surprisingly well with most of the recent LoRAs I’ve trained, and you should be able to coax high-quality results with the right prompting.
I mainly use ComfyUI for generating, so you will have to do your own experimenting if you use something different.
Recommended Settings:
• Sampler: Euler
• Scheduler: Beta
• Steps: 20–30
• Flux Guidance: 3.5–4 (set your ksampler CFG to 1)
Enjoy experimenting—and thanks for checking it out!
Description
FAQ
Comments (8)
A question: there are two parameters: the CFG scale (which usually should be set to 1) and the Flux Guidance Scale (which is recommended at 3.5). Which one do you mean in Recommended Settings?
Great question, I should revise my description to be correct. One of the samplers I use by comfyui_essentials has the flux guidance built in, for some reason I wasn't thinking about the difference, even though I use normal ksamplers with the flux guidance also. In the case you're using a normal Ksampler, you are correct, the Flux Guidance is set to 3.5 and the CFG on the ksampler is set to 1.
@DigitalAF Thank you!
Would be curious to see an FP8 version :)
I plan on it with the next version
I absolutely love how precise this Model reacts when it comes to creating light, shadows and especially LED and neon Lights. Love it so far. Thank you for your work.
Thank you so much for that!!! Once life slows down some, I plan on uploading the next version :)
@DigitalAF Take your time. ;)
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.