Everflux x1: The Photographer - NF4
Made in Forge UI. Use the unet nf4 loader for this to work in comfyUI.
This model merge emphasizes:
Hyper-Realistic Photography Elements: Perfectly suited for generating outputs with lifelike textures, natural lighting, and striking visual depth.
Exceptional Detail and Textures: Ensures outputs with intricate, true-to-life details, elevating creative projects.
Improved Human Anatomy: Delivers anatomically accurate representations, with particular attention to historically challenging areas such as hands and body proportions.
Enhanced Text Transcription: Offers reliable and precise text rendering, making it a versatile tool for projects requiring clear typographic elements.
Reduced Content Restrictions: Provides a more flexible creative experience, allowing the generation of artistic nudes and other uncensored visual content.
This model is an ideal choice for photographers, digital artists, and creators who prioritize realism and creative freedom in their work. Whether you're creating hyper-realistic character art, generating high-fidelity textures, or exploring conceptual photography.
Description
Original Version
FAQ
Comments (13)
How to load/where to store? Tried as unet and standard checkpoint, both threw errors in comfy...
I use it in Forge
@Humanoid23 OK, but does that have different loading options? Is it a diffusion checkpoint model, unet, etc?
@EnragedAntelope try to use the unet nf4 loader in comfyui
@jice Thank you, that was the trick! Never used that node before.
@jice Should probably mention that its a NF4 version. will be helpful for users
@itsnotroobs Yes, unet (without vae and clip) nf4, because nf4 All In One (with vae and clip) it's not the same loader, but I am not the author of this model
@itsnotroobs Good call, thanks! I'm new to the Flux landscape.
@EnragedAntelope i never once got any nf4 checkpoint to work in comfy. not even with the nf4 loaders, updating etc, consulting with A.I. doesn't work. and I don't understand why people even use Forge because it's not maintained. do you mind sharing a workflow?
I don't think this model is FP8 as shown in the description, is it? At just 6.24 GB in size, this appear to be a different, smaller quant? FP8 Flux models are usually ~11 GB in size (U-NET only).
My mistake, it's NF4
does this need unet, clip, and vae to use in Forge?
Tried it out, It seems to provide some better quality with photos but its prompt adherance for "unexpected" imagery is a bit hit and miss, it seems to fail more often the longer the prompt is.
It's relatively quick which is nice.