Very happy to share my new model! this model is all you need when comes to generate NSFW images using Flux!!! Highly trained.
As of now try it out and let me know your thoughts about it.
(NSFW) Completely uncensored, allowing generation of erotic and very explicit images.
NSFW FLUX - Checkpoint
Training Configuration:
GPU: RTX 4090
Dataset Size: 200 images
Batch Size: 1
Optimizer: AdamW
Learning Rate: 1e-04
Total training steps: 4.5k
Training Duration: 3h+
Recommended Parameter :
Sampler : euler
Steps : 20+
CFG : 1
Recomanded weight is 0.5 - 0.9
trigger word :- nude, big round boobs, topless
If you like my model and want to support my work, send me some buzz.
Please post your pictures directly on the model card...that also helps me to continue my work.
Access The Full Professional Toolkit
The models I release publicly are foundational. For artists and creators who demand the absolute best, my most advanced, highly trained, and exclusive character models are available only at The Midnight Lab on Gumroad.
A subscription is the best way to get access to these premium models, my entire library of professional workflows (like The Muse Engine and 4K Influencer Studio), and in-depth guides.
>> Explore The Midnight Lab on Gumroad <<
Description
FAQ
Comments (14)
15gb lora? wtf?
description says is a lora.
My bad. Typo error
Great work! IMHO version 1 of this model (i.e. the new v1) is the best Flux dev model on Civitai. And I pretty much try them all - I run every model with a standard test suite of 10 prompts, batch of 4 images per prompt.
But I'm not convinced v2 is an improvement over v1. Yes it's more realistic, but it also significantly increases bad hands, extra limbs and body horror in general. I get the feeling you may have just baked in a realism LoRA made by someone else. Just wild speculation of course, probably triggered by its quick release (just a few days after v1) and the uneven output.
Again, you're definitely on to something fantastic, but please consider continuing your training on v1 and not v2.
thanks, even i felt so... i'll rework on v1's improvement. i think implementing realistic texture is spoiling checkpoint.
I was halfway through download when I read this comment so I had to stop it and do the other one. Thnx for helping me not waste my time.
CLIPTextEncode ERROR: clip input is invalid: None If the clip is from a checkpoint loader node your checkpoint does not contain a valid clip or text encoder model. SOO. Which Clip Text Encoder Model does it require ? A special nsfw unlocked one ? edit: forget it. I thought the checkpoint would carry its own clip and vae. not downloading more stuff. recycle binned.
any flux clip and vae works
Whatever happened to this version:
flux1DevNSFWUNLOCKEDFp16_v30FP16.safetensors
It was form old account, which was taken down by civit ai. Will upload this week...
@The_Midnight_Lab Thanks, this model is probably the best flux based model.
@The_Midnight_Lab It is a very good model, Great job!
@The_Midnight_Lab reminder to upload v3 :)
@VonLemberg i need to redo the whole model as civit is not allowing me. will do soon once i get time.
Why not google it and download it. First result at huggingface.
Details
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.