Goddess Project
--Formerly Uncensored Females--
Standalone Checkpoint - Goddess works in FORGE only
DO NOT LOAD a separate VAE, TE, or CLIP unless using GGUF
This version is a mixed precision Flux Dev model, with limited UNET changes to allow for feminine anatomy.
Run this model in automatic FP16 Lora mode NOT NF4
This model is full precision in BF16 UNET with mixed precision (NF4) on the TE blocks
This model fits in a 24GB card and could be run in GPU only mode as such
High speed BF16 with a slightly lower prompt accuracy compared to the 33GB full model
Links (For GGUF ONLY)
Updated CLIP - Standard CLIP-L - FP8 CLIP-L -- ** Version Comparison **
Per the Apache 2.0 license FLAN is attributed to Google
This model is a training using many individuals with known ages and 2257 forms, it has also been merged to try and ensure that no known individuals can be reproduced. However FLUX seems to like to learn faces even with less then 10% data rather then merge them into a new face.
Description
FAQ
Comments (17)
how to use fp16 lora
Any Lora should work with the workflow provided.
Fluxs 'same face' issue will never be fixed, I'm back to sdxl, what a waste of potantial.
Character lora's work so it is possible, to want extent who knows
@Felldude If someone finds a way to tackle this issue without ruining the core creativity of Flux, that would be awesome.
@Mr_Jinguji learn more about block weights
@Felldude LORA barely works. I tried several different and instead of other Flux FP8 checkpoints, I got worse results.
Is already fixed: https://civitai.com/models/766608/sameface-fix-flux-lora
@Sendrael , I meant characters loras. Yes, this is not about "same face," but it's about Loras working like 10%. The same loras work well with other checkpoints such as original flux.d, agflux, agflux shnell, stoiqon, even with flux1DevNF46StepsNSFW_fluxdevFP86Steps, okay, but worse. Guidance is no matter at all.
@DemonAlone Regarding character LORA's the ones I have listed work well however I am doing several trainings a day to try and balance out the issues with V1 of this checkpoint
@Felldude , I see, thank you for your answer.
100buzz for your great work thx bro
Thank you
Training checkpoints is like burning money
Hey, I'm looking for a high-quality Flux checkpoint. I just tested yours, and I think you need to put more effort into the training set. The facial weights of the characters have deviated from the original checkpoint, and the skin texture is quite poor. However, most people overlook this issue when training checkpoints, so it's not your fault. My suggestion is to use more high-quality, faceless materials.
@samliu I will be working on the fact that it burnt out all but light skin colors first, this is a merge not a finetune, and I am using Civitai trainer so options like oversized training and random cropped bucketing or masking are not available.
Everything about AI is like burning money, so burn it for something you enjoy!