Huge natural breasts, areola slip for FLUX

Huge natural breasts, areola slip for FLUX - V1

NSFW

Shoutout to yoadster693; he asked me to do an areola slip version of my huge natural breasts lora, so here it is!

The dataset for this lora has been prepared like my typical lora. It is captioned in natural language without trigger words. All faces in dataset have been censored in various ways to prevent the result from looking like the persons in the dataset. If you want to see a face, put details of it in your prompt. Random edits were used to prevent the introduction of bias caused by these edits.

Description

{
  "engine": "kohya",
  "unetLR": 0.0005,
  "clipSkip": 1,
  "loraType": "lora",
  "keepTokens": 0,
  "networkDim": 4,
  "numRepeats": 6,
  "resolution": 1024,
  "lrScheduler": "cosine_with_restarts",
  "minSnrGamma": 5,
  "noiseOffset": 0.1,
  "targetSteps": 2640,
  "enableBucket": true,
  "networkAlpha": 8,
  "optimizerType": "AdamW8Bit",
  "textEncoderLR": 0,
  "maxTrainEpochs": 40,
  "shuffleCaption": false,
  "trainBatchSize": 3,
  "flipAugmentation": false,
  "lrSchedulerNumCycles": 3
}

FAQ

Comments (5)

yoadster693Dec 11, 2024· 1 reaction

CivitAI

Once again mate, stelar work!!!!

KiefstormDec 11, 2024

CivitAI

Excellent! Quick question regarding your statement- "Random edits were used to prevent the introduction of bias caused by these edits."

Does this just mean you used different shapes when censoring the faces?

Thanks

ElectronLibre

Author

Dec 11, 2024

I use random shapes with random colours, I even tried random emojis with success. I do this on any part of the images I don’t want in the lora, usually it works great. I censor faces, but also things like tattoos or unclear features. What’s important is that I also use close-ups without other edits, this help with not learning the shapes. The combination of overview with censoring and close-ups on the features I want in the lora help to prevent other bias. Overall, I try to randomize what I don’t want in the lora; style, photo type, clothes, etc.

KiefstormDec 12, 2024

@ElectronLibre nice, I've experimented with this as well. I'm in the middle of testing this process with and without captions. I'm testing joy captions, which say stuff like "a photo of ... with their face censored by a red circle". Did you use captions this way?

ElectronLibre

Author

Dec 12, 2024

@Kiefstorm Yes, I use joy captions and I train on civitai directly (their auto-label captions feature use joy). Joy does put a caption also for the shape used for censoring.

LORA

Flux.1 D

by ElectronLibre

Download (Beta) View on CivitAI