DatAss Rev FINAL - Pony Realism版本Rev 3 (ID: 518470)

Information:

This model was merged and trained on women only. It can depict males but is not very flexible.

---

LoRA Training:

• Use either wdv3 captions or real captions, i personally use Florence-2 Large in TagGUI

• Train on base pony, results are close enough imho

• Use Prodigy Optimizer with a d-coeff of 0.5

• Use masked training with an 1.0 alpha mask for the face and 0.3 for the body, if possible remove hands and feet from the mask

• Do not train the text-encoders, only unet

• 600 - 1000 steps is suggested

---

Suggested Settings:

• Sampler: Euler a + AYS (Align your steps)

• Resolution: 832x1216

• Steps: 14

• CFG: 4 - 7

• Perturbed Attention Guidance 2.7

• HiRes Upscaler: 4x_NMKD-Siax_200k

• Sampler: DPM++ 2M SDE + SGM Uniform

• Steps: 30 - 50 (Or 12 with SDXL Hyper LoRA at 0.5 weight)

• Denoising Strength: 0.25 - 0.35

Prompt:

As usual start with the score tags. I personally start every prompt with this huge list of tags:

score_9, score_8_up, score_7_up, photo, source_photo, source_real, RAW, 8k, low quality, traditional media, amateur, real life photo, film grain, dark, darkness, dim night light, light flash, very dark focused flash photo, (dark background:0.75), low light, detailed face

---

DatAss Rev 2 Recipe:

Step1: GameArt Pony + CyberRealistic Pony (0,0.1,0.2,0.3,0.3,0.3,0.3,0.3,0.3,0.2,0.2,0.2,0.3,0.3,0.3,0.3,0.3,0.3,0.2,0.1)

Step2: Step1 + Custom Trained Pony Model (0,0.15,0.25,0.3,0.25,0.2,0.15,0.15,0.1,0.1,0,0.1,0.1,0.15,0.15,0.2,0.25,0.3,0.25,0.15)

Step3: Step2 + (Damn Pony Realism - CyberRealistic) (0,0.35,0.6,0.65,0.45,0.25,0.1,0.1,0.05,0.05,0,0.05,0.05,0.1,0.1,0.25,0.45,0.65,0.6,0.35)

Step4: Step3 + (Personal Anteros Mix - SDXL) (0,0.25,0.4,0.45,0.25,0.15,0.1,0,0,0,0,0,0.05,0,0.1,0.15,0.25,0.45,0.4,0.25)

Step5: Step4 + (Step4 (+SG Finetune) - Step4) (0,0.2,0.55,0.43,0.35,0.1,0.05,0.05,0.05,0.05,0.05,0.05,0.05,0,0.05,0.05,0.05,0.05,0.05,0.05,0.05,0.1,0.35,0.43,0.55,0.2)

Step6: Step5 + Personal Anteros Mix (0,0.1,0.28,0.28,0.2,0.1,0,0,0,0,0,0,0,0,0,0.1,0.2,0.28,0.28,0.1)

Step 7: Step6 + ICKPOTLycoXL 0.25 + photo_v2 0.2

描述:

Trained a bit more, then added a lot more checkpoints and LoRAs to the mix, improved contrast and realism.

Not much room for improvement here, it might not be better than Rev 2 in all aspects.

训练词语: realistic,cinematic film still,photo