Warning⚠️: Prompting on this lora might be extremely hard or getting bad result easily

New Sampling Method: DPM++ 3M SDE with 40 steps

U can use default sampling if u want: Euler a with 28 steps

HOW TO USE:

Method 1:

1. Use AIDXL Model

2. Please add (aesthetic:1.5) into the prompt

3. Recommend negative: unfinised, sketch, lineart, aidxlv05_neg, (worst quality, low quality:1.2), lowres, messy, (jpeg artifacts:1.2)

4. Recommend weights is around 0.6-1

5. Try put "hatsune miku" or "twintails" on negative if it necessary

Output:

Method 2:

1. Load this VAE

2. Use quality tags

3. No negative needed

4. Recommend weights is around 0.6-1

5. Try put "hatsune miku" or "twintails" on negative if it necessary

Output:

Lora trained on Animagine XL V3.1

描述:

训练词语: