Model Usage:

Mainly used to train the MM_LORA model, and then layer the weights to separate and color, so that you can adjust eyes, texture, etc....

V1: The primary purpose of this model is to achieve subtle, high-fidelity synthesis between photos and anime. When you train a LORA with this model, you can produce both high-definition anime and highly realistic images.

V2: Update all the details about layered fusion to make the generation very detailed, sacrificing dynamics and tension in exchange for the accuracy of the generation.....

V3: Violent Style Implantation: Aggressively embeds the target style (A "blueprint" approach).

This

[ V3 model consumed a massive amount of my passion and nearly depleted all my creative drive/desire.

If you choose to use this version, please consider sponsoring contributing towards the electricity costs and effort expended.]

V a 2.0: Forged by expert critique in a private group, this model's weights were refined over 5 iterations per 'a' version.

The result, v a2.0, is a highly versatile model for generation and training. It delivers near-realistic anime, backgrounds, and expressive emotions. It will be released for free.

v0.3: Used for LoRA training, involving the removal of backgrounds from a large quantity of extremely high-quality photos for classification label training.

v1.1: The model reflects aesthetic progress informed by peer feedback. Due to the mounting complexity of continuous iteration, I will pause main training. I need a break to focus on developing more LoRAs to enhance my aesthetic judgment.

v2.2: Slightly improved realism while mitigating graininess/noise caused by missing data

Recommended Resolutions:

(832x1216 - requires upscaling for high-definition),

1024x1536

1280x1960

Photo Anime - v2.2

Photo Anime - v2.2

Sampler

Steps

CFG Scale

Seed

Clip Skip

Size