【Exclusive on the whole network】Ishikawa Mio

Join the QQ group to be the first to test the model

You can also exchange views together. This is a channel for both beginners and experts to communicate freely. Your participation is welcome, hoping to help more friends who are just starting with SD

Do not lose confidence because of some indifference. If you need it     Group ID:226877470

This version is for entertainment purposes, due to quality issues with the training set

It causes the mouth to look a bit strange when running the final convergence on the map

Everyone can try changing the sampling method

Follow this setup for the best output

Large Model: Maiju v2 Enhanced Edition

Trigger word: scl

Lora weight: around 0.7

Initial resolution: 512×768

High-resolution repair must be enabled

Don't set the redraw range too large

Recommended sample: DPM++ 2M Karras

Number of iterations: 50 is more stable

Tested: Restart 30 steps can produce a good image

Remember to turn on AD for full-body images

For large models using the Maiju, using negative embedding for image output

will yield much better results. The author's site link is provided below for your reference

https://www.liblibai.com/modelinfo/11cb04cd660d48adbde58ee617971e65

Responding to the request of fan ゞ 水﹏﹖

A model of Riko Ishikawa is hereby released

The training set was not good enough and did not meet expectations

However, improvements will continue in the future, so stay tuned

Welcome Riko Ishikawa to join the Cyber Eternal Family! winking!!

(Note: Using the model for image output is purely personal. The purpose of image output should be personal appreciation, and the model should not be used for any form of commercial purposes or to spread or publish false information using images generated by the model)