t3(ver-10 beta)
★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆
Please upload at least one picture you made. We will try to reflect it in the official version by referring to the extent of the picture. Any picture is fine, so please upload it.
★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆
1. This version was designed to produce good pictures even at low sampling steps.
2. Please refer to the reference picture (most of the reference pictures apply low Sampling steps. When using higher Sampling steps You will be able to create better pictures.)
3. Useful pictures are created from sampling steps 5 or higher.
CFG scale is recommended from 6 to 8.2. If the picture is broken above 9
We recommend using Dynamic Thresholding. Set the value approximately “-1” lower than the sampling step you set. (Please make additional selections yourself ^o^)
4. ADetailer produces a good picture with a face or hand. (I don't know because I haven't tested lora in use.)