@znort By the way, I have tested the experience for 24 hours and provided you with feedback on the results.

1. When the weight is 0.8 or above, there is a 50% probability that collaborating with the character Lora will cause facial deformation. If the weight is below 0.8, it can effectively maintain and recognize the face of the character Lora.

2. When using the combination of "wide open mouth" and "tongue out" prompts, in some cases, the tongue may not protrude, especially when the weight is less than 0.8.

3. The posture, painting style, and posture are too lively and varied, which means that the obedience and controllability of the prompt words are not as good as the old version. This is a double-edged sword, both advantages and disadvantages. Unfortunately, the diversity of the tongue is bound to the style of character posture painting. It would be great if the diversity of the tongue were separated from the style of painting and character posture.

4. At present, a relatively stable method of use is to use it in conjunction with other tongue models. The disadvantage is that although the posture and painting style are controllable, the appearance of the tongue itself is relatively lack of change and relatively single, because it cannot utilize the rich and varied tongue appearance of this model under high weights: vivid and varied tongue coating, not a single and rigid shape. When used with other tongue models, there is a greater probability of changing the character's face under high weight.

Finally, this is still the best tongue master model currently available.

@znort By the way, for anyone interested, I discovered the secret to this by accident when comparing versions. I mistakenly ran both versions together, with the newer one at a slightly higher strength (.38-.44), and I got perfect amazing results every time!