*Perhaps because these sliders do not rely on specific images, they function almost exactly the same way with Z-Image Turbo.
These are a collection of experimental "concept sliders." The cover image is unrelated to the actual content.
Theoretically, anything achievable with these concept sliders can also be done via prompts, though these may offer slightly easier handling. Since they do not utilize specific image data, the model may not always behave exactly as intended. As a general rule, aim for a setting range of +/- 5, though values outside this range may sometimes be appropriate.
QualitySliderZ
While technically a slider, this functions similarly to a TextEmbedding. It consolidates elements that would typically be handled by quality tags or negative prompts—think of it as the equivalent of "EasyNegative" or "EasyPos" for SD1.5 or SDXL.
With the advent of Distill LoRA, there is an increasing trend of using Z-Image Base at a CFG scale of 1.0. While this setup normally precludes the use of negative prompts, this type of LoRA allows the functionality of negative prompts to be handled within the positive prompt itself. However, please note that the intensity differs between concepts manipulated via LoRA and those described in the prompt text, so you should not expect identical results.
This version does not include tags like "masterpiece" or "best quality," as they are ineffective in this context. Instead, the primary focus is on ensuring anatomical stability, improving composition, and suppressing unwanted elements like text.
AnimalSliderZ
This slider is used in cases where a prompt like "a cat" results in a cat-eared girl, or where a standard "1girl" prompt unexpectedly produces cat ears. The positive and negative settings suppress human ears and animal ears, respectively. While there seems to be a trend where having four ears isn't considered an issue, I implemented suppression for them anyway—though whether it actually works is another matter. When used on a human character, it can cause horns or wings to sprout in addition to the ears, even though I only specified ears and a tail.
AgeSliderZ
A slider for adjusting age.
ThiccSliderZ
A concept slider for manipulating body shape.
ExposureSliderZ
A slider for adjusting photographic exposure (overexposure, underexposure, etc.). The concept alone may not be sufficient for the desired effect.
GuofengSliderZ
A slider for adjusting the "traditional Chinese" aesthetic. The negative setting shifts the style toward a Western look, though the resulting behavior was somewhat unexpected.
ClothingSliderZ
A slider for adjusting the amount of clothing. This is highly experimental.
SlantEyesZ
A slider for controlling eye slant (droopy vs. upturned eyes). The character's appearance may change—possibly due to insufficient anchoring—suggesting that the concept alone might not be enough to achieve the desired result.
Description
FAQ
Comments (2)
I tried using the output from the FP8 model with QualitySlider, but there is indeed some degradation...
The purpose of QualitySlider is to specify negative elements within the positive prompt itself. While there are limitations due to the nature of this conceptual processing, it is effective. For instance, writing "blonde hair:4.0" in a negative prompt prevents blonde hair from appearing in the image. However, writing "blonde hair:-4.0" in a positive prompt has no such effect; in fact, it makes blonde hair more likely to appear. Instructions like "do not show text" are also futile. Positive prompts only respond to affirmative expressions. By manipulating concepts, it is possible to achieve negative processing without using actual negative prompts.
Since these are packaged as LoRAs, they are limited to handling standardized elements, but I believe we should expand the variety of these LoRAs. Please let me know if you have examples of negative prompts that work effectively.
Note: Many attempts will not work effectively. Concepts that do not produce the expected change even when written as "masterpiece:10.0" in the positive prompt are a waste of time to include.


