This Demographic Shifting Pack contains Concept-Slider LoRAs designed to seamlessly transition between different ethnicities.

To use these sliders correctly and effectively, you need to understand a few specific quirks of Krea2 and what actually happens when these models are applied.

Morphing from a Russian subject to a Native Indigenous subject, for example, is not controlled by skin tone alone. This slider alters characteristic facial structures, bone formation, facial soft tissues, and, of course, the skin pigmentation itself.

Consequently, the required weight settings may vary depending on the fine-tuned checkpoint you are using—especially if your model has an inherent bias, such as being pre-trained on Asian subjects. Similarly, your choice of lighting can heavily influence the vector and its corresponding weights. This is particularly noticeable in the positive range of the sliders. For instance, in a controlled lighting setup, the Caucasian-African slider might achieve a fully developed African skin tone and complete structural features at a weight of just 3. In a different prompt with very warm, golden light sources (like Golden Hour), that exact same weight of 3 might result in a subject that looks more Latina. If this happens, do not hesitate to push the weights higher. These sliders are incredibly robust, and depending on your prompt, it takes a lot of oversteering before the image quality actually collapses. Since all these models were trained on "Krea2 RAW", they should be fully compatible with all common Krea2 versions.

I am highly interested to see how these sliders perform on Krea2 models that have been specifically fine-tuned for semi-realism or anime style. But that’s just a side note.

Naturally, when using these sliders, it makes little sense to explicitly include ethnicities (e.g., "Chinese woman") or skin descriptions (e.g., "tanned skin") directly in your text prompt. To give the slider full mathematical freedom over the bone structure and facial morphing, you should stick to entirely neutral terms like "woman" or "man".

For each individual slider, I will provide a neutral male and female test grid demonstrating the weights from -3 to 3.

The prompt used for these grids is as follows:

A dramatic and high-end studio portrait of a confident [woman/man], sharp focus, cinematic lighting. The subject is captured from the chest up (medium close-up), looking directly into the camera with an intense, captivating expression. Moody and atmospheric studio setting with a dark, textured charcoal-grey background. The lighting is an exquisite interplay of Rembrandt lighting on one side of the face, casting deep, artistic shadows, paired with a vibrant neon cyan and amber rim light that sharply outlines the silhouette. Subtle volumetric smoke or mist floating in the background, catching the colored light. High-end fashion editorial style, photorealistic, intricate skin texture, pores, and realistic hair detail, cultural accessories, moody color grading, volumetric depth, masterpiece.

In additional showcase images, I will try to push the weights to their absolute visual limits.

For all showcase prompts, I will refrain from using any additional LoRAs, and I will share the complete metadata of the sample images. This way, you can easily read out and reproduce the exact weights used.

If you have any further suggestions for sliders in this series, please let me know. I will try to implement them.