Shadowlight

Notice: To prevent malicious theft on other platforms, the model will be available for irregular downloads, and the image processing will continue as usual.

If you want to communicate and share, welcome to join the AI Mutual Aid Association, search QQ group number: 482570609

→→→→→→Get more high-quality models←←←←←←

DarkLight: Bringing more realistic and artistic lighting/shadows, with higher contrast backlight without losing any details

As we all know, the SD1.5 image tends to be gray (not dark enough and not bright enough), lacking depth. Although C station has specific optimizations for lora, it is not very noticeable, and raising the weight too high can interfere with the environment. In order to better improve this long-standing issue, to unleash the potential of SD, I have trained this lora

Advantages of DarkLight

1. Higher brightness contrast threshold: Better shadows, especially in backlit situations (such as adding silhouette prompts)

2. Flexible and controllable: Adjusting the weight can modify the contrast

3. Lower weight: In numerous tests, a weight of 0.3 is often sufficient to have an effect

4. Preservation of details in the shadow areas: Particularly prominent on realistic characters (other aspects are still under testing)

5. Better black and white images: Understandable, as the contrast levels are richer

6. More vibrant colors: Surprisingly, but the contrast is not only effective in black and white

About Prompts

Most labels in the dataset contain keywords such as (1girl, spotlight, dark theme), so placing them at the beginning of the prompts works better

The 1girl can be replaced with any character, like 1man, 1cat

Adding a color word before spotlight, such as red spotlight, will change the light source to red. You may need to try multiple times, but the effect is already good without it

Similarly, dark theme can be modified with colors, like red theme, which will present a reddish tone in the shadow areas. Also, if you use green theme, there is a chance of a forest background, depending on the large model. I will work on improving this in the next version (perhaps)

Respect copyrights, respect individual labor achievements. When using this model elsewhere, please credit the author. This model is for learning and communication purposes only, not for any commercial activities

prompt:

1girl, spotlight, silhouette, portrait, parted lips, closed eyes, beautiful detailed skin, one hair bun, turtleneck sweater, dark background, <lora:spotlight:0.84>,

Negative prompt:

nsfw, logo, text, bad anatomy, verybadimagenegative_v1.3, extra limb, missing limb, disconnected limbs, malformed hands, ugly,

Steps: 60,

Sampler: DPM++ 2M Karras,

CFG scale: 5,

Seed: 1003785097,

Size: 640x960,

Model: Realistic Vision V5.1