UPDATE V2

Re did the lora using the free lora trainer on civitai, reacts better to prompting for certain lights. Read trigger words. Might update further with more data so it can mix different lights together better. Found that this version might not play too well with stylized but feel free to try, all images done with Virile Reality V3.0 BETA 3

READ!

This is my first lora, the goal was to get more interesting lighting and contrast in images.

Yes, this will have a strong impact on images, it will change the contrast, colors and subject at higher strengths. I will leave tips on how to use it below, but please experiment with it further than me. I have attached some sample images made with the lora and some xyz plots to see the lora’s effect. 

HIRESFIX is HIGHLY recommended!

TIPS ON HOW TO USE IT

If you are fine with it changing the image, then you can simply use it as any other lora. recommended weights from my testing and personal use, 0.4 - 0.8, this all depends on what you want it to do and if you are using it combined with other lora’s. 

Another thing that I have been doing when I feel like the lora is holding back the flexibility, do  THIS. When using it, use it at lower weight in the prompt, 0.2 - 0.4. Then when you find an image you like, use hires fix or img2img, now you can use the lora at a higher strength in the hires fix prompt. I like doing 0.4 in the normal prompt and 0.8 or higher in the hires fix prompt. This allows you to layer more lora’s and less changes with the subject of the image.

ALSO! I recommend trying different types of lighting, like “Green lighting, green tint” etc… I usually combine this with BREAK at the start of the color. 

Lora was made for realism but works great with stylized.

描述:

Read desc.

训练词语: