Model made from many tests generated by stable diffusion.

I have tried many versions of the same prompt with many different models to achieve this.

Recommended Values:

• Weight: 0.1-0.4 (gets really distorted at higher values)

• Activation Word: Fantascape

• Prompts like: landscape, waterfall, volumetric fog, forest, mountains, greenery, river, valley, volumetric reflections & other landscape prompts

• Model: can use any (try batch checkpoint & prompt through all your models to find out what you like the most)

• Sampler: Euler a is enough

• Samples: 20-40 for high quality, <20 for more artsy images

Original Prompt: (training images)

• Positive: high quality, high resolution, extreme detail, masterpiece, great composition, (face:0.9), disembodied hands, (((hands:0.8))), (((forest))), valley, river, (trees:1.2), grass, (mountains:0.2), flowers, bushes, roaming animals, weeds, long grass, nature, sky, clouds, ((greenery)), ((day)), bright picture, (((in the style of Zdzislaw Beksinski))), (((surreal))), (((colorful))), ((dark photo, dark scene))

• Negative: (bad quality), (worst quality), ((out of focus)), text, signature, watermark, username, night

• you can use verybadimagenegative textual inversion in negative next to the prompt i wrote here