nol1fe Profile picture

Nov 23, 2022, 11 tweets

I assume you have a stable diffusion set, if not get it from github.com/AUTOMATIC1111/…

Model Anything 3.0
huggingface.co/Linaqruf/anyth…

drop it into directory \Stable Diffusion\models\Stable-diffusion

btw. waifu diffusion is cool too!

Now choose one option for the first step
1A) Create the shape of the image you want to generate in paint/any software - "best results"

1B) Create a white sphere on black background - with high denoise we will get items from the prompt placed nicely on the black background - "creativity"

1C) Generate a base image from txt2img to get what you want, but surprising results. Needs many generations before you get the shape you want - "slowest"

2) Go to img2img and put the image from the first step

3) Add prompt: sword, art station, HQ, 8k, stylized, art by Granblue Fantasy, art by Ghibli Studio

Add negative prompt: (watermark:1.5), (logo:1.5), (text:1.5)

you can modify it as you want, add skulls, dragons, scales or fire

4) Set denoising strengths like
• for steps 1A and 1C about 0.5-0.7 with a CFG scale between 7-13 - generate until cool results
• for step 1B aim for higher values like 0.8-1 with a CFG scale between 7-13 - generate until satisfying results and go with the image to step 2

5) Enjoy the swords

6) And follow @followmarcos and @emmanuel_2m for good content and different approaches in this area

Share this Scrolly Tale with your friends.

A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.

Keep scrolling