Breaking news

Stable Diffusion Exercise, Warg Trooper Hana


Hana the Warg Trooper and Big-Bad-Doggy the Warg on her crotch

I want to ride this monster into the battlefield and translate the imp language in both directions?Hana felt the cold sweat on her forehead being blown by the gust of wind, which made her head hurt.
“Hana, are you okay?” Professor Shannon checked her weapons and stared at the warden as he saddled her mount. She still had time to turn around and care about Hana.

I can’t! I only know about forty little devil words, and they are still childish versions of “eat rice” and “fight”!Hana howled inwardly.

“Yes, I can.” Hana nodded in response with a firm look in her eyes.

This article will briefly talk about how to use simple inpaint to quickly create the visual design you want.

First of all, this picture is a visual design produced for a certain chapter in the novel, which gives me a reference when writing the novel. Thanks to the visual stimulation this picture gave me, I was able to successfully write the outline of the entire plot.

The completed chapter, part of which is posted onWarg Trooper Hana — drawn by Stable Diffusioninside.

When designing this image, the basic concept was to have the protagonist ride on a warg in a fantasy world, following a raven in the sky on a mission against a smoke-filled background.

For the entire work, starting from the initial draft using prompt words, I used inpaint to change the required parts bit by bit. The following is an animation picture demonstration:

Simply put, the process is probably:

Using prompt words to run out the initial pictures, I randomly generated about fifty pictures in total, and selected the one that I thought best met my needs as a starting point.

Starting picture

The problem with this picture is that the wolf’s feet are completely wrong, these are horse limbs and a wolf’s limbs should look like this:

Wolf, photo credit: Britannica, Inc.

Wolf, photo credit: Britannica, Inc.

Another problem is that the background should be plains or forests rather than mountains to fit the novel’s setting, and the character on the mount is not what I want.

First, I start with the most important character. Here I use inpaint to frame the entire character, and then use my own Lora to match the clothing prompts, with a denoising strength of 0.7, and ControlNet OpenPose to fix the character’s limbs and The head is the completed character draft:

Role replacement completed

At this time, the character’s clothes have not yet met my requirements. The short skirt should be red, and the feet should be wearing boots. The most important thing is, why are the reins missing on the hands?

At this time, you need to use drawing software, such as GIMP, to make a simple circle selection, then change the color of the skirt to red, then cut out the reins and black gloves from the previous picture and paste them on the new picture.

After completion, use inpaint to integrate the image into the semi-finished product image using a low denoising strength of 0.3 for these parts.

Then for the feet, use the prompt word black knee boots, combined with the denoising strength of 0.6 to create boots:

Clothing replacement completed

Next, use ControlNet Canny or Lineart to fix the lines, and first inpaint the skirt, then inpaint only the face, match the character’s exclusive Lora, and select Inpaint area: only masked to finely redraw the face to get a face Detailed semi-finished product drawing:

Facial detailing completed

In the design of visual communication, a picture will have a very important point to convey to the reader, usually the face or the intersection of moving lines. In this picture, it is the position of the character’s face, so it must be redrawn carefully. , allowing people to lock the face at a glance andLet readers expand the picture through the visual movement line from top to bottom such as face->wolf head->ground..

After getting it done, start inpainting to select the background, and replace the mountains in the background with scenes such as cloud forests and plains that better suit the needs of the novel.

After changing the background, inpaint the limbs of the wolf and use wolf legs with a denoising strength of 0.7 to have fun. See if you can get normal wolf legs. Here you need to use Inpaint area: Whole picture to let the AI Know the position and posture of the entire wolf’s body to avoid making skewed limbs.

After the limbs are corrected, use drawing software to paint simple wolf ears, reins, flying raven and other parts, and then use inpaint to redraw these parts in stages to make them Enough to fit into the picture:

Completed picture

Finally, enter the extra page to enlarge it several times. Since after zooming in, you can notice that the character’s face is still not detailed enough, you can use drawing software to cut the character’s head into a 768×512 sub-picture, send the head sub-picture back to inpaint, and use 0.4 to 0.6 denoising for the face. strength to redraw, then return the redrawn sub-picture to the big picture, and carefully align and sew it back:

Sub-pictures that have been re-refined on the face

Sub-pictures that have been re-refined on the face

In this way, a large picture of 3400×2300 is completed, and the detail of the entire character’s face is ensured.

The entire conceptual design drawing took about four hours from the first original drawing to completion.

I wish everyone a happy AI graph calculation!

The article is in Chinese

Tags: Stable Diffusion Exercise Warg Trooper Hana


PREV Entrusted by ceramic artist Chen Jinwang, a knitting expert recreated the figure of PUMA, the loyal protective dog, with wool felt 1:1 – Free Art and Literature Network
NEXT Artist DavidCow Zhang Tianxing’s Hong Kong solo exhibition “The Core of Immersed Nature”