Sign In

How to: Pose Your Character "Flawlessly"

17

For a long time, I realized that the thing that took my time the most in stable diffusion was giving a complex pose to the character I've created. In this article, I will tell you step by step how you can pose your character in the fastest and simplest way possible.

Step one (Posing):

  • Set the desired camera angle and pose on the "Posemyart" website.

  • Export as Image and Depth(Don't forget to adjust the depth scale).

  • Then open Image in Photoshop and draw your character's look on the pose image.

Step two (Depth):

  • Use img2img with your drawn character and use ControlNet depth model set as shown in the image below.

Step three (Refference):

  • With your original character's portrait image, use ControlNet refference with default settings.

  • my character Alice:

(samdoesart:1), (dreamlikeart: 1), portrait of a beautiful young bartender girl, Britt Katrin Filippa, medieval anime art style, black and white medieval outfit, short skirt, white background, simple background, masterpiece, medieval bartender girl, blonde hair, blue eyes, small breasts, front view, looking front, looking at viewer, character portrait, cleavage, intense shadows, ambient light, illustration, thick outlines, highres, drawn by Inoue Takehiko, (Yoji Shinkawa:1.1), Jakub Rozalski, (kuvshinov: 1), sam yang, <lora:sam_yang_offset_right_filesize:0.7>, <lora:EyesGen:0.3>
easynegative badhandv4

Step four (Adjusting):

  • Add few more words to help AUTOMATIC1111 to understand the pose:

    looking from behind, close angle, close-up, bending over, ass, petite ass, back view, symmetrical, 
  • Set DenoisingStrenght: 0.5

  • CFG: 9

  • Steps: 20

  • Scale: 1.0

  • And create 512x768 image2image picture.

Step five (Upscaling):

  • After creating 20-50 pictures, Choose the pictures that looks the best.

  • You can now disable the "dept" and "refference" in ControlNet.

  • And repeat the process with the pictures you chose with (don't forget to adjust the little details in photoshop):

  • DenoisingStrenght: 0.55

  • CFG: 9

  • Steps: 30

  • Scale: 1.5

  • And create 768x1152 image2image picture.

Step six (Final Step):

  • Now downscale your 768x1152 image to 512x768 (in Photoshop).

  • Put the downscaled image to image2image and maximimize the scale according to the capacity of your graphics card (mine is 1.9).

  • From here you can adjust the Denoising Strenght, CFG, Steps as you want. (I use Ds: 0.55, CFG: 9, Steps:45)

  • Then Upscale the image by 2 (I use 4x-AnimeSharp).

    Result:

Tips:

If you wanna create consistent looking face on your character you can give your character a name on Random Name Generator and put it in your prompt.

Don't forget to use Inpaint for detailed face or clothes. You can also use "ADetailer" for beautiful eyes and hands etc.

50

Comments