Source: AGI Innovation Lab
Image source: Generated by Wujie AI
Many people are working hard to create a set of continuous comic characters and scenes in Midjourney.
Combining my findings and techniques shared by other creators, this article will solve this problem once and for all.
In theory, you can now create images for an entire comic book.
Let's delve into it.
In this article, I will show you how to:
- Create consistently continuous comic characters
- Make them perform any action you want
- Design consistent scenes and backgrounds
This will be an epic ability.
In this tutorial, we will use the Midjourney Niji mode, which is very suitable for creating comics.
Niji has 3 styles. The x effect of the scenic style is the best.
To open these settings, add the suffix "--niji --style scene" after the prompt of Midjourney.
Before we create the images, there is one more thing.
This can be done by adding "split into 3 different images, taken from multiple angles" in the prompt.
Now, let's generate the image!
This is the image I generated.
Observe how a single image is divided into 3 frames, all containing the same character.
Because we are using the scenic style, the photos are in landscape format, which is ideal for our next step.
We can use the inpainting feature to expand each image. Our previous article introduced this feature of Midjourney (Midjourney launches "local retouching" feature, can directly "partially repaint")
Simply enter new prompts and paint on the other 2 images.
For example, I want the girl to stand by the lake, so I used the prompt in the image below:
Just like magic!
You can repeat this process for the other 2 frames in the original image.
But anyway, we currently only have 3 images to use.
What if we want more photos of the girl doing other actions, like a wide-angle shot of her jumping?
We will inpaint the original image again.
But this time, when generating it again, leave some space between the selected area and the original image (see the image below).
After that, we will change the prompt to a wide-angle shot of the girl jumping.
There is a problem: this girl doesn't look like our original character.
Solution?
Simply repaint to change the face again.
You can use the same prompts as before.
By repainting for the second time, it's actually easy to make the new face look like our original character.
She looks much better now.
By the way, you can also use inpainting to change her clothes, or make any other adjustments to the surrounding environment.
In the image below, you can see the same girl wearing 4 different outfits.
The prompt I used was simple:
Light blue clothes --niji --style scenic
The effect is good.
Using this technique, we can create any number of images.
This way, we achieve consistency of the characters.
But what about the consistency of the scenes?
Suppose we want the castle in the first picture to appear in the background of the second picture.
What should we do?
Answer: Inpainting through image prompts.
Steps:
- Crop the castle from the first picture.
- Upload the cropped castle image to Discord.
- Copy the image link of the castle.
- Use the castle image link as a prompt to retouch the second picture.
That's it!
The same castle now appears in the new picture.
Finally, consistency of the characters and scenes is achieved.
In theory, you can now create images for an entire comic book using this method.
Reference: @chaseleantj
Wujie AI recently launched the MJ model, which allows users to draw using Midjourney v5.2 and Niji v5, supporting input of Chinese and English text, with an interface that is more in line with the habits of domestic users, and with multiple style choices, while the cost of use is lower than the subscription cost of Midjourney.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
