🖌️ AI Art Tips Thursday #13: This is sugoi!

💡 AI Art Tips for the Day

We will explore creating art generation via the niji model by popular demand from the vote last week.

As usual, if you are keen to explore the official document from Midjourney about niji, Check it out here.

Just a quick background info, the niji model is created by people from Midjourney and Spellbrush. I have checked their site briefly but I can’t find out why they go with the name “niji”. However, according to the Japanese dictionary, the Japanese word “niji” means “rainbow”. It does make some sense as we are thinking about creating colourful anime-style art, so some colours from the rainbow would certainly help. 😆 

Cool, so let’s talk about how to create with the niji model then. To get started with niji, the process is actually quite simple. Instead of adding the usual Midjourney version number as a parameter, just add --niji 5 to get the niji model incorporated.

In terms of the inspiration for today’s creation, I have gone with the famous Lofi Girl from Youtube. If you don’t know what I am talking about, you can check it always-on Youtube stream here.

First, let’s test with some basic prompts.

I must say, without any further prompting, the result from the niji model is fantastic. The amount of details that is included is abundant and the colouring it uses really matches what you can see in a manga or an anime.

The test that adds the “in anime style” to the based prompt is also pretty fun. I must say the normal midjourney version 5 with that prompt works quite well already, however, if you consider the amount of detail included, the niji model certainly wins in this case.

Secondly, I want to test the different styles you can add on top of the niji model.

Once again, I am in awe of these generations. The details match the prompt I have got and each style represents itself quite well. For instance, if I am requesting something that would emphasise the environment, I would certainly go with the scenic style. If I want to try something different, the expressive style would be my go-to.

Lastly, since we are talking about anime, I can’t help but test if the art generation could match specific anime that I can name. For this, I am going with one of my childhood favourites, Naruto 🥷 

Well, I personally feel the style does not match the original anime I can remember from the niji model. Although, I must say the original midjourney version 5 captures the details quite well, just look at those symbols.

Cool, so in conclusion, this is what I found by playing with the blending feature

• The niji model can generate anime-style art with abundant details in the environment and the main object(s).

• To achieve specific style requirements, you can add --style [style type] after the normal niji model to fine-tune the results.

• The niji model won’t be able to mimic specific anime styles based on my test.

Try this in your next prompt

[Your prompt] --niji 5 --style [style type, e.g. scenic, expressive, etc]

🔤 Word of the day

Composition

n. refers to the arrangement of elements in a photo. It's about where you place things in your picture, how you balance different parts, or what you choose to include or leave out. Essentially, it's about creating an interesting and engaging photograph.

To check further explanation about this and some rules around composition, check this Wikipedia.

🖼️ Image of the day

Who doesn’t love some cats around while you are doing your assignment? 😸