Message From E.A. Bear:

Did you enjoy the holidays? I sure did! This was my first year celebrating Advent as a confirmed Roman Catholic. It was a whole month long celebration of feasts, parties, service work, and lots of singing!

Some of you have asked me how I create consistent characters. I'm still working out how to answer this. For now, here is some transparency:

The tools I use are Midjourney, Dalle-3, Kanvas 16 Tablet, and Clip Studio Pro.

I will now attempt to explain how I create consistent characters. This assumes you know the basics of prompting and are familiar with Midjourney and Dalle-3.

Midjourney Pros and Cons
It's very versatile with many options for forcing the results you want. Despite this, Midjourney is terrible for consistent, complex details and dynamic poses. There are many ways to apply force, such as modifiers, sentence syntax, and text weights.

Experience has taught me that if you want something you have to show Midjourney what you want. I prompt with 2 - 3 images and will continuously repeat the process until I get the desired results. I apply another technique called split-testing which is typical of ad copy because I want to go wide with many variations.

I will sometimes use the in-painting feature called "Vary (Region)" to redraw using 1 - 3 reference images and a simple text prompt. The text reinforces the visual otherwise the output looks strange.

Recently I am applying hand-drawing techniques to these images with incredible results!

Dalle-3 Pros and Cons

It turns out Dalle-3 is highly versatile and excels at rendering dynamic poses with multiple characters. It is so good I was able to produce Cross+Word within days.

The drawbacks is that it is highly lobotomized with sanitation controls against extremities, obscenities, and violence. It has a special animus against anything that could be copyright, even if you own the copyrights and are providing your own images.

Dalle-3 is better at character concept creation than Midjourney. I did a comparison test and was happier with Dalle-3 than MJ.

Dalle-3 is conversational as opposed to Midjourney. The bulk of my work here is done by clearly articulating what I want, and letting it know my likes and dislikes.

I am able to upload ten images on desktop and four on mobile but the results are lukewarm. It takes many attempts to get the desired results. There is a lot of drift!

Unlike Midjourney where the bulk of the work is done in showing it what you want, Dalle-3 is primarily conversational with the images providing minimal reinforcement of what you are saying. Make of that what you will and try it out for yourself!

So To Recap:

Midjourney

• 80% of the work is showing it what you want with 2 - 3 images
• 20% of the work is text prompting to articulate the details
• Few restrictions

Dalle-3:

• 80% of the work is conversing
• 20% of the work is using image uploads to reinforce the conversing
• Highly restricted

Using Dalle-3 and Midjourney Together With Kanvas 16

Images produced in Midjourney can be fed into Dalle-3 and vice-versa with the corresponding prompting work.

The Kanvas 16 allows isolation or consolidation of particulars. For example, I can isolate a character and use Midjourney to create new dynamic poses that I can photobash for new prompting!

For instance, I used Clip Studio Pro with my Kanvas 16 to create multiple flags from scratch for the Samaritans and Judeans for Phillip The Evangelist, which I then fed in as new image inputs to create more interesting results!