Tutorial: How to create consistent characters in Midjourney

A practical guide with examples on how to create a consistent character using Midjourney.

1. Create the Initial Image or Upload it from the Internet

The first option – just to use some ready-made photos. To begin, you have the option to either upload images that inspire you from the internet onto Midjourney Discord or create an AI image using a prompt.

1.1. Upload a ready-made image with a human.

If you choose to upload an image, you should click on “+” in the Midjourney input field:

And then – just upload an image from your computer. After- just click on the uploaded image and open it in a new tab

and next – use such a discord link for the next prompts

They will have the same look as a simple prompt, example is below:

/imagine https://cdn.discordapp.com/attachments/997271689769529394/1084104151237144586/charlie_chaplin_42-18095616.webp in color, 90th style 

1.2 Let’s generate a human

This is a flow if you don’t have ready-made photos and want to create a digital model for later use in different locations and scenes. Each Midjourney generation is assigned a unique numerical value called a seed. If you upload an image from the internet, it will not have a seed value. In such cases, you should only use the image URL for the first iteration. Once you receive a similar image in the first iteration, you can retrieve its seed value.

However, if you have created the input image using your own prompt, you will need to upscale any one of the images from the initial grid.

You can modify and personalize this approach to fit your preferences. If you’re interested in learning more about light and film, as well as posing models, Nick has an informative thread series on those topics that you may want to explore.

For my prompts, I will be using the latest version V4c, so I won’t be using any specific parameters such as “–v 4” or “–q 2”. Instead, I’ll be inputting the prompts directly into the midjourney bot without any extra formatting.

/imagine street style photo of a woman, shot on Kodak Gold 200

Once we have our initial model, we are going to pick one, I went with number 2, red jacket, white hair. This is our gen0, Kate.

2. Step – Upscale and make more

So now that we made a pick, we upscale, and make sure our upscale looks crisp, we don’t want any major artifacts in the image

To obtain the seed value used by the Midjourney bot, we simply ask for it through the bot’s interface. We can locate the message containing the image grid in the Midjourney bot’s conversation and click on the reaction button. Then, we can click on the envelope icon that appears.

If the envelope icon isn’t immediately visible in the reaction menu, try typing “envelope” into the search bar within the reaction menu. Once you’ve clicked on the envelope icon, you’ll receive a direct message from the Midjourney Bot with the information you’re seeking.

In my case it’s 2553048472. This value will help create consistent images.

Now we need to take the initial image URL and the seed phrase and make a new prompt. For receiving the initial image URL click on the “Web” button as shown below…

and save the image address from the midjourney.com side. Just make a right click and select “Copy Image Address”.

Kate 1

Almost everything is ready. In the new prompt, we will use the image URL + our written prompt, but this time we want a side view or profile shot + the initial seed. An example for my prompt is below:

<my img url> street style photo of a woman, side view, shot on Kodak Gold 200 --seed 2553048472

And finally – we have got consistent results:

You can repeat this process as many times as necessary until you have obtained 1-2 images of the person’s left and right profile or side angle view that closely resemble the initial image. Once you have identified the most similar images, you can proceed to upscale them.

By now, we should have gathered a minimum of 3 photographs, or even more:

  • Front view (1-2 images)
  • Left side view (1-2 images)
  • Right side view (1-2 images)

We can combine these images with a new prompt to generate an above-angle view of the face. It’s important to use the same initial seed number. Be sure to keep re-rolling the prompt until you’re satisfied, keeping in mind that we’re aiming for outputs that closely resemble the existing images we’ve chosen.

<front> <left-side> <right-side> street style photo of a woman, extreme high-angle closeup, from above, center view, shot on Kodak Gold 200 --seed 2553048472

the first left one is what we need

Repeat this process for the below-angle view, bearing in mind the reference image we examined earlier. Our goal is to capture as many angles of the face as possible and create a comprehensive map of its features. But Midjourney works better with up to 6 images in the prompt.

/imagine <front> <left-side> <right-side> <top-view> street style photo of a woman, extreme low-angle, from below, centered view, shot on Kodak Gold 200 --seed 2553048472


3. Combine it all

At this stage, we are incorporating 7 or more images as sources of inspiration, representing the various angles of the face that we have selected. It’s likely that the resulting output will feature a yellow jacket, as all the inspiration images we’ve chosen feature either yellow or gold jackets. We will no longer include the “–seed” parameter since it’s no longer necessary, and we don’t want to start from the same noise pattern as before.

/imagine <front> <left-side-1> <left-side-2> <right-side 1> <right-side 2> <top view> <below view> street style photo of a woman, shot on Kodak Gold 200

As we can observe, not all of the images generated may turn out to be of high quality, and it might be necessary to re-roll the prompt. Nonetheless, most of the time, we obtain at least 1 satisfactory outputs.


4. Put our model in different scenes

Thus far, we have created images that are of a 1:1 aspect ratio, without utilizing wide-angle shots. We have not given Midjourney any specific instructions regarding the images generated.

Now, let’s attempt to place our model in a studio photoshoot setting. We will modify the background, and shot type, encourage her to smile, and adjust the lighting and film type. We are introducing and changing several parameters to the prompt that were not present in the original inspiration images.

/imagine <front> <left-side-1> <left-side-2> <right-side 1> <right-side 2> <top view> studio photo, white background, of a woman in a white outfit, ultra wide shot, full body, center view, studio lighting, shot on Fujifilm Pro 400H --ar 16:9

Let’s try something else. We can try to put on some really crazy outfits, here is the same prompt, but changing her look .

<front> <left-side-1> <left-side-2> <right-side 1> <right-side 2> <top view> street style photo of a woman in a pink fluffy blowup dress, ultra wide shot, full body, center view, smiling::2 studio lighting, shot on Fujifilm Pro 400H --ar 9:16

Let’s make Kate smile) We need to add the tag to the prompt:


So, the final look of the prompt:

<front> <left-side-1> <left-side-2> <right-side 1> <right-side 2> <top view> studio photo, white background, of a woman in a white outfit, ultra wide shot, full body, center view, smiling::5 studio lighting, shot on Fujifilm Pro 400H --ar 16:9

Leave a Reply

Your email address will not be published. Required fields are marked *

10 Ways to De-Stress Your Day Hour by Hour

10 Ways to De-Stress Your Day Hour by Hour

How to plan workout when your mind and body are all over the place

50 Best Midjourney Architecture Prompts

50 Best Midjourney Architecture Prompts

Midjourney Architecture prompts visual examples, structure, and best practices

You May Also Like