Midjourney Trick: Consistent Character in Different Images

➡️ ACCÉDER A chat gpt français en cliquant dessus ! ⬅️

Creating Consistent Characters from Different Perspectives

In this video, the presenter demonstrates a technique for creating consistent characters in multiple images, using mid-journey. The technique involves merging images and describing the character for mitually to generate consistent results.

Finding Inspiration on Mid-Journey

The presenter first starts by browsing the mid-journey community feed to find inspiration for merging several images. They identify two images that they want to merge but are looking to have the same character in both images. The character they want to use has blonde hair, wears a red and white uniform and is sitting on a throne.

Using Prompts to Describe the Character

To create the same character in multiple images, the presenter links to both images in the prompt and describes the character’s appearance. They use descriptive words such as « white short beautiful young fake Queen. » Using the prompts, mitually generates beautiful images of the character in different poses.

Using One Image to Merged into Any Background

The presenter demonstrates that the same technique can work with just one image of the character. They link to the image in the prompt and describe the person’s appearance as « woman wearing a white tank top and black shorts walking in a park. » As a result, the presenter gets multiple images of the person in different poses.

Using a Prepared Scene for More Control

The presenter shows that using a prepared scene gives you more control over the style and lighting of the background. They describe the scene as « Paris on a summer day photorealistic 8K AR 16×9 » in the prompt. After retrieving several images, the presenter selects one and links it to the image of the person. The resulting images show the person walking in Paris with the Eiffel Tower in the background.

Upscaling the Image for More Detail

The presenter recommends upscaling the images to give them more detail, making the characters more photo-realistic.


The presented technique is an amazing way to put a person into all kinds of backgrounds and create graphics. By linking images and prompting mitually of the character’s appearance, users can get consistent characters in multiple images. This technique is perfect for creating graphic novels.

Source : Olivio Sarikas | Date : 2023-02-08 21:00:08 | Durée : 00:06:30

➡️ ACCÉDER A chat gpt français en cliquant dessus ! ⬅️


Olivio Sarikas

#### Links from this Video ####

Consistent Characters from different Perspectives: Create https://youtu.be/Dl9orWUSREY

Blend Images in Midjourney: https://youtu.be/-9R_sYpdfiE

Tokenized AI: https://www.youtube.com/@TokenizedAI

mansh singh

I am unable to join two images it gives invalid link

Joan Warren

How do you grab the link to an upscaled creation?

Joan Memoz

I loved your energy ! Your information was simple to understand and interesting to watch. Thank you for sharing this Olivio. What do you think about BlueWillow? Do you have an tricks for it too?

Maxine Wairimu

This was so informative Olivio, thank you for sharing this gem😁🤩 What are your views on BlueWillow? Please share tricks on this tool as well


Thanks for this cool tip! And: if I were not married already I’d help you out with not being single anymore 😃

Eddie XYZ

but, what if i have 2 images and i want them kissing each other? how do i have to do? image 1 url an it descriptions then image 2 url and it descriptions and then a general information about what i want? is that possible? making two persona of different images interacting in one?

Archetype Alch3my

Im getting different images every single time even after adding seeds and links, what am I doing wrong

Full Sail Game Dev

I'm surprised that it let him upload that image with the small top and shorts because I tried uploading a image with a girl in a 1 piece bathing suit and it's like NSFW!!! denied. I'm like it's a freaking 1 piece bathing suit midjourney

Pala kush

Pleaaase remove urself from the corner..its distracting


I guess we have to use American English. 🤔

Elliot Jenner

Can you do this in stable diffusion (A1111 or invoke)? I know you can use img2img but that only takes one image (ie the thing you did in the middle) and not combining 2 images as was the main idea.

Javi SánChef

Olivio, I have a question. Midourney is free???

Short and Cut

Bro look how you smile doesn't matter if your single as long are your happy!


What i should write on <img1> <img2> ? Is there any spesific code?

Akmal Ahmad

How to do this trick in Stable Diffusion?

Brian Jilg Music

@3:35 girl has the six knuckle shuffle going on! Lost track on her right hand too! Thanks for he video though, this is useful to see how to get consistent characters. Much appreciated!

Georgette présidente

If a girls walks in that outfit in Paris, I don't give her 10mn of life expectancy… Thank you for the tutorial tho :p

Сергей Казаков

Use ""–seed" so that the character is as similar as possible everywhere

Glenn Wilson

Ok but what do you mean by "image?" I have tried everything adn I cannot replicate this method. Do you right click the image and copy the URL link and post that after the /imagine? Can anyone help me with this? Just saying IMG is so vague to me


Perfect. Thanks so much. This is what I was looking for.

Le Pouilleux

3:35 8 fingers and 6 on the other hand.

Susan West

I enjoyed this clip, thank you for sharing Olivio. Please share tricks on BlueWillow next


How can this be applied in Stable Diffusion (Automatic 1111). As far as I am aware, you can use only 1 reference image. Is there an extension to change that?

James van der Walt

I'm so frustrated. I can't get midjourney to render a consistent fantasy character – a reptilian cat humanoid in this case. All the tuts focus on human or cartoon characters. With human characters I managed easy enough but the changes for fantasy creature characters over multiple scenes just doesn't work. Any tips?

ganimed1969 Gani

Hey, thx for great info! But I also have a question. Recently I tried to make avatars for friends with midjourney, but it's pretty hard to get faces that would resemble photo that I upload into midjourney. So sometimes people kinda recognize themselves, sometimes not. I tried to describe people as well, but still – faces are rather different. Is there a method that will make midjourney clone a face and then kinda mingle it into a different ebvironment?


I've noticed, unless I'm missing something again, that both in Invoke and A1111, image links do not work. If I use a image link in front of my prompt all I get are different shapes

Steve Clarke

Great video again thank you


Congratulations on your YouTube plaque Olivio


Hmm…somehow didn't work for me. I tried using the same seed, but as soon the words in the prompt change, the composition of the drawing changes.

Veuillez vous connecter pour laisser un commentaire.