Midjourney is pretty good at creating realistic human characters that do not exist at all and are 100% artificially create AI based images. Wouldn’t it be nice if you could take your own profile pic and feed it to AI to create different versions of yourself using the /imagine command in Midjourney. Well actually that is quite easy and you can force the character or face to be same by using image prompt as part of your text based prompt.
All you need to do is upload your profile pic somewhere on the internet or on the discord server (as you are using), get the public link of the image and use it with your prompt. To illustrate how you can create different profile pics of your face I am going to use couple of images from Unsplash.
Source Images
I am using the following two images and if you want you can copy the link of these images and experiment by yourself. Make sure the photo you use it a front shot of yours and without any sunglasses. This way Midjourney will have all the information about your face to create various versions of you.
Create your Profile Pic
Now that we have a source ie. our profile image that we want to you let’s get started with creating various AI Profile pics from here on. So the prompt will be <image url> photo of a man/woman, attributes, styles you can then keep feeding desired variations prompts to Midjourney to create multiple stylised AI images of yourself.
Let’s start with the female image sample from above. My prompt is <image url> cyberpunk gothic photo of a young woman, caucasian –ar 16:9
–ar 16:9 defined the widescreen aspect ratio that I want in the image.
Those experiencing issues generating these image try to use the –v 4 or –v 5 model in the prompt. This prompt and blog post was tested with these Midjourney models which at the moment you can still evoke using these arguments.
You can see that the images created represent the face of the same real person that we fed to Midjourney using the image prompt. There is an added bonus of “young woman” as she looks lot younger in her AI generated images. If you want to get a few more style options checkout our Punk Styles in Midjourney post focused on these various styles.
Here are a few more versions of her and the prompt in the caption.
now let’s try a few things with the young man’s photo which we selected above as the main image prompt. Let’s create a glamorous cinematic looking shot of this image by using <image url> glamour photo of a young man, lv jacket, wide shot, studio lighting, soho, shot on Fujifilm Eterna 400, 4k –ar 16:9
Wow right? here is another version but changing the Fujifilm film style. I shoot with Fujifilm cameras in real world so I now these profiles best but you can try others that you are familiar with from Canon or Nikon etc.
In case you are skeptical here is the screenshot from my Midjourney prompt. There is no other hidden secret here!
Checkout the below variations that I tried in square aspect ratio. You can try various different style prompts or simply if an artist you like just add “by <artist name>”. There are 1000s of artists that are recognised by Midjourney and you can check out the complete study of this in the post Artist Styles on Midjourney
Feeling Brave – Age Yourself?
If you are feeling brave and want to continue having fun try to add “old” in the prompt, so for example old man or old woman. Here is what you can expect from Midjourney.
Hopefully this post helps you create AI based profile pics of yourself using Midjourney. It is heaps of fun actually I am looking forward to what you create and if you are on Instagram or Twitter be sure to tag me @harmeetgabha
Update: 14 March 2023 – Saw another cool prompt so had to try it with my above source image.
Update 19 March 2023
A few people have expressed concern that they were unable to achieve the results with the above method. I have experimented in Midjourney v5 as well as originally with v4 and I can continue to see the same face in the results. Here is another source image firstly created in Midjourney v5
You want to use a front facing original source image that is the basis of your creations. After that its just a matter of adding what you want Midjourney to do with this image. Here are a few prompts I used:
- <img url> photo of young man, cartoon character –v 5
- <img url> a man as a fictional character by Simon Stålenhag, trending on artstation –v 5
- <img url> a young man as a fictional character by Aaron Jasinski, trending on artstation –v 5
- <img url> a middleaged man –v 5
- <img url> a young man by Alex Gross –v 5
You can see the examples below that you can same persons face but based on prompt you can vary the age of the character as well.
Readers Image
One of the readers of the blog reached out at they were having issues with using their image and not getting the same result as demonstrated above. I think lots of times it depends how well the portrait is that you are using as the sample. Also make sure you are using similar aspect ratio to the output that you want to get.
In this case Midjourney did struggle to produce the same face using this image. However you can force more of the image by using something called Image Weight defined as –iw <value>. Value of 1 is normal weight amongst the image and prompt. However to ensure more of the image is used you can use a value of 1.5 or 2. I would not suggest going higher because then more of the original image will be represented in the generated output.
(resulting images are not very much like the image on the left)
Once I force the original image weight in the prompt you get better results as seen below.
If you'd like to support our site please consider buying us a Ko-fi, grab a product or subscribe. Need a faster GPU, get access to fastest GPUs for less than $1 per hour with RunPod.io
Hey there – does this still work? I am following your exact prompts and it is changing the face – thanks!
yes it does.
This is a brilliant idea. I wonder how I could get the images to be similar enough to be frames of video to make a personal profile gif or something with a wink.
I am experiencing the same thing.
I will post some more examples but it could be the kind of image you are using. Does the image show your face completely? Its not a side pose/angle is it?
Thanks for the follow up on this! So I was able to make this work with your source file – I am still having trouble with headshots taken from my iphone? I wonder if it just needs to be a petter source picture or something?
here is my source image for you to tell me where I am going wrong! https://i.postimg.cc/P5GdG2vM/IMG-0019.jpg
I tested your image and got some good close results. Have a look at the blog post again.
No, it doesn’t work. Midjourney changes the face completely. It doesn’t look like the original photo at all. I feel disappointed about this program.
Yeah…that’s disappointing for a long time it was working very well
Similar issue – you do use the /imagine prompt or not at all?
Without the imagine prompt I get no results. With I do, but not splendidly.
of course you have to submit your prompt using /imagine prompt
I read all your post.
I try mj4 and 5 with your prompts and with simple ones, I try many photos, I try 1:1, I try -iw 2, I photo shoot my self with ring light and clear face.
Every time I get other people… 23 prompts and nothing…
Hi, I am agree with everyone it doesn’t work, even if I copy-past the same exemple as in your article. impossible to keep face ressemblance. Why?
Proof it works
I’ve tried too. Although all 4 images are basically the same person… it isn’t me. Close, but not me. I wonder if it’s dependent on which package you pay for? I notice you have access to the ‘stealth’ command.
plan does not make any difference. At the end of the day Midjourney is trying to create an image that resembles you, its never going to look exactly like your photo.
I really enjoyed your post but could you add in a little bit of info on how to optimize your selfie for this? I feel like no matter how i try i’m just not getting similar results. E.g.:
https://cdn.discordapp.com/attachments/1099835923677593620/1101362113483644969/1_by_1_Miguel.jpg
https://media.discordapp.net/attachments/1099831763875934289/1101506701414375556/-AI-Enthusiast_viking_a56eb911-a0e1-46f7-9ea3-15b781822e85.png?width=406&height=406
Or are these the types of results a person should expect?
It depends on the AI how it interprets your facial features. This will determine how well it can generate similar face. Ideally if you want consistency you need to train the AI model on several of your images and expressions. Midjourney is a quick and easy way to generate similar faces. I did this for my face using 16 images in Stable Diffusion (not Midjourney) and I could get consistent images of myself in various styles.
I think this to get your face in midjourney you have to run it multiple times, not just one generation of images,and insert link of generated photos too.
Not even close…. I’ve tried a dozen times… never remotely looks like me. 4 pics of random guys… not even same nationality.
Thank you SOOOO much for your post. Mine worked without any issue.
I also have tried this, and it will not use the same face. I just want to use my face and get a professional headshot. It changes my face completely ever time.
I think Midjourney model may be biased and therefore favouring certain kind of ethnic backgrounds. I tested with an Indian background person photo and the model was not able to reproduce the same face. It also could be that the model is updated and no longer supports such closely guide images forced by the prompt as it did before. They keep developing and evolving their platform.
It just doesn’t work, man. Lol. Maybe it worked before but now it doesn’t.
Yes I’m having the same problem as everyone, even with –iw 2 it doesn’t work.
Source image: https://cdn.discordapp.com/attachments/990816877691437086/1121823877593583676/HeadShot.JPG
Generated:
https://cdn.discordapp.com/attachments/990816877691437086/1121834400745787432/chrismills_change_jacket_to_black_9c6be898-cf68-4145-80e6-f3d76e2c48ec.png
I notice it says, Fast when processing, is there a way to make MJ take more time?
Thanks
thank you but it does not work for me
does not work, uses random face photos instead
Can you do this with an African American image? I notice MidJourney for some reason has troubleshooting duplicating African American people / images from a photo uploaded, it looks like a completely different person wearing the same clothes.
Yes, you can but maybe good idea to use –no argument to exclude Caucasian in the prompt. MJ has tendency to default to caucasian people by default
Does this still work for you? I will try your instructions but I’ve tried my own selfies many times as the character and never get good results. I haven’t tried the image weight. Also, can you please tell me how best to learn stable diffusion, since you got results there. I’ve tried but very complicated even to get to run on a computer. Did you take a course or how did you learn?