The secret sauce is the --cw parameter. It is set to 100, which tries to keeps 100% of the input, so that is why the shirt stays the same. Lower values (even 0) will keep the facial features a lot of the time, but will adapt the clothes and lighting according to the prompt.
Definitely a good parameter to experiment with. I thought keeping the clothes was a useful feature to showcase for this video (because my old Consistent Character Technique didn't really support that), but I can think of a ton of use cases where you would prefer a lighter character weight
@@glibatree I thought so, you did mention keeping the striped shirt will help the recognisability of the character. But of people that want to see their character dressed like a fireman of astronaut, this helps. I did try to put myself in it, you did a far better job of that than what I managed to do. I look like a mad dwarf ready to eat anybody's soul at the drop of a dime. Must experiment more with that...
I think you might be lucky with it to be honest. It can create your look really well. No matter how I tweak it, it NEVER consistently gets my look even with many photo references. I'm pretty experienced as a prompter too, it just always comes out with look-alikes, but not one would be able to fool anyone.
I love this and I love your videos!! I'm just learning how to use Midjourney and your content has been so helpful. I'm actually mindblown 😂 It's just amazing. Thank you so much!
Great video thanks and I’ll try uploading a bunch of images of me as I used one yesterday and I looked terrible in the outputs, older and harder face so I’ll give it a try with 5 images and hope can get to work as -cref is working great creating new character w consistency
Thanks for your video, i uploaded 10 photos and tried a lot different parameters and desciptions, but always 10 or 20 years old more than the picture lol
We have had this for a loooong time now as open source and free. Thats how all the AI influencers are done. Midjourney just keeps lagging behind, but it finally got ipAdapter that was released 8 months ago.
What a scam. Bought the pro version and it is absolutely impossible to make a version of your self. Have tried all gpts and ai bots on earth for prompts and it WILL NOT work.
Is there a way to do this where we can have different clothes / scenes? E.g., looking like a doctor in scrubs or looking like a runner in workout clothes, that kind of thing. Also, I see you have multiple GPTs -- should we use the consistent character one? Or the one with the most uses?
Great video! I find Midjourney still has trouble with me, a female; it gets some parts right but the images still aren't "me". MJ has a tendency to make women unrealistically beautiful imo. It does my husband really well, though.
That does make sense, I may have to experiment a little for the best ways to counteract that. Glad you liked the video, and good luck on your generations!
Hey! Thank for the video but this just doesn't work for a lot of people. You somehow got lucky with the type of face you have. Tried various versions, different set of photos, prompts but it doesn't even generate close enough images. Bad luck.
Great video. I tried this but the output was a character that had similar features but looked nothing like the uploaded character references. I set the --cw parameter set to 100. Where am i going wrong?
i am not sure if its because im Chinese and that MJ isnt trained enough on how chinese people look like , but i followed the exact steps and the gens does not look like me at all 😝
Good afternoon, tomorrow May 5 in Spain is Mother's Day, with the previous version of MID JOURNEY on Discord I knew how to fit your face on another character but now I use the alpha version, could you help me to remember my mother and give her a new character. In short, how can you include a face in a character, a portrait, in the alpha version? Thank you very much in advance
And to your edit, yeah totally welcome to take the glibatree prompt off, but I have never had it hurt a prompt. I just like to see them on the popular feed haha :)
what if i have already an image that I like and I need to swap the face on it. for instance, I want to add my own face to a beauty model photo I generated in mid-journey. Also how to do it with an image of multiple faces?
Poses, expressions, and lighting match the dynamics of the scene with little to no effort on my part. I find this a very impressive feature, that I think many people have been waiting for
somehow that doesn´t work with my face - it always generates characters a bit similar to me, but never, really never, I got results like you did in the video. Also someone from J said, sref doesn´t work with own images? it´s really weird, but i never got any character looking like me, no matter how many images I tried :D
Same! I cant get it to work for the life of me, maybe some humans are easier to create since their look is more common? It keep making a weird old ugly version of me lol.
@@samferriere5730 I haven’t tried the multi angle photos with same shirt and white background tho. I will try that when I have time. Was hoping it could just take a slew of random pictures of me, but may need more concise references.
I have written a very interesting book and I want to publish this book through midjourney or other platforms with video pictures and audio. Can you do this for me? I’m ready to pay or you can suggest me who can do this for me.
Thank you, but what is the solution for those who don’t have access to the midjourney website. Discord only works with an url but my images on my computer have no url😅
So its possible to make pics of our selfs? I thought in first release that would not be possible, as they were afraid people would misuse this feature.
Yeah I am surprised to, but I found it was very effective at creating me - and it seems like other people are having similar results. Doesn't work for everyone - but I'm loving it...
I'm kinda playing a long game here, but both DALLE and Midjourney are models that get regularly retrained. I am dreaming that, because I really trust that artworks that come out of my GPT, the Glibatree tag will become a useful filter that improves the quality to the end result. I also like seeing "Glibatree prompt" show on Midjourney's front page or in random blogs and news articles hahaha
Thanks man. Glibatree is my absolute favorite. After generating four images, I turn around and put the best one in my GPT called the Canvas Critic. After a lengthy critique, it regenerates the image and I’d say 60% of the time, the image comes out even better, frosting on my cake! Thanks for updating yours consistently.
Im having trouble doing this on mobile without access to fo the Beta website. But i watched the video while driving, gonna have to rewatch. 🎉🎉🎉 AWESOME
You can only select up to 20 image prompts at a time. That can be a mix of character references, style references, and regular image prompts, or all of any of those. In discord there is a separate character limit, I think 4000.. so it mainly depends on the length of your links.
I'd say it depends way more on the prompt matching the face itself. Inconsistencies between the prompt and the characters (or lack of description) makes Midjourney really suffer
The secret sauce is the --cw parameter. It is set to 100, which tries to keeps 100% of the input, so that is why the shirt stays the same. Lower values (even 0) will keep the facial features a lot of the time, but will adapt the clothes and lighting according to the prompt.
Definitely a good parameter to experiment with. I thought keeping the clothes was a useful feature to showcase for this video (because my old Consistent Character Technique didn't really support that), but I can think of a ton of use cases where you would prefer a lighter character weight
@@glibatree I thought so, you did mention keeping the striped shirt will help the recognisability of the character. But of people that want to see their character dressed like a fireman of astronaut, this helps.
I did try to put myself in it, you did a far better job of that than what I managed to do. I look like a mad dwarf ready to eat anybody's soul at the drop of a dime. Must experiment more with that...
@@glibatree kindly not working
I think you might be lucky with it to be honest. It can create your look really well.
No matter how I tweak it, it NEVER consistently gets my look even with many photo references.
I'm pretty experienced as a prompter too, it just always comes out with look-alikes, but not one would be able to fool anyone.
exactly !!!! i have not found success in it either
Yup, I second this. He got pretty lucky imo 😅
Or maybe he’s famous enough that midjourney can make it very similar to the reality
same
same .. hours trying diferent photos and promts, nothing
Can someone help
I just can’t seem to get a single character that looks like me. Guess I don’t have “one of those faces” lol
I love this and I love your videos!! I'm just learning how to use Midjourney and your content has been so helpful. I'm actually mindblown 😂 It's just amazing. Thank you so much!
Great video thanks and I’ll try uploading a bunch of images of me as I used one yesterday and I looked terrible in the outputs, older and harder face so I’ll give it a try with 5 images and hope can get to work as -cref is working great creating new character w consistency
Good luck, do let me know how it goes!
Not sure what version you're using? V6? What settings are you using? Tried uploading multiple images, but nowhere near what I look like.
You're a great man on this platform 👏
Thanks for your video, i uploaded 10 photos and tried a lot different parameters and desciptions, but always 10 or 20 years old more than the picture lol
Haha, I get completely opposite results. I look babyface on all of the outputs :D
ive uploaded my pictures and the generated photos dont look anything like me .. WHY? cananyone help
?
This is insane. GPT is so powerful. Now we will see many people on mountains or scenic locations in insta😁
We have had this for a loooong time now as open source and free. Thats how all the AI influencers are done. Midjourney just keeps lagging behind, but it finally got ipAdapter that was released 8 months ago.
seriously annoying that its not explained for those that dont have access to the beta site!
if someone could tell me how to get on this beta site he is using that would be noice!
I saw something about have 10,000 generations..but i could be wrong. I dont have it either...I dont think many do yet.
What a scam. Bought the pro version and it is absolutely impossible to make a version of your self. Have tried all gpts and ai bots on earth for prompts and it WILL NOT work.
I can not get this to work at all. It does not recreate my face.
Thanks for this! Very helpful, trying it now . . .
Is there a way to do this where we can have different clothes / scenes? E.g., looking like a doctor in scrubs or looking like a runner in workout clothes, that kind of thing. Also, I see you have multiple GPTs -- should we use the consistent character one? Or the one with the most uses?
Great video! I find Midjourney still has trouble with me, a female; it gets some parts right but the images still aren't "me". MJ has a tendency to make women unrealistically beautiful imo. It does my husband really well, though.
That does make sense, I may have to experiment a little for the best ways to counteract that. Glad you liked the video, and good luck on your generations!
Try to use lower -s value. Less then default 100!
Hey! Thank for the video but this just doesn't work for a lot of people. You somehow got lucky with the type of face you have. Tried various versions, different set of photos, prompts but it doesn't even generate close enough images. Bad luck.
Hands down the best explanation on youtube!
Great video. I tried this but the output was a character that had similar features but looked nothing like the uploaded character references. I set the --cw parameter set to 100. Where am i going wrong?
great stuff man, you remind me of ray romano
i am not sure if its because im Chinese and that MJ isnt trained enough on how chinese people look like , but i followed the exact steps and the gens does not look like me at all 😝
Good afternoon, tomorrow May 5 in Spain is Mother's Day, with the previous version of MID JOURNEY on Discord I knew how to fit your face on another character but now I use the alpha version, could you help me to remember my mother and give her a new character.
In short, how can you include a face in a character, a portrait, in the alpha version?
Thank you very much in advance
Wonderful video! Thank you. And I'm a big fan of your GPT.
(assume we take off the "glibatree prompt" at the end, as well?)
Thanks so much! Glad you enjoyed!
And to your edit, yeah totally welcome to take the glibatree prompt off, but I have never had it hurt a prompt. I just like to see them on the popular feed haha :)
How important are the 5 images on a white background? I just tried various selfies. Why do you think those 5 images are important??
what if i have already an image that I like and I need to swap the face on it. for instance, I want to add my own face to a beauty model photo I generated in mid-journey. Also how to do it with an image of multiple faces?
I followed the same steps exactly but with sorry its not give me result of me
I can't seem to add images I don't have that option in my "home " screen....
This is great but how’s it different than taking photos with green screen and using backdrops from around the world. The face would be real.
Poses, expressions, and lighting match the dynamics of the scene with little to no effort on my part. I find this a very impressive feature, that I think many people have been waiting for
Have you ever figured out how to use to ref characters in same image?
Yeah, I got a video for you th-cam.com/video/2GDjxYwXorQ/w-d-xo.html
@@glibatree Thanks man, appreciate it!
somehow that doesn´t work with my face - it always generates characters a bit similar to me, but never, really never, I got results like you did in the video. Also someone from J said, sref doesn´t work with own images? it´s really weird, but i never got any character looking like me, no matter how many images I tried :D
Same! I cant get it to work for the life of me, maybe some humans are easier to create since their look is more common? It keep making a weird old ugly version of me lol.
I'm the same soooo frustrating!! I always look awful and I hate the pics..lol!
@@Los_Leto same for me! I'm starting to wondr the same.
@@samferriere5730 I haven’t tried the multi angle photos with same shirt and white background tho. I will try that when I have time.
Was hoping it could just take a slew of random pictures of me, but may need more concise references.
I have written a very interesting book and I want to publish this book through midjourney or other platforms with video pictures and audio. Can you do this for me? I’m ready to pay or you can suggest me who can do this for me.
does this only work in midjourney alpha?
Can we make the character face have deferent emotions?
Thank you, but what is the solution for those who don’t have access to the midjourney website. Discord only works with an url but my images on my computer have no url😅
So its possible to make pics of our selfs? I thought in first release that would not be possible, as they were afraid people would misuse this feature.
Yeah I am surprised to, but I found it was very effective at creating me - and it seems like other people are having similar results. Doesn't work for everyone - but I'm loving it...
awesome! thank you
Is this possible to do if you don't have access to the alpha and are using the discord bot?
How can we add our own captured image as the background?
Exciting
Great work. Why do all the Chat responses have your name in them? Does that help the prompt?
I'm kinda playing a long game here, but both DALLE and Midjourney are models that get regularly retrained. I am dreaming that, because I really trust that artworks that come out of my GPT, the Glibatree tag will become a useful filter that improves the quality to the end result.
I also like seeing "Glibatree prompt" show on Midjourney's front page or in random blogs and news articles hahaha
Thanks man. Glibatree is my absolute favorite. After generating four images, I turn around and put the best one in my GPT called the Canvas Critic. After a lengthy critique, it regenerates the image and I’d say 60% of the time, the image comes out even better, frosting on my cake! Thanks for updating yours consistently.
meh looks nothing like me
Im having trouble doing this on mobile without access to fo the Beta website. But i watched the video while driving, gonna have to rewatch. 🎉🎉🎉 AWESOME
Yeah, try not to drive a vehicle while watching also watching a video. Even if you're ok with the risk, the rest of us don't feel the same way.
@@dhickma I was listening not watching, that how I missed the instructions. It's hands free on Bluetooth
insane
Hey man, what happened to your GPT?
Thanks alot❤
Glad you enjoyed!
Isn't working on the app or beta.
all your pictures, look like comedian Jeremy Hotz.....lol
Is there a limit to how many reference images you can use?
You can only select up to 20 image prompts at a time. That can be a mix of character references, style references, and regular image prompts, or all of any of those.
In discord there is a separate character limit, I think 4000.. so it mainly depends on the length of your links.
🎉Well it really depends about your face. You're actually quite lucky that it works with yours. Most of the time it fails.
I'd say it depends way more on the prompt matching the face itself. Inconsistencies between the prompt and the characters (or lack of description) makes Midjourney really suffer
makes me look so ugly ahahaha
👋
GREAT VIDEO worked for me BUT...Midjounrey sucks at doing people of darker skin (AFRICAN) never on point
oh and I see mid journey alpha is not available for everyone so how you are doing things is not of much use to us under 1k image producers...
yall talk way to fast
What if you change the TH-cam playback speed to 0.5?
so fake
too complicated with all this gbt stuff and zillions of comands, why is nothing ever simple, i dont have a year to waste