i knew aboitu this but now i see why this is the best bet it is because of a defect in almost all modelsL: concept and consistency scapaing or soemthing like that. or invading conssidency. i dont recall. is when oyu have many faces of different peopel in one , this ahppens with flux, even when yo change one single face, the model mixes up the faces and can only process one. flux sometimes can process two but is gonna mix them somehow. this is a defect workign in favor of us. the model of the face filters out and invades the others grids so the seed and the actuaql model of the face itself keep consentent. if you happen to have extended a photo like a 2:! ratio, you will see sometimes it creates 2 personas. and in each varation those people which are the same person may have different estyles or gestures in the bdody but the face is gonna be totally syncrhonized in expression, even if they are taken from different perspectives. so this is good. sd1.5 in favor of us. mj has sd in its inner core so...
The method of creating a specific appearance is a very interesting, but also very complex method. many iterations. if there are no other alternatives, you can use it. But in general, I would like a shorter method, which I'm sure Midjourney developers will release soon. thanks for your hard work!
such a great idea. this is definitely a professional level work flow. character consistency has always been a pain. i will definitely try out this method. thank you for sharing.
Once again an incredible video! I've been using your previous workflow on consistent characters in Midjourney plus all the amazing tips I got from the live stream you did with Tao Prompts. Really changed the game for me on consistent characters in Midjourney. I love this new method you share in this video. Only issue I have with it, is it is incredobly time consuming. Not a problem for a short 5-minute film or trailer but anything longer would 5-10X the time required to produce these incredible results. I'm working on a 30-minute short film (with 5 main characters and 13 supporting characters) and while I'd love to use this method, I can't really justify the time requirement so I'll be sticking with your previous method using cref, character sheets and the tips you and Tao offered up. Thanks again for presenting the cutting edge of character and setting consistency. If you have any suggestions for me on the 30-minute task at hand, I'd love to hear them. Thanks again for everything you do for the AI and Midjoureny community!
Thanks so much, I appreciate the kind words. Also I really understand how meeting that balance is important. To my knowledge, you're on exactly the right path... My GPT, and creating a cref from a grid, is the easiest most reliable method to get fairly close rn. I'll continue to be on the lookout for faster easier ways though, that are less of a sacrifice on details. This technology keeps improving, so it may only be a matter of time!
Here’s my current workflow: I create my character normally. Then, I upscale the picture and drag it back to the mid-journey stage, requesting the character sheet. Next, I download all the pictures that resemble the character, crop the faces, upscale them, and bring them back. Finally, I press the character icon on all of them, and I’m able to get a character that looks extremely similar to the original.
Hey, that's a great technique. Thanks for showing. Hopefully, we'll see some updated character features that make creating consistent characters easier. I would like to see a character design feature that basically makes a 3D design with different layers. Like a basic body (skin) layer, then maybe a layer for hairstyle and attributes like tattoos/ jewelry etc. Then the last layer of clothes and maybe (hand held) props. I'm curious what the upcoming MJ storytelling feature will bring us.
The patience and time prompter needed to create a proper AI image is amazing. And you have to go through internally and then to client... It's so time consuming that I doubt if it's really worth actually.
want to say thank you for the tidbit of info... i went to try this to see used a female vampire in a bar ...i would say following what you described in vid, i got 70 percent of it different alterations but same character and mostly same outfit in the primary picture i used. will see more of this results
Hopefully, Midjourney will implement something like this internally to make *cref* more accurate. I think we should share this video with the Midjourney team.
okay this is huge! Thank you...would you consider doing video with animal. Generate storybook animal, and try to consistently recreate in different poses? Thanks
What a brilliant way to faceswap without any lora training, thanks! Can you make more videos about this please because it's not 100% clear for me, I'm not native English
Yeah there was a moment at 10:39 in the video where I erased that little bit of the bottom- which I later explain to take care not to do... There wouldn't have been a seam if I didn't do that in the first place.
I've been doing something similar replicating tiger stripe tattoos but MAAAAAN if i want the pose to change, MJ still REALLY struggles and its a lot of in-painting and remixing, I'm going to try a few slight tweaks based on your tips here. But yeah... getting MJ consistent on details is a struggle.
Definitely takes a lot of patience and finesse. For my goals this has been great - but I can imagine things like tiger stripes would be challenging. Good luck!
You've got the perfect face for a 19th century nobleman, who would be considered incredibly beautiful by standards of the time. Soulful eyes, genuine smile. Good job, mate 🤘😃
Thanks. Very nice. The problem is the new features are only available annual subscribers of Midjourney or those who have paid monthly for a year. Maybe it will roll out for everyone soon.
How do you get around the Edit tool's compression artifacts that plague the image as you continue to make edits or build out the surrounding scene? (using V6.1 for reference)
Your stuff is amazing, thank you so much. It really is very compelling and useful. When I add a character reference image created in Midjourney it makes the new images look they were drawn instead of photographed, so it significantly changes the texture of the original image. Even if the prompt stays the same word for word. How can I get around that.
Hey I love this! However, I simply cannot get the editor to actually do the same thing you are, with the same prompt and settings. I always just end up with the same character facing the same way, partially out of the frame, or really zoomed in. Is there any thing else you're doing behind the scenes to actually get the head to turn?
Excellent once again. You need to look out for black friday sale on freepik. 50 bucks for the year. You get flux, mystic, upscaler. Far superior inpainting. Paint to image. Re-imagine . Id be lost without mj & freepik. They compliment each other. Way less censorship as well. I find the new editor in mj as censorship on steriods, couldnt even do a blood splatter scene ??? Insane.
Thanks for the kind words! Hard to say exactly how long, cause I took a break partway through and that approach was my second try. Probably not much more than 45 minutes for the original. And maybe another 10-20 minutes per matching version. Glad you enjoyed!
Great video. Is there a particular Midjourney membership level you need to do the photo editor? I believe I was getting a “you must upgrade” the other day and I have the $30 membership.
I also have the 30 dollar membership have it, but when the editor launched they said: 1. You need to have generated 10,000 2. Been subscribed for the entire past year 3. Subscribed to a plan where you pay by the year. Not sure if things changed
I'm having trouble creating a consistent environment, like a room, office, bar, or café, with the same characters appearing from different camera angles. Could you provide a step-by-step tutorial on how to achieve this?
While MJ can read the transparency of the pshop file, you still much make at least one edit with the Erase tool in MJ or it will refuse to edit the image
@Glibatree I really can't figure out a step by step procedure from what you say in the video. The results ARE extremely good, so obviously your method works, but... can you please distill it to a step by step process? Thanks!
The part with the owl girl character can be done with the normal editor, it's when you get to complex characters like the angel, with a glowing halo, whering a key necklace, where you need the uploaded image editor to get that much detail consistent
It's available for any of the annual plans at the moment, or anyone who has been subscribed for at least a year, or anyone who has made 10,000 images. So it's limited at the moment, but not as bad as only for $100 a month. I imagine there will be more access in the coming weeks.
The new uploaded image version of this editor definitely makes this technique far more powerful and let's you do more- I hope everyone gets access soon. You can still apply the method to your creations, like I showed for the woman with owl makeup, but it will have more limitations. Hopefully it can help your generations even now.
Jesus, this is a lot of steps to get MJ to do something. I’ve been using MJ for a couple of years now and I’m starting to hit the wall with frustration due to the lack of UI and actual controls. It’s getting better but talk about jumping through hoops.
Brilliant. This is FAR better. You should send this to midjourney and maybe the can leverage this technique internally.
Haha, yes! That'd be amazing!!! 😍
@@glibatree Please check your email I have something important to discuss with you @glibbatree
They will in version 7
I thought the exact same thing!
i knew aboitu this but now i see why this is the best bet it is because of a defect in almost all modelsL: concept and consistency scapaing or soemthing like that. or invading conssidency. i dont recall. is when oyu have many faces of different peopel in one , this ahppens with flux, even when yo change one single face, the model mixes up the faces and can only process one. flux sometimes can process two but is gonna mix them somehow. this is a defect workign in favor of us. the model of the face filters out and invades the others grids so the seed and the actuaql model of the face itself keep consentent. if you happen to have extended a photo like a 2:! ratio, you will see sometimes it creates 2 personas. and in each varation those people which are the same person may have different estyles or gestures in the bdody but the face is gonna be totally syncrhonized in expression, even if they are taken from different perspectives. so this is good. sd1.5 in favor of us. mj has sd in its inner core so...
The value you bring is amazing
Thanks for that, I appreciate you watching!
Amazing workaround and problem solving. I will be testing this out and will make sure to credit your tutorial!!
Amazing!!! You are so patient and creative to have thought up this method!
Glad to see so many people excited by it, I've really enjoyed playing around with it!
The method of creating a specific appearance is a very interesting, but also very complex method. many iterations. if there are no other alternatives, you can use it. But in general, I would like a shorter method, which I'm sure Midjourney developers will release soon. thanks for your hard work!
I second that!
Can’t wait for this release 😂
Cool idea! Definitely not the easiest to control but with enough attempts it seems to work! Good stuff
@FutureTechPilot can you extract a step-by-step method from this video?
@@MrJoe1199 thought about it for sure, but ultimately I don't think so! Maybe I'll try again later though
@@FutureTechPilot I watched it several times. I will iterate again and again until I can get it down to a procedure.
@@MrJoe1199 let me know if you find some good strategies!
such a great idea. this is definitely a professional level work flow. character consistency has always been a pain. i will definitely try out this method. thank you for sharing.
This is so clever. Big brain stuff at play here. Thank you!
You are the best brother.. You should in Midjourney team.. Love the way you describe everything.. Thanku so much.. Fan of yours.. ❤
The answer was there the whole time for me.... I needed this for a book I am writing then take the images into video generation models. Thank you!
Brilliant method - you always lead the way!
Nailed it! This is advanced research, or a crazy lucky shot! Wither way, it's grrrrreat!
Cool! What a nice discovery. Thanks for share!
You bet! Hope you're able to apply it. Thanks for watching!
That’s so great, thank you so much for such a useful video!
Once again an incredible video! I've been using your previous workflow on consistent characters in Midjourney plus all the amazing tips I got from the live stream you did with Tao Prompts. Really changed the game for me on consistent characters in Midjourney. I love this new method you share in this video. Only issue I have with it, is it is incredobly time consuming. Not a problem for a short 5-minute film or trailer but anything longer would 5-10X the time required to produce these incredible results. I'm working on a 30-minute short film (with 5 main characters and 13 supporting characters) and while I'd love to use this method, I can't really justify the time requirement so I'll be sticking with your previous method using cref, character sheets and the tips you and Tao offered up. Thanks again for presenting the cutting edge of character and setting consistency. If you have any suggestions for me on the 30-minute task at hand, I'd love to hear them. Thanks again for everything you do for the AI and Midjoureny community!
Thanks so much, I appreciate the kind words. Also I really understand how meeting that balance is important.
To my knowledge, you're on exactly the right path... My GPT, and creating a cref from a grid, is the easiest most reliable method to get fairly close rn.
I'll continue to be on the lookout for faster easier ways though, that are less of a sacrifice on details. This technology keeps improving, so it may only be a matter of time!
I rarely post a reaction but I really wanna thank you for this video! I needed this!! Thank you 🎉
your videos are so underrated! keep it up
That's so nice thank you!
Amazing! Your content is incredible, thank you!
Here’s my current workflow:
I create my character normally.
Then, I upscale the picture and drag it back to the mid-journey stage, requesting the character sheet.
Next, I download all the pictures that resemble the character, crop the faces, upscale them, and bring them back. Finally, I press the character icon on all of them, and I’m able to get a character that looks extremely similar to the original.
how do you word your prompt?
This is awesome, thanks for the tip! It may help so much with using the renders in Runway for differents shots. Will try that!
Yeah this is the best method I've run into by far for that sort of thing.
Great idea for when you have only one or a few photos of the person you want to put into your MJ image
Another excellent tutorial. You rock! Thanks so much!!
Amaaazing. Thank you for sharing this!
Omg. I appreciate this one! 🎉 Amazing.
Hey, that's a great technique. Thanks for showing. Hopefully, we'll see some updated character features that make creating consistent characters easier. I would like to see a character design feature that basically makes a 3D design with different layers. Like a basic body (skin) layer, then maybe a layer for hairstyle and attributes like tattoos/ jewelry etc. Then the last layer of clothes and maybe (hand held) props. I'm curious what the upcoming MJ storytelling feature will bring us.
That would be cool! Really excited about what's to come!
Seriously best video on this ever amazing you found a hack around and crazy good results 👌👌👌
Great tip about adding transparency from a Photoshop file! I'll definitely be giving this a go.
Thanks, good luck!
ANY transparency will do. You can create them with inkscape with $0 cost. Or even capcut.
Holy crap! Cannot wait to try this! Thank you!
To align perfect, just shift paste-in-place it back and then resize both layers down together
nice. just a tip. paste your final image into photoshop and then scale both layers together to make it easier to match.
The patience and time prompter needed to create a proper AI image is amazing. And you have to go through internally and then to client... It's so time consuming that I doubt if it's really worth actually.
Well fucking done!! That was amazing! So good!
want to say thank you for the tidbit of info... i went to try this to see used a female vampire in a bar ...i would say following what you described in vid, i got 70 percent of it different alterations but same character and mostly same outfit in the primary picture i used. will see more of this results
Seems like you're getting the hang of it! Not a catch all solution, so it makes sense that your images would still need some finessing... Good luck!
Wow, thank you SO much for this!
You're very welcome!
this is a really good technique
This is great info! Thanks for the video - will def. help in projects moving forward 🤙✌
Great to hear! Good luck on the projects.
I cannot wait until I unlock that darn 10k generation editor... I love MJ, but man that is a lot of generations lol
Great video, thank you! It was very helpful.
Wow! Love it!! Thank you
Amazing thank you for the great share
Simply brilliant, thank you!
You're very welcome! Thanks for watching!
Thanks Bro, It looks like I've come to the same way of drawing myself and characters more similar and realistic 👍
Awesome, good luck
Hopefully, Midjourney will implement something like this internally to make *cref* more accurate. I think we should share this video with the Midjourney team.
okay this is huge! Thank you...would you consider doing video with animal. Generate storybook animal, and try to consistently recreate in different poses? Thanks
You are smart. Love it
wow that is so nice thank you very much for sharing !
What a brilliant way to faceswap without any lora training, thanks! Can you make more videos about this please because it's not 100% clear for me, I'm not native English
This is amazing! Can I do this in discord without the editor?
huge work, thanks man
Genius! Thank you !!!
Midjourney is also consistent with your expensive yearly subscriptions!
Great method! Is there a way to apply the brush strokes to your face so that the style and detail level is consistent across the whole image?
You can clean up the neck by erasing the layer so we dont see the seam :)
Yeah there was a moment at 10:39 in the video where I erased that little bit of the bottom- which I later explain to take care not to do...
There wouldn't have been a seam if I didn't do that in the first place.
Some really good stuff.
Great tutorial, thnks !
Amazing! How did you find out about this method?
thanks for the video, anyway for consistent backgrounds ?
I've been doing something similar replicating tiger stripe tattoos but MAAAAAN if i want the pose to change, MJ still REALLY struggles and its a lot of in-painting and remixing, I'm going to try a few slight tweaks based on your tips here. But yeah... getting MJ consistent on details is a struggle.
Definitely takes a lot of patience and finesse. For my goals this has been great - but I can imagine things like tiger stripes would be challenging. Good luck!
Cool! Does this require adding "--cref" to the prompt in the Editor?TIA!
Bravo! super helpful, thank you!
Good luck creating!
You've got the perfect face for a 19th century nobleman, who would be considered incredibly beautiful by standards of the time. Soulful eyes, genuine smile. Good job, mate 🤘😃
That's very flattering, thank you for the kind words!
Dude. I love you
This is really cool.
Great idea to solve the problem! Is there a similar option in discord?
Thanks. Very nice. The problem is the new features are only available annual subscribers of Midjourney or those who have paid monthly for a year. Maybe it will roll out for everyone soon.
Agreed! The other option is having generated 10,000 images which is a tall order of the other two aren't true. I bet it will be relatively soon.
At 5:19 in the video, I'm not sure how you copied and pasted the image that previously had only the necklace removed, can you show me?
😅
my sincere wow. thanks
How do you get around the Edit tool's compression artifacts that plague the image as you continue to make edits or build out the surrounding scene? (using V6.1 for reference)
Your stuff is amazing, thank you so much. It really is very compelling and useful. When I add a character reference image created in Midjourney it makes the new images look they were drawn instead of photographed, so it significantly changes the texture of the original image. Even if the prompt stays the same word for word. How can I get around that.
Unbeatable ~~!!!
Hey I love this! However, I simply cannot get the editor to actually do the same thing you are, with the same prompt and settings. I always just end up with the same character facing the same way, partially out of the frame, or really zoomed in. Is there any thing else you're doing behind the scenes to actually get the head to turn?
I want to do fashion stuff, i need some help with generation.
smart. asking myself why mj doesn't do that "in the background" anyway
Brilliant!
Excellent once again.
You need to look out for black friday sale on freepik. 50 bucks for the year. You get flux, mystic, upscaler. Far superior inpainting. Paint to image. Re-imagine .
Id be lost without mj & freepik. They compliment each other.
Way less censorship as well.
I find the new editor in mj as censorship on steriods, couldnt even do a blood splatter scene ???
Insane.
Thanks man, I'll keep an eye out!
Thanks! Legend.
Thanks so much!! Roughly how long did it take you to get the Angel character finalised. Brilliant insight and tips!!
Thanks for the kind words! Hard to say exactly how long, cause I took a break partway through and that approach was my second try. Probably not much more than 45 minutes for the original. And maybe another 10-20 minutes per matching version.
Glad you enjoyed!
@ Amazing!!
Bro that tip will work out for me... i need to do an comercial of an orange kitten with a lion mane wig
Hope you're able to pull it off, seems like a cool project!
Good job
Great video. Is there a particular Midjourney membership level you need to do the photo editor? I believe I was getting a “you must upgrade” the other day and I have the $30 membership.
I also have the 30 dollar membership have it, but when the editor launched they said:
1. You need to have generated 10,000
2. Been subscribed for the entire past year
3. Subscribed to a plan where you pay by the year.
Not sure if things changed
Excellent
I'm having trouble creating a consistent environment, like a room, office, bar, or café, with the same characters appearing from different camera angles. Could you provide a step-by-step tutorial on how to achieve this?
@regumrai did you figure this out? This is where I am stuck too.
@@erinchantz me too. I hope Midjourney will update this feature in the next version
wow.... good shit
Does this work for cartoon and animation styles
Thank you! :)
GENIUS.
How do I download those damn edited images? It's driving me crazy!
this video is beyond useful but us that dont have PS and not good at it :( dead end
While MJ can read the transparency of the pshop file, you still much make at least one edit with the Erase tool in MJ or it will refuse to edit the image
True, but you can just click the eraser around in a transparent part - so it can be a pretty quick fix. Just a quality of life bug I think
@Glibatree I really can't figure out a step by step procedure from what you say in the video. The results ARE extremely good, so obviously your method works, but... can you please distill it to a step by step process?
Thanks!
I have a question so we can’t do this method in normal editor ? Please someone answer 😢
The part with the owl girl character can be done with the normal editor, it's when you get to complex characters like the angel, with a glowing halo, whering a key necklace, where you need the uploaded image editor to get that much detail consistent
it does appear far better, but just so that we are clear, the editor is only available in the most expensive version right ?
It's available for any of the annual plans at the moment, or anyone who has been subscribed for at least a year, or anyone who has made 10,000 images. So it's limited at the moment, but not as bad as only for $100 a month. I imagine there will be more access in the coming weeks.
Hi! How did you put 2 pictures on one working field? I can't do it.
I had to pull both images into an image editor (in my case Photoshop) export it, and then upload that result to work with in Midjourney
genius
Nice 🚀
Thanks 🔥
Doesn't work for me :/
But ou realize not everyone can access this feature in midjourney yet. You have to have 10,000 images or a year subscription
The new uploaded image version of this editor definitely makes this technique far more powerful and let's you do more- I hope everyone gets access soon.
You can still apply the method to your creations, like I showed for the woman with owl makeup, but it will have more limitations. Hopefully it can help your generations even now.
@glibatree Thank you
Wow wow
👋 hi
Hi there! Haha
Great cref have been a let down. Also using serval reference images.
I hope this works better for you! Good luck!
Jesus, this is a lot of steps to get MJ to do something. I’ve been using MJ for a couple of years now and I’m starting to hit the wall with frustration due to the lack of UI and actual controls. It’s getting better but talk about jumping through hoops.