🎉Thank you to my Patrons who voted for this video! 🌟Join the Woollyfern Patreon (free & paid tiers) ➡www.patreon.com/Woollyfern 🌟Midjourney Guides: woollyfern.gumroad.com ➡many are free and paid ones are included with some Patreon tiers CREF Deep Dive: th-cam.com/video/BOgqKKRT7IE/w-d-xo.html Midjourney's Image Editor: th-cam.com/video/n4apjbyNfGE/w-d-xo.html How To Make Images on Midjourney's Website: th-cam.com/video/RYcGacwrnPA/w-d-xo.html
Thank you! Very informative. I've been struggling with Midjourey on using consistent characters for my cinematic trailer and this definitely gives me a lot to work with.
As always with you, Woolyfern, a very useful guide, delivered with your usual clarity. I’ve not seen this technique suggested by any of the other online ‘experts’. Well done for figuring it out.
I do prefer 6.0 over 6.1 because of the difference in detail and how each model generates photographic-style images of people. 6.1 tends to produce less detailed, more simplistic images (on average). That's great for when I want a nice, clean illustration-style image, but not so great for when I want a realistic-looking person. People in 6.1 are a bit too "beautified". So yes, I still gravitate toward 6.0, but 6.1 has its place too 🙂.
No problem! Poses and object holding can be difficult. Outside of trial and error via revising your prompt text, there are a few other things you can try. 1) If the character's pose is correct, but object holding needs modified, try the Basic Image Editor. You can erase over an area, modify the prompt and re-run it to replace the erased area. Basic Image Editor: th-cam.com/video/n4apjbyNfGE/w-d-xo.html 2) Try an image prompt (image + text) where the image shows a pose you are trying to achieve. Image prompts covered here: th-cam.com/video/Kyr-nWtp1z8/w-d-xo.html Example pose reference here: th-cam.com/video/BOgqKKRT7IE/w-d-xo.htmlsi=-kT2-BhHXbVTfA0C&t=548 3) If you have a reference image that shows the pose you want, you can try using Midjourney's Edit & Retexture features in their Full Image Editor to swap in your character and change the look of the image. The Full Image Editor does have limited access, but take a look at this example to see what I mean. Full Image Editor: th-cam.com/video/qUw85p1hyhM/w-d-xo.htmlsi=BL_mH8aMluMN6Ont&t=598
Thanks! Check out this video: th-cam.com/video/jeamaku6TpA/w-d-xo.html. I have a section showing how to get the image URL of an image you created in or uploaded to Discord and the URL of an image already in your Midjourney gallery. In the specific example I showed at 06:11 - that is the URL of an image I created in Discord.
Thanks for your insight. I have a question. Are you able to use consistant background setting in midjourney? Let's say most of the scenes are happening in the 1 same spot. How can you generate such an image for multiple characters with various sizes, doing all different poses in that same background?
There is not a consistent background feature. The best hack for this is to first create your background image without any characters. Then, to put a character into the image, open your background in the Image Editor (on the website) and Erase where you want the character (you can also use Vary Region on Discord to do this). Then update the prompt text, add your CREF, and submit the job. Midjourney will only update the area of the image that you erased. You can make your final result even cleaner by taking the two images into Photoshop or Canva, cut out the character from one and place it into the same spot in the original background image. It's not the smoothest workflow and will probably take several tries, but it is a way to get something similar to what you want.
thank you dear, but my generated face not like as my photos. I use 10 different ways but i cant. Only Aı generated photos works well, i think. Because they are trained same style fashion photos.
Yes, Midjourney's CREF feature works best when the Character Reference image you use is an image that was created by Midjourney. All of the Character Reference images that I used in the video were created with Midjourney. Using non-Midjourney images as Character References is a bit more challenging.
One way is to try to create a Midjourney version of yourself (or the person you are trying to use as a CREF) using a combination of image prompting and detailed prompt text. Use the image of yourself in the image prompt. Describe yourself in detail. Once you get a satisfactory Midjourney version of yourself, try using that as the CREF. You can also try using multiple image URLs after the --cref parameter, just separate each URL with a space. Midjourney's CREF works differently than the Flux models. CREF should improve after v7 is released. This is still Midjourney's first version of Character Reference (released March 2024). Image Prompting: th-cam.com/video/Kyr-nWtp1z8/w-d-xo.html
It depends what I am working on, but lately I find myself using the website more, especially if I want to use the Editor. It's a lot easier to zoom, pan, and repaint all at once instead of in multiple steps. I still do virtual upscales ("U" buttons) in Discord though, for images I like.
Midjourney will choose a style for your results if you do not give it one. If you want a photographic image, be sure to include a phrase such as "photo of" or "photo taken with Kodak film", etc. I have a few videos with example phrases for film and photography that you might find helpful here: th-cam.com/play/PL8IB1Sx8eJ1ryTuq8skzS-vWO-W5b1pnx.html&si=-Xp5B6p3u6ibOobZ
@@woollyferncreative Thank you very much for your help, I will definitely share the result with you and your followers, inşallah you do great things and your channel grows a lot.
I have a follow-up question. Does anyone know how to create an additional 5 scenes from one image that Midjourney generated for me? It’s important that it keeps the same characters it already created. I’ve tried several adjustments, but it always generates different characters. Thanks a lot for any help!
Two things that you might take a look at: 1) if you want the character AND pose to stay the same, consider using the Image Editor on the MJ website, erase around the character, and type a new prompt for the scene. 2) Use the Remix feature - on the web, open your image and click Remix Subtle or Remix Strong, update your prompt text to describe the new pose and scene (be sure to still include your CREF in the prompt), submit the job. Remix will use your image as a starting point and can be a way to make adjustments to an image that you already like. If you want to use Remix on Discord, enable Remix Mode in your /settings. Then, each time you click a "V" button, you'll get a pop-up where you can update the prompt.
This is fantastic, any chance you could create same video using cat/dog? I’m attempting to create storybook for grandchild, really struggling with consistency of the cat. Not sure where I’m going wrong, little more of a terror book than storybook. Thanks. Love your videos!!
A "terror book", oh no 😂! Unfortunately CREF doesn't work as well with non-human characters. You might have to get more descriptive with your prompt text to get the details you need. Hopefully this will improve sometime after the V7 model comes out!
Woolyfern, love your clear and concise tutorials. Was wondering if you could help an Ipad Pro user. It never seems to respond properly with my prompts especially in the editor mode. I have a smart keyboard and yet, Im unable to delete previous image refs to add another. Any suggestions would be greatly appreciated! ❤ Your channel.
Thanks! Sorry, I don't have experience using Midjourney on an iPad. I know some features don't work well (or at all) for some devices. I would check out the Help room at midjourney.com/rooms and see if someone there can provide an answer.
CREF works best with images created by Midjourney. If you use a photo of a real person, it does not work as well. When V7 comes out (soon) CREF should work much better with non-Midjourney-created images. Hope that answers your question.
If Midjourney had a decent enough prompt understanding, this would have been useful, but neither their cref nor prompt understanding is any good. The clothes always don't match. Which means you can't create art that features teams, coz everyone would be wearing something different. How do you even begin to call something consistent when it's loudly a failed attempt? And you guys keep hyping it as if you can use it any setting that demands consistency. At least they should create refence codes for characters or costumes, even it'd be an extra feature that you buy.
I agree that Midjourney's prompt understanding & CREF feature have room for improvement. CREF creates characters that are _similar_ to the reference, rather than exact duplicates. It is also not meant for duplicating exact outfits. It works pretty well with faces - although still requires a little effort which I share in the video - and can be consistent enough across images to meet many people's needs. Remember that the current CREF feature is also Midjourney's _FIRST_ attempt at CREF (March 2024), so it's only going to get better 🙂. If you need exact character and clothing duplication, Midjourney isn't the right tool yet, but these capabilities are expected to improve with the next model release (by the end of the year).
Just an FYI. I am a very recent convert to AI. Took out a months subtraction to Midjourney about two weeks ago. Today I was told I was banned. I did not know why I still don't know why. I reached out to customer services, in a nutshell. They do not discuss individual cases, the ban is permanent and there are no refunds. Just glad I did not pay for the whole year. So Caveat emptor.
🎉Thank you to my Patrons who voted for this video!
🌟Join the Woollyfern Patreon (free & paid tiers) ➡www.patreon.com/Woollyfern
🌟Midjourney Guides: woollyfern.gumroad.com ➡many are free and paid ones are included with some Patreon tiers
CREF Deep Dive: th-cam.com/video/BOgqKKRT7IE/w-d-xo.html
Midjourney's Image Editor: th-cam.com/video/n4apjbyNfGE/w-d-xo.html
How To Make Images on Midjourney's Website: th-cam.com/video/RYcGacwrnPA/w-d-xo.html
I have some projects How can i Contact you ?
@@villageengineerbd7100 contact info is on my channel page
Really useful, well explained and wonderfully edited videos! Thank you for sharing and for all your effort to make it easy and good looking 😊
Thank you!
@artePro_Oficial, I follow you, and your stuff is wonderful too!
@@silversurfer4195 Thank you!! 😊🙏
Woollyfern, I needed this video for my current project (a short film) thank you for saving me sooooooo much stress, lol
Excellent, happy to help!
Thank you! Very informative. I've been struggling with Midjourey on using consistent characters for my cinematic trailer and this definitely gives me a lot to work with.
Great, glad you found it helpful!
For months I’ve been trying to do this and finally found a video that explains so much in detail
Great! Glad you found it helpful!
As always with you, Woolyfern, a very useful guide, delivered with your usual clarity. I’ve not seen this technique suggested by any of the other online ‘experts’. Well done for figuring it out.
Thanks, glad you found it helpful!
Much needed ❤👏🏼👏🏼
Another great video. Thanks! Considering the Patreon membership
@@silversurfer4195 thank you 🙂
You are the best Woollyfern!!! thank you so much!
Thank you 🙂🙏
Great video. Very useful and easy to understand. Thank for sharing 🤩🙏
Thank you!!
Excellent Tutorial ... Thank you VERY VERY VERY MUCH !
Thanks!
Super Thanks for a very enlightening video as usual. 👏😃
Thank you!
Holy Wownesssss YES!! 🎉👏🏼💯
great video!
Thank you!
Very Informative Thanks
Thank you!
you are amazing.... ty
lisa loeb is the mid journey QUEEN… 😍
You are a life saver 🤩🤩🤩🙏🙏🙏
Can I ask why dont you use 6.1? Do you prefer 6 over 6.1 and if so, why? 😁
I do prefer 6.0 over 6.1 because of the difference in detail and how each model generates photographic-style images of people. 6.1 tends to produce less detailed, more simplistic images (on average). That's great for when I want a nice, clean illustration-style image, but not so great for when I want a realistic-looking person. People in 6.1 are a bit too "beautified". So yes, I still gravitate toward 6.0, but 6.1 has its place too 🙂.
@@woollyferncreativeInteresting! I was hesitant to go back. What you say resonates with my experience.
thanks for the tutorial! what would you do to change poses of characters and what they are holding?
No problem! Poses and object holding can be difficult. Outside of trial and error via revising your prompt text, there are a few other things you can try.
1) If the character's pose is correct, but object holding needs modified, try the Basic Image Editor. You can erase over an area, modify the prompt and re-run it to replace the erased area.
Basic Image Editor: th-cam.com/video/n4apjbyNfGE/w-d-xo.html
2) Try an image prompt (image + text) where the image shows a pose you are trying to achieve.
Image prompts covered here: th-cam.com/video/Kyr-nWtp1z8/w-d-xo.html
Example pose reference here: th-cam.com/video/BOgqKKRT7IE/w-d-xo.htmlsi=-kT2-BhHXbVTfA0C&t=548
3) If you have a reference image that shows the pose you want, you can try using Midjourney's Edit & Retexture features in their Full Image Editor to swap in your character and change the look of the image. The Full Image Editor does have limited access, but take a look at this example to see what I mean.
Full Image Editor: th-cam.com/video/qUw85p1hyhM/w-d-xo.htmlsi=BL_mH8aMluMN6Ont&t=598
@@woollyferncreative thank you! going to try this now 😀
This is superb, thank you so much. Voodoo-less MJ, who’d have thought? Well, okay, less-Voodoo…
Thank you!
Great!!! Thank you :-)
Great tutorial! I have one question: how can I get the URL you mention at 6:11? Thank you!
Thanks! Check out this video: th-cam.com/video/jeamaku6TpA/w-d-xo.html. I have a section showing how to get the image URL of an image you created in or uploaded to Discord and the URL of an image already in your Midjourney gallery. In the specific example I showed at 06:11 - that is the URL of an image I created in Discord.
@@woollyferncreative lovely, thank you very much!:)
Thanks for your insight. I have a question. Are you able to use consistant background setting in midjourney? Let's say most of the scenes are happening in the 1 same spot. How can you generate such an image for multiple characters with various sizes, doing all different poses in that same background?
There is not a consistent background feature. The best hack for this is to first create your background image without any characters. Then, to put a character into the image, open your background in the Image Editor (on the website) and Erase where you want the character (you can also use Vary Region on Discord to do this). Then update the prompt text, add your CREF, and submit the job. Midjourney will only update the area of the image that you erased. You can make your final result even cleaner by taking the two images into Photoshop or Canva, cut out the character from one and place it into the same spot in the original background image. It's not the smoothest workflow and will probably take several tries, but it is a way to get something similar to what you want.
thank you dear, but my generated face not like as my photos. I use 10 different ways but i cant. Only Aı generated photos works well, i think. Because they are trained same style fashion photos.
Yes, Midjourney's CREF feature works best when the Character Reference image you use is an image that was created by Midjourney. All of the Character Reference images that I used in the video were created with Midjourney. Using non-Midjourney images as Character References is a bit more challenging.
@@woollyferncreative what is your best way to fix this problem? I tried some flux models for train my pictures, it works well.
One way is to try to create a Midjourney version of yourself (or the person you are trying to use as a CREF) using a combination of image prompting and detailed prompt text. Use the image of yourself in the image prompt. Describe yourself in detail. Once you get a satisfactory Midjourney version of yourself, try using that as the CREF. You can also try using multiple image URLs after the --cref parameter, just separate each URL with a space.
Midjourney's CREF works differently than the Flux models. CREF should improve after v7 is released. This is still Midjourney's first version of Character Reference (released March 2024).
Image Prompting: th-cam.com/video/Kyr-nWtp1z8/w-d-xo.html
What do you work in the majority of the time, web editor or discord?
It depends what I am working on, but lately I find myself using the website more, especially if I want to use the Editor. It's a lot easier to zoom, pan, and repaint all at once instead of in multiple steps. I still do virtual upscales ("U" buttons) in Discord though, for images I like.
My reference image is a real photo, but the images I produce with that reference are always like watercolor or pastel, do you know why :(
Midjourney will choose a style for your results if you do not give it one. If you want a photographic image, be sure to include a phrase such as "photo of" or "photo taken with Kodak film", etc.
I have a few videos with example phrases for film and photography that you might find helpful here: th-cam.com/play/PL8IB1Sx8eJ1ryTuq8skzS-vWO-W5b1pnx.html&si=-Xp5B6p3u6ibOobZ
@@woollyferncreative Thank you very much for your help, I will definitely share the result with you and your followers, inşallah you do great things and your channel grows a lot.
I have a follow-up question. Does anyone know how to create an additional 5 scenes from one image that Midjourney generated for me? It’s important that it keeps the same characters it already created. I’ve tried several adjustments, but it always generates different characters. Thanks a lot for any help!
Two things that you might take a look at: 1) if you want the character AND pose to stay the same, consider using the Image Editor on the MJ website, erase around the character, and type a new prompt for the scene. 2) Use the Remix feature - on the web, open your image and click Remix Subtle or Remix Strong, update your prompt text to describe the new pose and scene (be sure to still include your CREF in the prompt), submit the job. Remix will use your image as a starting point and can be a way to make adjustments to an image that you already like. If you want to use Remix on Discord, enable Remix Mode in your /settings. Then, each time you click a "V" button, you'll get a pop-up where you can update the prompt.
How can I contact you for a potential cooperation?
There's contact info on my channel page
hey.
It doesn't work for me at all. It gives me completely different faces ;( do you know what could be the reason?
Which method are you using? Be sure to add in your new CREF image anytime you are trying to place a new face into the image.
thanks:)
This is fantastic, any chance you could create same video using cat/dog? I’m attempting to create storybook for grandchild, really struggling with consistency of the cat. Not sure where I’m going wrong, little more of a terror book than storybook. Thanks. Love your videos!!
A "terror book", oh no 😂! Unfortunately CREF doesn't work as well with non-human characters. You might have to get more descriptive with your prompt text to get the details you need. Hopefully this will improve sometime after the V7 model comes out!
Woolyfern, love your clear and concise tutorials. Was wondering if you could help an Ipad Pro user. It never seems to respond properly with my prompts especially in the editor mode. I have a smart keyboard and yet, Im unable to delete previous image refs to add another. Any suggestions would be greatly appreciated! ❤ Your channel.
Thanks! Sorry, I don't have experience using Midjourney on an iPad. I know some features don't work well (or at all) for some devices. I would check out the Help room at midjourney.com/rooms and see if someone there can provide an answer.
No idea, but i do all steps by steps, and finally Midjourney creates 4 versions of non-me pics, no similarity. what`s the reason?
CREF works best with images created by Midjourney. If you use a photo of a real person, it does not work as well. When V7 comes out (soon) CREF should work much better with non-Midjourney-created images. Hope that answers your question.
I couldn't get this work. Very frustrating.
Which part? If the --cref parameter is new to you, I recommend watching this video first th-cam.com/video/BOgqKKRT7IE/w-d-xo.html
😜😜😍😍😍😍
i am so lost....
What questions do you have? If the --cref parameter is new to you, I recommend watching this video first th-cam.com/video/BOgqKKRT7IE/w-d-xo.html
If Midjourney had a decent enough prompt understanding, this would have been useful, but neither their cref nor prompt understanding is any good. The clothes always don't match. Which means you can't create art that features teams, coz everyone would be wearing something different. How do you even begin to call something consistent when it's loudly a failed attempt? And you guys keep hyping it as if you can use it any setting that demands consistency. At least they should create refence codes for characters or costumes, even it'd be an extra feature that you buy.
I agree that Midjourney's prompt understanding & CREF feature have room for improvement. CREF creates characters that are _similar_ to the reference, rather than exact duplicates. It is also not meant for duplicating exact outfits. It works pretty well with faces - although still requires a little effort which I share in the video - and can be consistent enough across images to meet many people's needs. Remember that the current CREF feature is also Midjourney's _FIRST_ attempt at CREF (March 2024), so it's only going to get better 🙂. If you need exact character and clothing duplication, Midjourney isn't the right tool yet, but these capabilities are expected to improve with the next model release (by the end of the year).
Just an FYI. I am a very recent convert to AI. Took out a months subtraction to Midjourney about two weeks ago. Today I was told I was banned. I did not know why I still don't know why. I reached out to customer services, in a nutshell. They do not discuss individual cases, the ban is permanent and there are no refunds. Just glad I did not pay for the whole year. So Caveat emptor.
I heard using describe function or add Midjourney App channel as a discord server will get you banned automatically. Don't know why.