Unfortunately Midjourney doesn't currently support consistent backgrounds. If you try some of the techniques I outline in this video you'll often find the characters will appear in similar backgrounds if you use the same prompt but there are always slight variations. Depending on the complexity of the background you want to use, you could either try creating the background separately and then placing the characters into the foreground via a photo editing tool (e.g. photoshop or photopea), or you could use Vary Region to adjust the elements in the background that don't match up. Consistent scenes across images may be implemented when they update to Midjourney V7. I expect that will roll out later in the year but there is no timeline for when that will occur. V6.1 is due to roll out before then but we don't know when that will occur either.

Yes, she's definitely an overachiever :)

Good to hear you got access. Yes, it sometimes takes a lot of extra work to get good consistency with the characters. If the characters look fine but they have the wrong outfit, or something has been added that shouldn't be there, you could try the Generative Fill option or Remove Objects option under Frame Edit within a specific shot. The other option is to create images in another program, like Midjourney, which is better at creating consistent characters and import them into your shots before animating them. I'm hoping they will make a lot of improvements with the platform while it's still in beta.

@@AIChallengeQuest Ok thanks a lot

Found it!

I hope so too, but it looks like it could be some way off - after V7 is released at least.

You're welcome!

They're only letting in a select number of people at a time to ensure that they don't overwhelm their servers and to test how well the servers cope with the number of people that are using their platform. It's been less than 3 weeks since they opened the platform and there have been people who signed up onto the waitlist well before then who haven't managed to get in yet. The best option is to sign up to their Discord server (the link is in the description of this video) and set up notifications for any messages on their #announcements channel and keep an eye out for beta giveaways. They usually let in 100 to 200 people at a time and those invites can go pretty quickly - usually within 15-20 minutes. If you check the announcements channel you'll see when they previously gave away invites according to your local timezone. They also let in people from the waitlist - I don't know how many people are on the waitlist but it could be a long time before they get to you.

Glad to hear it!

This is a great question. The splitscreen allows you to change your character to follow your new prompt in a new setting (e.g. sitting in a landrover) whilst keeping consistency of the character and their clothing. If you didn't use the splitscreen it would use the setting they're already in to paint out the rest of the image. It uses the context of the area that surrounds the area you shaded to decide what to insert into that space (whilst taking into account your updated prompt). This is probably a little more obvious when you shift them to a completely different setting (e.g. coffee shop instead of the moon). Also, by inserting the URL of the original image into your prompt it will insert your character into that new setting. If you didn't do that then you would get the new setting but the character would not appear in that scene. In addition, by keeping one half of your original splitscreen it will keep using that character instead of inserting a new character into the other half. You can think of it as keeping the character's DNA intact when you make changes to your prompt. When you've completed the changes you can then paint over your old image. If you want to make more changes based off that same character and clothing you can go back to the splitscreen and repeat the process. As long as you start with a good splitscreen which shows the same character(s) that look the same in each side of the image then you should be fine.

Thanks!

I used that in my prompts to stop Midjourney from producing darkened edges around the image. It's not necessary if you're going to replace the background with another setting for your character though.

Thanks and appreciate the awesome tutorial! @@AIChallengeQuest

Thanks!

At about the 2 min mark in this video I cover the process of painting over a couple of those images and then I change the prompt. At this point you would modify the prompt to: "Jessica on a bus reading a book". If she doesn't appear in the new image you will need to add the starter image in again at the start of the prompt. I cover this in more detail in my latest video at the 3m 20s mark. Open the starter image (with 3 different poses), open that image in a browser and copy the URL, then paste it into the start of your prompt (after you've painted over a couple of the poses) and then modify the prompt with "Jessica on a bus reading a book". It will update with a new image of Jessica reading a book on a bus. I also cover another method of changing the background, in my article on Medium (there's a link to it in this video's description) that you could try.

It can vary - sometimes you'll get two new images and sometimes you'll get a merged image. The main thing is that the new images should be consistent with the original image. You can then select one of the new images and extend it by painting over the other images (make sure to overlap the body/face of the character you've chosen). If you change the prompt to "blank white background" it will paint over the other images with an extension of your chosen image. You can then change the size of the image using aspect ratio or zoom out. Hope that helps and thanks for your feedback.

@@AIChallengeQuest thank you

Hello, the website keeps directing me to discord, how do you do all of these things on discord, or how do you get to the website?

@@pattyzajac4737 Currently the website is only open to Midjourney users who have created at least 1000 images - they will open it up to everyone progressively over time. In the meantime, you can do everything in Discord. Have a look at the article I linked in the description of this video, I show you how to use Discord to do the same things.

Everything you see here can also be done directly in Discord. If you check my article on Medium (link in the description) it might be easier to follow. Thanks for the feedback.

What is creation? The alpha website?

Yes, the alpha website is currently open to anyone who has created 1000+ images on Midjourney. It's an alternative to using Discord to create and view your Midjourney images. They're working on making it more user-friendly for new users to Midjourney before they open it to everyone.

@@AIChallengeQuest aaaaaaaand Boom! Today Midjourney came out with the Character reference feature (--cref) and we're good to go! I just tested this and it works well enough :)

Yes I've been checking it out too. It will be interesting to see how well it handles different scenarios. It doesn't look like it can handle multiple characters at once though. I think that will come in time.

I think it depends whether you're trying to achieve an exact likeness of that person. If you're just trying to create a character that resembles that person then you could use the photo as a reference image. In that case you could upload the image and use that in your prompt.

You're welcome!

Thanks for subbing! There's more coming soon.

It's more difficult to get consistent results when you have two characters, but you can try a prompt like this: "3 different poses of a couple: Jessica and Bob --ar 3:1 --style raw" However, Midjourney seems to resist when you add extra elements into your prompt after using Vary (Region) and painting out a couple of those images. Updating your prompt to: "Jessica and Bob together in various poses --ar 3:1 --style raw" seems to be fine, but if you change the prompt to say that they're standing in front of a vintage car it will often show only the vintage car on its own. Either that or the couple looks a bit distorted. It may take a lot of rerolling to get what you want. When Midjourney introduces their consistent character feature this may become a lot easier.

@@AIChallengeQuestTook some playing around. Got the best results by creating two separate characters with /prefer option set. Then creating a random image with two people. Pick the best of that lot and Vary (Region) their face with the Lasso Tool, and assigning the first character. Then using that produced image, Vary (Region) and using the Lasso tool to select the second face, then assigning the second character. A bit of work, but the results came out well enough. Thanks for the response and video!

Nice work!

Hi @trafferz. I've loaded a new video that covers my method for achieving multiple consistent characters. Hope it helps.

@@AIChallengeQuestexcellent. trying now