BIG Midjourney Update ~ IN-PAINTING Full Breakdown

[@dannyellis971]

If the in painting didn’t work you can use the “::” slider

[@FutureTechPilot]

hmm, interesting. Can you give me an example? How would you use it?

[@Airbender131090]

Finally. SD cult will not be able to say “there will never be inpainting in MJ!” Now we need controlnet and mj is full package

[@puja1985]

Leonardo ai already have control net

[@tristanmoreau1956]

the full package would be :free and uncensored

[@Airbender131090]

@@puja1985 Leonardo ai is stable diffusion, isnt it?.

[@Airbender131090]

@@tristanmoreau1956 yeah shure. Sd is free. You just need to buy 1500$ pc and pay for windows and huge electricity bills, but shuuuure its freeee!!! You can use MJ for 5-6 years with that money. And noone cares about making naked women and porn.

[@FutureTechPilot]

makes me wonder what the next big 'breakthrough' will be

[@AG_before]

Thank you! 👍 The fact that they managed it _in_ Discord, is pretty amazing! Otherwise, it's a cute little child learning to walk. We must help!

[@FutureTechPilot]

hahaha yes I'm still amazed that Discord has to create features just to help Midjourney. But it seems be worth it! And it works on mobile, like that is pretty magical

[@AMPKlickitat]

"laughing hysterically" -- MJ is already overpowered with respect to emotion, and with "laughing hysterically" that dialed it up way too far. Dial it down with just "laughing", or better yet "laughing::0.5" . . . remember you can weight your prompts in remix / vary operations.

[@FutureTechPilot]

That's a really good point!!

[@JuneStarling]

Wow, these updates look cool. Saw something similar in CGDream recently. I love using these both tools!

[@FutureTechPilot]

haha yes this technology is amazing

[@LouisGedo]

MJ vary region actually sucks big time. I'm comparing this to tools that actually work well like Adobe Generative Fill. They actually should NOT have released it so early before getting a better performing tool.

[@FutureTechPilot]

😂 Photoshops gen fill is legit the Heavyweight Champ right now

[@Riverfox2012]

Tried it "going by the directions," which is 1/3 to 1/2 of the image to be redone. Changed clothing or scenery and it works okay. Not outstanding on every image, but okay. Sunglasses are too small so I expect that'll be the thing they're ironing out.

[@FutureTechPilot]

That's a good point! I guess the smaller things will be more difficult

[@BxBNFT]

As I was watching your video I realized I could go to old images (I was all the way back at some of my v4 image blends)... and, In addition to your ✉️ trick (thanks!) You can also: Click the U or V buttons (assuming you haven't already upscaled the image) or The 'beta / light upscale redo' buttons, ...and the Vary(Region) button appears in the message / new image MJ sends you! 📥👨‍🎨 Just wanted to share! thanks @Futuretechpilot

[@FutureTechPilot]

Oh okay! Very cool. Thanks!

[@jpjacobsenbr]

Brazilian follower here. Watching this video I couldn't stop thinking: This doesn't make easier nor faster than generating image in MJ then doing last touch ups like sunglasses, background and jewerly within Photoshop Generative Ai which is in Beta. Thoughts on this? Thanks for sharing knowledge.

[@FutureTechPilot]

You're completely right! In fact, I think Photoshops Gen Fill is much better than Midjourney haha but I think there are lots of people who don't have Photoshop, and I think this is a great option for them!

[@jpjacobsenbr]

Thanks for the honesty. I have both MJ and PS. I will stay on my workflow for now but will keep tuned in for how MJ develops on this front. It all happens so fast in Generative AI, it completly changes the audiovisual industry. @@FutureTechPilot

[@LouisGedo]

15:24 Have you tried to prompt in the car using --style raw ........I wonder if that could make a difference in terms of coherence with the background image?

[@FutureTechPilot]

dang, good idea!!! I'll write that down now and give it a shot

[@stefantewes314]

I guess, MJ tries to fulfill your prompt with the whole image. You get sunglasses because there weren't sunglasses already and you only erased a specific area to paint these. As for the cat image: you asked for a cute cat, you got a cute cat - the one that was already there. Should have asked for two cats ;)

[@FutureTechPilot]

hahah that's a really good point! Maybe if I said '2 cats' it would have worked

[@aronno0abir]

any way to use this feature in already existed image ? I mean can I took any photo from my gallery, put it in midjourney & inpaint over it ? ( just how photoshop AI work )

[@FutureTechPilot]

It's a good question but I don't think it's possible, at least not at the moment

[@GloriaPM]

I did manage to get a cool text of the word I wanted, most of generations give you something similar to the word, I chose the one that started with the same letter of the word I wanted and then I used the new vary feature letter by letter until I got it, it is doable BUT it takes way too many rerolls, I wouldn´t suggest to spend fast hours on this. Anyway, Thanks a lot for the video and all the tips!! Keep it up!

[@FutureTechPilot]

You're exactly right! And I think I'll be making a video about that for this Monday

[@SHADRAGZ]

Wat should I do if I just to want to polish my artwork whithout changing the design? ... I do monsters design tho

[@FutureTechPilot]

hmm I'm not too sure Midjourney can help with that! Photoshop is probably a better solution

[@SaidThaher]

Damt .. they just s*cking the credits like black hole 😂😂😂😂😂

[@FutureTechPilot]

hahah way too many things to experiment with. It's exhausting

[@dustinkoop]

Amazing!!! I am trying to change a face and am being told that I need to select a larger area. Any hacks fo fix?

[@FutureTechPilot]

YES, I just learned that you can select OUTSIDE the image. That should solve the problem

[@pedrodeelizalde7812]

Thanks for your video! Question, can we import an image from our pc an edit it with inpainting?

[@FutureTechPilot]

No I don't think that's possible. But maybe in the future!

[@KingZero69]

amazing! really excited for this update to FINALLY come to midjourney 💪😎

[@FutureTechPilot]

apparently this has been ready for a while and they were just waiting for Discord to be updated!

[@ganon4]

The strange thing is I don't have the input to write text, I can only select and submit.

[@FutureTechPilot]

make sure 'remix' is turned on. You can do find it in /settings or type /remix into the prompt box

[@jasonp9508]

Figured out you can right click a region, and it removes it from the selection!

[@FutureTechPilot]

Awesome tip! Thank you

[@boheem3451]

I was changing some people on a beach to a sea diver. I then changed the sea diver. It totally got rid of the diver and I have a blank beach. A bit like magic erase.

[@FutureTechPilot]

I haven't tried to remove anything yet but I'll definitely have to experiment soon!

[@Juan2142]

I just tested it on only 3 images, and got the result I wanted straight away that you had demonstrated in your first image showing the character laughing by simply adding an image weight, kept my prompt the same mostly but updated it with changes I wanted corresponding to the area that I had selected and boom, used --iw 2. Not sure if this works on every image, will be testing this all day.

[@FutureTechPilot]

woah, just adding an image weight to the prompt. That's a really good idea!! Thanks for the tip, I'll have to try that out now

[@Juan2142]

I've tested it now on about 30 images, it's inconsistent in the results, so definitely more to explore in the new cool feature, I'll let you know if I find anything that comes up.

[@bobbyw1074]

Can you do this with a photo I have that is not generated with midjourney?

[@FutureTechPilot]

that's a good question! And no, I don't believe so.

[@theparkinglotcities]

Thanks for the video! The release notes advise against getting too weird with your inpainting prompts - keeping things within the realm of the original prompt - but I've gotten some cool surreal images by doing the opposite. I've got a lot of images of strange alien landscapes. Just selecting half the image, and asking for a photo of a modern cityscape result in something cool. The results can be a bit mixed, rarely an actual photo, but I love it. Specifying a film stock can also help. I've had the best luck with Cinestill 50D, but haven't experimented beyond that.

[@FutureTechPilot]

That's a really cool tip! Thanks, I'm writing that one down now haha

[@LouisGedo]

15:54 That's actually quite interesting that in some images, the bot was able to seemingly interpolate the bricks correctly into the in painted region. Fascinating

[@FutureTechPilot]

Exactly! Fascinating is the perfect word lol

[@paullenoue8173]

Biggest tip/trick for me in this video? How to get a "fresh" image with all the zoom options. THANK YOU!

[@FutureTechPilot]

I'm glad I could help out!

[@DarrenHuckey]

I have remix on, but I don't have the Vary (Region) option. Are they releasing it to small groups at a time?

[@DarrenHuckey]

Never mind… I had to restart Discord. Yay!

[@FutureTechPilot]

Glad it worked out!

[@LoneIrbis]

Nice! I can recall quite a few examples when this thing could've saved me an ontherwise perfect image. Let's say, when character's face is fine, but some bits of clothings looking weird or out of intended style, or maybe there's a clearly electricity-based light on the wall when it's supposed to be medieval age, etcetra. Some of those I had to try to save in photoshop, but it still looked bad because that would take much better photoshop skills.

[@FutureTechPilot]

Yes you're right! I completely agree, sooo many pictures now have the ability to reach their full potential. The rabbit holes never end haha

[@artanon]

Thanks for a more-than-expected run through. I prompted for an airplane to be added to an empty sky and got 3 empty skies and one with a hot air balloon. Sometimes MJ is an obstinate genius with a contrarian sense of humor.

[@FutureTechPilot]

hahah it can definitely be frustrating at times

[@vernardfields7044]

I figured out a way to add specific characters into the in-painting option! 1. Generate your background 2. Generate your character and copy the image link 3. Go back to your enlarged background image and perform the in-painting function 4. Make your area selection 5. Enter the general prompt order (simple description of character, "paste image link", character action) and hit enter *it may take a couple times for the enter button to work.

[@siffertlouis535]

Hey ! I tried it several times without any results :/ any tips on how you made it work ?

[@FutureTechPilot]

Thanks for the breakdown! I haven't had much luck with that but I'll try it again.

[@LouisGedo]

14:51 Ooo-la-la ........... I really like that car render! Please offer some of the prompt recipe for it 🙏

[@FutureTechPilot]

hahah futuristic sportscar, unsplash --s 400 .... unsplash is a god

[@LouisGedo]

@@FutureTechPilot LOL!! I can't believe it is such a simple prompt with such fabulous results!

[@musigamer4723]

why can't I type what I want in the in painting.

[@FutureTechPilot]

make sure 'remix' is turned on ... you can find it in /settings or type /remix into the prompt box

[@Just-a-guy926]

Can you use an image prompt with this

[@FutureTechPilot]

Yeah! But it's kinda weird. Not sure the best way to do it yet

[@rezasalamati9652]

Great video. Thanks

[@FutureTechPilot]

🤝

[@skywalker5709]

what a great features finally!!

[@FutureTechPilot]

Very cool indeed!

[@ElevenEvilExes]

btw, the right mouse button deletes an area you have selected.

[@FutureTechPilot]

thanks for the tip!

[@ElevenEvilExes]

@@FutureTechPilot you're welcome!

[@SceneArtisan]

For showing images on a regular screen (computer or small device) MJ's outputs are fine, but for printing they're woefully inadequate; having to use AI enlargement is pitifully fake; no one can persuade me to think anything to the contrary. In-painting looks great (I have yet to use it) but I'll resubscribe when native (not upscaled) outputs have had their resolution increased. Until then, I'll wait, in-painting can wait too.

[@FutureTechPilot]

I think resolution seems to be a problem for every ai generator!

[@jaycobyart]

I was finally able to get some coherent text by selecting the pieces I wanted to change and changing it letter by letter.. prompting "the letter 'x'" and over time was able to slowly change out the letters and get a decent looking image. Thanks for sharingm I'm enjoying the Inpainting feature, probably too much .. be well

[@FutureTechPilot]

Yes exactly! Going one letter at a time seems to be the answer

[@puja1985]

Oh wow, inpainting in Midjourney! Thank you for putting so much effort into this full breakdown of the new feature. I felt bad after unsubscribing to Midjourney a couple of weeks ago and switching to Leonardo AI. I guess I'll be using both of them in the meantime until I decide which one I should focus on. I'm amazed at how much attention Midjourney gets from KZbinrs who make tutorials about it whenever a new feature is released. I don't see that on other platforms, at least not as quickly as Midjourney gets reviewed. For example, when the photoreal feature was released in Leonardo AI, not many people were talking about it. But whenever Midjourney releases something new, there are at least 5 or more active KZbinrs discussing and making tutorials about it, including you. At 8:30 , you mentioned that an image might not appear as you imagined it in Vary Region. One possible solution is to try expanding the region by using the zoom tool or panning feature and then doing Vary Region again. See if the intended subject appears. Again, I can't test this myself since I don't have access to Midjourney anymore. hehe At 10:55 , i've noticed some flaws with the panning feature since it was released. The image can be distorted or the blending from the initial image can be lacking. I haven't seen this problem as much with the zoom feature, although the zoom feature does have its own drawback of reducing the resolution size. At 11:15 , the custom zoom button trick is amazing for bringing back the disappeared Vary Region button feature after panning an image. Thanks for bringing this up! At 13:50 , I noticed that you lassoed the Joker and Batman together in Vary Region at the same time. If you lassoed them separately and then generated the region one by one, they might not have come out in mixed costumes. Right now, it looks like Batman is wearing the Joker's suit and vice versa. At 15:23 , you mentioned that the car placement might not be perfect, but the lighting is quite accurate. I can imagine what we could produce with a combination of Midjourney and Photoshop, using photobashing and photomanipulation techniques. With a good eye and creativity, we could probably create masterpieces. I hope that in the future, Midjourney will add something like ControlNet, a consistent image-to-image mode where we can add AI subjects regionally to an uploaded photo of our own. In order to do this, I think they would need to have their own web-based platform like Leonardo ai, as adding more complex tools to Discord might be challenging. Fingers crossed for Midjourney v7! If it happens, I think it's going to blow our minds even more. Wow, what a journey!

[@FutureTechPilot]

Thanks for sharing your thoughts and enthusiasm! I'm really glad other people find this tech as interesting as I do haha and I hope Leonardo/Stable Diffusion keep pushing Midjourney to improve and offer more features. Something like ControlNet would be soo cool

[@ChurchofCthulhu]

Great vid! Thanks! ❤

[@FutureTechPilot]

🤝 Cheers!

[@davidmichaelcomfort]

Nice video! You can do face swaps really well using in-painting. Simply use the rectangular select tool to select face and then use a combination of image prompt and character description to generate a new face. I wrote up a Medium post about it "Creating Consistent Characters using Midjourney’s new In-Painting Tool." Also, I experimented with adding characters to an "empty" scene and didn't get good results. I think it is better to swap out a character rather than placing character in an empty scene

[@FutureTechPilot]

Cool tip on the character face! And yeah I think swapping anything works 10x better than trying to 'insert' something into an image

[@Ton369]

wow finally cant wait for them to ditch Discord. Why do they even use Discord as a software platform at this point?

[@FutureTechPilot]

They're working on their own website but I guess Discord is their best option at the moment

[@SavageBro.]

Thanks for making this video I’m really good at in painting now! 👍🏻

[@FutureTechPilot]

I'm happy I could help!

[@Oxes]

Its not a specific inpainting like stable diff, but this is pretty damn cool!

[@FutureTechPilot]

Yeah exactly!

[@rickswizzle4824]

✌️

[@FutureTechPilot]

🤝

[@nomorejustice]

Woyoooo, this is game changer!!! Thanks for fast updates! 🙏

[@FutureTechPilot]

Game changer for sure

[@ElysterStaux]

Loved the video. The good bye saying "peace" in the ending cracked me up tho.

[@FutureTechPilot]

😂

[@channelnotfound403]

Tip: Select everything but the persons face, then change the prompt to a new location. "women in garden scene, looking at camera", then change to "women at the beach, looking at camera". Now it will keep that persons face, just re-create the scene/body of the person. Downside is they will have the same facial expression of course.

[@FutureTechPilot]

That sounds really cool! I'll write that one down and try it out soon. Thanks!

[@channelnotfound403]

@@FutureTechPilot I also found it easier to do the selection in 2 halves. Where you select the Left half minus their face, then the Right half minus the face, and finally any extra bits that were missed. It will be nice when we can get an Inverse Selection Tool to make this process a little easier; as well as a Zoom feature to get it a bit more precise.

[@renaca01]

The image of the car was too big in relation to the road image. Try a smaller car or smaller image. It matters.

[@FutureTechPilot]

like the size of the file? would I have to put it in photoshop, make it transparent, and shrink it?

[@renaca01]

@@FutureTechPilot Reducing the size of the car is the goal. If you reduce the file size might be enough but I found better results not having a car that fills the entire image. Trial and error.

[@chrisr.5730]

during making images yesterday, this option was coming like magic. i try it and it works really good for my generations.

[@FutureTechPilot]

haha I'm happy to hear that!

[@trevorsoh2130]

Thanks for the great, Detailed content! You are awesome!

[@FutureTechPilot]

I appreciate the kind words!

[@LouisGedo]

👋

[@FutureTechPilot]

🤝

[@jenerocity]

Thanks!

[@FutureTechPilot]

Thank you!

[@eurossocial]

Wow! this was the best demonstration on the new inpainting feature in MJ. thank you for sharing this. By the way some people say that if you can not see the new button, you should close and restart the discord, and it will show up. About the cat images, have you checked your images in the Midjourney website gallery? maybe the effects have been applied in there?

[@FutureTechPilot]

Yes good idea to mention restarting Discord. And no I never checked my gallery, guess I couldn't imagine it being any different haha but I'll take a look and let you know if I find anything cool

[@eurossocial]

@@FutureTechPilot yes sure!

[@crykz9055]

how do I get the update, I don't have it in my discord yet?

[@FutureTechPilot]

I've seen people say to restart Discord. I don't think you need to sign out. Just close the app/website and reopen it. Hope that helps

[@jasonp9508]

I had to update discord to get it. Restarting discord does the same thing.

[@crykz9055]

ty @@FutureTechPilot

[@crykz9055]

ty aswell@@jasonp9508

[@thisisyoutubehandle]

08:44 it did work. it changed that rag or what ever that thing is on the right bottom corner.

[@FutureTechPilot]

😂 I guess when you're looking for a second cat, the smaller changes are hard to notice haha

[@jimjohngirard]

Thanks, I had no idea....great...loved it!

[@FutureTechPilot]

Happy to help!

[@CaroleeHeynen]

What does "unsplash" mean?

[@FutureTechPilot]

It's a stock photography website that is a really powerful word in MJ. And it allows you to create a consistent cinematic look without the need for a bunch of extra words in your prompt

[@CaroleeHeynen]

@@FutureTechPilot I am such a newbie! I would love to hear more. You mean you put the word "unsplash" in the prompt as some kind of acknowledgement? Or does MidJourney somehow reach into Unsplash and winnow through the images in some magical way?

[@FutureTechPilot]

@@CaroleeHeynen using a word will tell Midjourney to harness what it has learned from that word. In the case of Unsplash, it will reference everything it has learned from that word, which includes an entire website of similar looking images. It's definitely magic, but it's the same thing as if you use an Artist in your prompt. Midjourney will create an image based on everything it has learned from that artist

[@tomi71]

Great! Thanks for the nice video? I wonder if we can have character tilt/rotate their head with this or change style e.g realistic photo -> cartoon Does this help with character consistency?

[@tomi71]

Oops the ? was supposed to be ! (questionmark typo hit there...)

[@FutureTechPilot]

hmmm, thanks for sharing your thoughts! I'm wondering a lot about character consistency ... I'll have to experiment with that. As far as tilting their head, I'm really not sure if that would work! Please let me know if you try it out and come to any conclusions. Same with the real->cartoon effect

[@tomi71]

@@FutureTechPilot looking forward if you test those ideas. I cant right now since i have not purchased it at the moment.

[@WalterSoares]

Muito Bom. Obrigado por compartilhar!

[@FutureTechPilot]

🤝

[@Juan2142]

The God's of AI have blessed us 🤣 Thanks for the video bro!

[@FutureTechPilot]

enjoy the blessings! hahaha

[@Juan2142]

@@FutureTechPilot 🤣🤣

[@EconomicWarfare]

This is so cool. I can't wait to play with it.

[@FutureTechPilot]

😂 let me know if you find any cool tips

[@louie8362]

Can you tell the devs to let us upload an image of a character and have inpaint solve the character consistency problem, and without needing to select the area, I know of another way to do it, but its 50 times a day limit, many thanks 👍👍👍

[@andytangaming2705]

mind sharing the "other way" method :)?

[@FutureTechPilot]

haha I don't know any of the devs but I know they're working on character consistency stuff

[@zofo264]

What does unsplash do?

[@FutureTechPilot]

gives it that 'vibe' you see. That look can be accomplished with other words, like mentioning certain cameras and lighting. But I find 'unsplash' gets the job done so efficiently with just the one word

[@zofo264]

@@FutureTechPilot Gotcha - thx!

[@hldfgjsjbd]

To like a lot something that plain and out-of-place looking is to have low standards

[@FutureTechPilot]

technically Midjourney is still in Beta. All of this progress is fascinating and should be applauded