Don't Sleep On This Sneaky Midjourney Update
Midjourney snuck in this much-requested feature! Now you're able to assign custom weights to image prompts featuring multiple pictures ... in this video I'll show you some common use-cases and the ultimate way to create unique images
► ARTICLE:
www.futuretechpilot.com/midjourney-tutorials/custom-image-weights
► NEW FREE Prompt Pack
✅ ftp.live/free-prompt-pack-1
► FREE PDFS
✅ Prompt Ad-Libs + 20 Category Key Words: ftp.live/ad-libs
✅ Camera Angles + Lenses: ftp.live/camera-lenses
✅ 200 Aesthetics: ftp.live/200Styles
✅ Midjourney Cheat Sheet: ftp.live/CheatSheet
► PROMPT PACKS
✅ Photorealism Prompt Pack: ftp.live/photoreal-pack
✅ Midjourney Variety Pack #1: ftp.live/variety-pack
✅ A.I Art Ad-Libs: ftp.live/ad-libs-premium
► MIDJOURNEY WRITTEN TUTORIALS
ftp.live/mjtutorials
► LET'S CONNECT
Discord - / discord
Instagram: / futuretechpilot
Пікірлер: 126
Been making vids with MJ for 6 months. You will never get exactly what you want. But you can settle for something close enough. Love using MJ!
@FutureTechPilot
11 ай бұрын
Yah hopefully we get more control in the future!
Great job as always. I love how honest and clear you are about Midjouney capabilities. Great communication skills as well, zero hype just enthusiasm and a desire to be understood. Keep it up.
@FutureTechPilot
11 ай бұрын
I appreciate the kind words and the feedback! Really helps me understand what works for you all. Thanks!
Finally! Two weeks later than I needed it, but still nice they added it.
@FutureTechPilot
11 ай бұрын
haha I know what you mean
no freakin way!!!! You are a legend!!!!!!
@FutureTechPilot
11 ай бұрын
Cheers pal!
Fantastic, thnx!
@FutureTechPilot
11 ай бұрын
Cheers!
Great tips. Thanks man.
@FutureTechPilot
11 ай бұрын
Cheers!
Love your content my man.
@FutureTechPilot
11 ай бұрын
Cheers buddy I'm glad I can help out
You're amazing !!! thank you so much
@FutureTechPilot
11 ай бұрын
And thanks for the comment!
Ha, valuable information. Thanks for sharing!
@FutureTechPilot
11 ай бұрын
Cheers pal!
You have enhanced my prompting. Thank you so much.
@FutureTechPilot
11 ай бұрын
I'm happy to hear that, thanks for the feedback!
Thank you dude for pointing this out! Highly useful videos as always! 😁
@FutureTechPilot
11 ай бұрын
Happy to help!
Interesting updates, thank you!
@FutureTechPilot
11 ай бұрын
Cheers pal!
Thanks FTP! I missed this whole update. I'm crap at math so many thanks for your weight experimentation!
@FutureTechPilot
11 ай бұрын
haha no problem pal, thanks for the comment!
Absolutly genius and helpful. Thank you.
@FutureTechPilot
10 ай бұрын
Hey I'm happy to help!
Your channel is amazing ❤ thank you for sharing 😊
@FutureTechPilot
11 ай бұрын
Happy to help, thanks for the support!
Thanks for sharing as always, the creative juices are flowing :D
@FutureTechPilot
11 ай бұрын
haha I can imagine this leading to even more infinite possibilities lol
it's better for a general guide of things in the image for sure, but I very much appreciate this video, I didn't know that!
@FutureTechPilot
11 ай бұрын
Yeah it will come in handy from time to time!
10:12 Love these! The lightning is awesome
@FutureTechPilot
11 ай бұрын
Yeah lol I've always liked MJ's interpretation of lightbulbs
👋 Finally! I've long awaited this!
@FutureTechPilot
11 ай бұрын
hahah I know what you mean!
Incredible !!! thank you
@FutureTechPilot
11 ай бұрын
Enjoy!
Wow , it's a great, thanks so much ❤
@FutureTechPilot
11 ай бұрын
I'm glad to hear it!
I had this problem in V4 I kept getting my raccoon mechanic holding a spanner/wrench because my image prompt was such high res of the object & my character a smaller res I kept getting giant tools in his hands … by using a smaller res image the tool got smaller, and/or I first put ‘a very small spanner in hand’ (I had named the image ‘spanner’ an old v4 trick. So to avoid giant necklace use a smaller res or prompt you require a smaller , thinner necklace.
@FutureTechPilot
11 ай бұрын
Wow, I never thought something like that would effect the output. That's fascinating, thanks for sharing!
We’re working on a spoofy movie trailer using Midjourney for the images and this is helpful. It’s been very difficult to get the look we want. We’ll try this. Thank you 🌟🌝
@FutureTechPilot
11 ай бұрын
haha that sounds cool. Good luck!
@AiYouTubeCoach
11 ай бұрын
@@FutureTechPilot thanks so much!
Oh that's just sneaky of them! Good job coming across it and thank you _so_ much for sharing this. 👍 💯 Slot machine - no kidding! 😅
@FutureTechPilot
11 ай бұрын
haha that's probably why it's so addicting, the slot machine vibe is real
00:00 🎉 Midjourney added the feature to assign different weights to image prompts, allowing more control over the text-to-image generation process. 00:25 📷 Previously, adjusting image weights was done using "--IW," but now you can mix and match images with individual weights using "::number" format. 00:40 🔄 The weight relationship matters; if image A has weight 4 and image B has weight 2, it's the same as 2 and 1, maintaining the ratio. 01:10 💡 Experiment with combining accessories and characters by uploading images together, simplifying link gathering. 02:47 🔗 The new feature is great for guiding the generation of images, resembling blend modes. 05:19 🏞 Adding descriptive text to prompts can enhance the generation of specific images, providing better results. 06:02 🎭 Combining real characters with cartoon backgrounds or vice versa can yield unique and blended images. 08:09 🖼 Combining real characters with real backgrounds using different weights can lead to creative and blended results. 10:18 🎨 Adding chaos and stylized values to prompts can result in even more diverse and unique images. 11:54 🎲 Combining multiple images with individual weights and adjusting values offers highly unique and creative image results. 12:52 🔄 Chat GPT can be used to generate different weight values for images, leading to diverse and personalized image outputs.
@FutureTechPilot
11 ай бұрын
Impressive!
@creativeprofit
11 ай бұрын
@@FutureTechPilot Absolutely blown away by this update! Midjourney keeps surprising us with powerful features. Your detailed breakdown of how to use the new image weight feature is incredibly helpful. The way you explain everything with clarity and enthusiasm makes diving into these techniques so exciting. Can't wait to experiment and see what unique images I can create using this. Keep up the fantastic work! 🙌
did not realise they'd added this! Thanks B)
@FutureTechPilot
11 ай бұрын
I don't think they ever announced it haha a sneaky change for sure
7:24 "grizzled" .........I don't think I've ever used that term in MJ....... I will try it now!
@AG_before
11 ай бұрын
😅 I was thinking the same thing when I heard it!
@FutureTechPilot
11 ай бұрын
🤣 always makes me wonder what other words I'm missing out on
I quite liked the video! Btw, what does unsplash mean?
6:46 MJ is capable of such amazing background imagery as well as architectural creativity!
@FutureTechPilot
11 ай бұрын
When the focal length is centered on a character and the background is blurred - has always blown my mind
@LouisGedo
11 ай бұрын
@FutureTechPilot Do you use SDXL installed locally? It's been many months since I played with my local install of Stable Diffusion. I intend to get some time in with SDXL over the next few weeks. SD 2.1 was a disaster from what I recall......when that model was released is around the time I stopped playing with Stable Diffusion.
@FutureTechPilot
11 ай бұрын
@@LouisGedo Nope! It looks amazing, but I don't have the time haha maybe I'll get around to it one day
I wonder if photobashing the items onto the main image first would make them come out clearer and cleaner when blended with the other images later, I bet even if done messily it would come out stronger.
@FutureTechPilot
11 ай бұрын
that's a good idea!!
Thanks again for another fantastic video! The combination of midjourney and ChatGPT to randomize the weights could very well cost me my entire fast time minutes…lol. I do have one quick question though. I noticed that you added RAW to the text of the image prompt. Is there a benefit to doing that as opposed to just the raw setting? 😎✌️
@FutureTechPilot
11 ай бұрын
Oh, I just use it sometimes when I'm going for a realistic subject on a simple background! I'm not even sure how much it really helps with that lol just something I've tried. Definitely not a replacement for --style raw
Love the term Slot Machines! Sometime you will get a perfect picture with one minor detail that will mess up the entire picture and then you get lost in variations wasting credits or precious time!
@FutureTechPilot
11 ай бұрын
hahaha it's definitely a similar 'buzz' to gambling
Hi, thanks for this video. Where did you find this information? I didn't see any announcement on the official Midjourney Discord🤔
@FutureTechPilot
11 ай бұрын
There wasn't one. It was extra sneaky. I saw some people discuss it in a chat and I asked about it. My understanding is that it was meant to be implemented a while ago, and when they found out it wasn't working, they turned it on. I guess they didn't think of announcing it because they thought they already did? I'm just talking out loud lol I don't know what really happened
May be “ a queen with a SMALL necklace….” This might work better 🙂⭐️
@FutureTechPilot
11 ай бұрын
haha yeah you're right!
Best Trick: generate a image with --s 0. After that vary (subtile) with a --s 1000 (remix mode on). With this strategy u get the result what u need.😉
@FutureTechPilot
11 ай бұрын
really? that sounds awesome, thanks a lot for sharing the tip! If I ever mention it in a video, I'll give you a shout out. Thanks again
@motivationzumhoren2725
11 ай бұрын
@@FutureTechPilot it always depends on what you want to generate. a lower s value gives you what you want. but with few details. it is therefore worth increasing the s afterwards. For example, full-body shot. Thank u 😊
(3:44) "Now both of those are going to blend together and be three times as important as the subject." Technically that is incorrect. The "original image" constitutes 1/7th of all "weight" in a 1/3/3 system while the other two "combine" to form the other 6/7ths of all "weight". That means they as a group are now *six* times more "important" than the original subject even though they are equal in importance relative to each other. Hence, the original vanished. If you actually wanted the two jewelry shots to be collectively "three times as important", your numbers would be 2, 3, and 3.
@FutureTechPilot
10 ай бұрын
Thank you! I think I'm used to including only 2 values in a prompt. I was definitely incorrect, so thanks for the correction!
@WifeWantsAWizard
10 ай бұрын
@@FutureTechPilot Sure thing.
This is what I meant by going deeper, lol. Do you have to have the space before the double colons?
@FutureTechPilot
11 ай бұрын
haha I had a feeling you were referring to something like this, but I didn't want to get your hopes up lol and you don't need a space for a regular prompt, like dog::5 is fine, but you need a space after a link or MJ will think the ::5 is a part of the link and it won't load properly
i just adding weights to images (great advice) but it gave me "invalid request" messages when adding chaos or any other parameter
@FutureTechPilot
11 ай бұрын
hmm, could you add text in the prompt?
"MidJourney is a slot machine" -truth! I would like to request more of your take on "describe". I use it constantly and discover a lot of artists from the results, but am I wrong in saying that if you want to create anything remotely like the image you are having described it is completely useless? I will re-roll it multiple times and sometimes never get it to even mention the main subject in the image! it rarely mentions anything its prompts about a specific action. its fun to experiment with by I would never call it a useful tool. Thanks for all your great content. -peace :)
@FutureTechPilot
11 ай бұрын
Q: "am I wrong in saying that if you want to create anything remotely like the image you are having described it is completely useless?" A: Nope! Not wrong at all, there is definitely a disconnect between the name of the feature and the capabilities. I think it stems from 2 things. First, it's a separate bot that describes the image - (like, the bot that draws is not the same as the /describe) - and second, it's probably really hard to linguistically describe a visual experience. I'm not sure we have the words to make it accurate enough. However, you can check out this video here kzread.info/dash/bejne/ZnqFyc2yfarKhrw.html where I go over what you can do with the /describe prompts to make them more accurate. Also, I have an extra super secret master tip. Lower the stylize value in the describe prompt. Like maybe --s 30 or even --s 60 and it could work better. Hope that helps a little!
@christopherbigelow6688
11 ай бұрын
@@FutureTechPilot thanks I will go back to that one. As usual you nailed what I’ve been feeling- I’ve heard others say describe is a way to learn how to prompt, to which I would say “haha- no.” FTP keeping it 100!
I have found that ai creates super interesting characters. I have a basic idea ai has thousands.
@FutureTechPilot
8 ай бұрын
Yeah exactly, ai is a monster at iteration
Is there any indication how MJ will interpret weights to the text prompt relative to the image weights?
@FutureTechPilot
11 ай бұрын
I DEFINITELY should have mentioned it in the video, and I apologize for that. So, the weight of something that you don't specify is always a 1. Like 'link1 ::2 link2 ::.5 happy dog' is actually 'link1 ::2 link2 ::.5 happy dog ::1' Hope that makes a little bit of sense
@AnthonySell
11 ай бұрын
I guess the question I have is how are the text prompt weights weighted against the image prompt weights? And do we still use --iw 2 for example, or will this new image-specific weighting throw everything we've become accustomed to out the window?@@FutureTechPilot
I can't find "copy image adress" on my iMac.🥴 Suggestion?
@FutureTechPilot
11 ай бұрын
May I ask what you see instead?
Thanks for sharing your skills, but you don't show really once you copy (2 Min 45 sec) the 3 links and how it's appaeras on discord?
@ericpeltier6771
11 ай бұрын
The 3 "https" are undivided on discord, can you help, thanks for your concern
@FutureTechPilot
11 ай бұрын
Yeah you're right, I should have done a live example. I'm sorry. Here is a better explanation When you click 'copy image address' and paste them into your prompt, it will look like this: /imagine prompt: media.discordapp.net/attachments/995015556770050100/1138156843097653348/futuretechpilot_a_black_and_orange_bird_sitting_on_fire_in_the__c56fe2a5-b6cd-4cba-8673-d005080c2e86.png?width=829&height=1178 ::1.8 media.discordapp.net/attachments/995015556770050100/1138156975167901876/futuretechpilot_a_girl_in_green_and_pink_with_headphones_on_her_7c24e3af-2ac2-46cf-8f1a-b835dde30cdb.png?width=829&height=1178 ::1.3 but Midjourney shortens them when you generate it: s.mj.run/tdkKxpRTJ5E ::1.8 s.mj.run/xh-6ieukuu8 ::1.3 I hope that clears up your confusion. Let me know if I misunderstood
Let midjourney update and make full on videos
@FutureTechPilot
11 ай бұрын
what do ya mean
cool but why write the ratios in decimal points, that isn't for e everyone. for demonstration purposes it would be easier to stick to 2 , 4, 8?
@FutureTechPilot
11 ай бұрын
Maybe! But I like the smaller numbers and tweaking it by a decimal point makes it easier to find the 'sweet spot' in my opinion
Is there any contact information for business cooperation?
@FutureTechPilot
11 ай бұрын
email futuretechpilot@gmail.com
Is there a maximum number of images that MJ will recognize or be able to successfully discern and utilize in an image?
@WonderfulWorldOfStuff
11 ай бұрын
Sounds like a experiment waiting to happen, I know 16+ images prompt is where it breaks MJ & errors. I’m guessing 16 images with weights maybe possible but I can’t imagine the sliding scale of weights to control such a mass amount, as you saw in the video with a too heavy weight on the person the necklace image was ignored, so once you start using 4-5 images it’s going to get tricky.
@LouisGedo
11 ай бұрын
@WonderfulWorldOfStuff Have you seriously even come close to using 15 images in which you can reliably say that all those images had a definitive influence on the MJ render? I get to 4 or 5 images and MJ seems to struggle to incorporate all the images.
@WonderfulWorldOfStuff
11 ай бұрын
@@LouisGedo no I haven’t, I asked David what was the maximum images you could use, he said he didn’t know and that I was welcome to test to see if it breaks. 16 images accepted - 17 images error everytime. I was simply mentioning that the max images was 16 but to utilise these with weights would be madness 😎👍
@WonderfulWorldOfStuff
11 ай бұрын
I use around 5 images when creating a character sheet, face of person for example at 5 different angles, I normally sensibly use 2 for blends & 3 if I want scenery/character/character 😎👍
@LouisGedo
11 ай бұрын
@@WonderfulWorldOfStuff 👋 So let me see if I understand your point, It's not necessarily the case that the MJ bot can incorporate as much as 16 images (or elements or influence of as many as 16 images) into a single generated image but rather that based on your testing one can merely add 16 images to a prompt before the bot goes into error? If true, that's nice to know but it's not useful information to me since I only really care about the maximum number of images that the MJ bot can effectively use to influence images it generates. And yes, I know it can be difficult to test this maximum threshold of *influencing input images* , but if you ever do come up with a number that appears to be consistently at the maximum threshold, please let us know. I know there's others who have also wondered about that max number of *INFLUENCING* images that the MJ bot can interpolate and use to generate a single image. 👍
👍👍👍💅
@FutureTechPilot
11 ай бұрын
🤝
Why are MJ results so weird since July?
@FutureTechPilot
11 ай бұрын
The algorithm changes but if you like the old versions, you can find them in the /settings !
I really don't understand the logic of using decimal numbers given the explained ratio. Why just not use, 1, 2, 4, etc... less characters and faster to interpretate.
@FutureTechPilot
11 ай бұрын
Yeah to each their own! I think it's easier to interpret for me personally. 1.5 vs .5 is 3x but if I bump it to 1.5 vs .7 I'm getting closer to 2x For some reason, I like that workflow more
So it still can't do product photography? What happened to the AI is going to steal jobs ? lol
@FutureTechPilot
11 ай бұрын
you'd have a way easier time photoshopping it
@codeXenigma
11 ай бұрын
@@FutureTechPilot Or just taking the photograph haha I'm mocking clickbait media that make claims AI will replace all jobs.
Or just get good at prompting and then you don't need any images.
@FutureTechPilot
11 ай бұрын
😶
midjourney also stopped swimwear and bikinis, so all female models will be dressed like they are from the 60's
@FutureTechPilot
11 ай бұрын
lol really?
@mac8463
11 ай бұрын
yes, try it out, all images have to be pg12 smh@@FutureTechPilot