BEST Midjourney Feature Ever?
Вставка
- Опубліковано 28 вер 2024
- Midjourney Version 6.1 Cheat Sheet PDF (free) ➡️ ftp.live/MJChe...
-- Paid Courses --
Midjourney Beginner Course ➡️ ftp.live/mjcourse
Midjourney Advanced Color Course ➡️ ftp.live/mjcolor
-- Social --
Future Tech Newsletter ➡️ ftp.live/newsl...
Instagram ➡️ / futuretechpilot
Twitter ➡️ x.com/FutureTe...
Thanks for the explanation, I was so confused on how to use this when they announced it despite reading their instructions, this is very informative, much appreciated! 🙌🏻
Glad I could help!
It seems kinda like a lora... i think you can share the tuned style to everyone
Thanks for the heads up and demo. I like using the grid feature and picking one of the 16 styles to play around with.
The grid wasn't working for me when I first tried it, but I've seen some comments saying they like it better, so I'll try to mention that in an upcoming video!
So basically the Midjourney equivalent to LoRas? Pretty cool! Hopefully in the future they add style mixing and retraining from those. This could be a killer feature tbh.
Yeaah exactly, you see the potential!
Your final 'style' batman's, especially image 3 at 8:16 remind me of Maxfield Parrish. Interesting how this polaroid style ended up landed in that realm.
Exactly! Style tuning is so powerful for that reason, you really don't know where you will end up
cooooool !!! thanks didn't see that coming
the rabbit holes never end!
fun to play with for sure, getting more messed up hands/faces in this. I don't know if it's something with the /tune or they are getting hiccups overall...
Yeah I know what you mean, something definitely gets lost in translation
welll... I mean I DEFINITELY want a Style Library function now. right? how am I supposed to remember all of these 6 months from now?
haha wow, what a cool concept. A style library! I honestly think Midjourney should consider making that. Like a little art locker on the website where you can store the things you like the most
sweet sassy molassy they did it! /list_tuners dropped yesterday. woohoo! @@FutureTechPilot
@Future Tech Pilot Do you have any videos showing how to keep the same character for all your photos? Thanks
kindaaa - ua-cam.com/video/ICh9ose7tGs/v-deo.html - that was made a while ago but the general principles still apply I think. There is no way to do it easily though, just keep that in mind
@@FutureTechPilot Thanks why it is possible to keep the same characters with leonardo but not with Midjourney?
Thanks for the update! I didn't even know.
haha what a pleasant surprise
hella slick
🤝
hello. thx for review, can we add weights to styles???
I just tried it and no, it does not appear to work
Is this the "personality" thing they were talking about? If so, it's different from what I thought and much more useful. Definitely warrants exploration.
By the way, if you want 130k subscribers by the end of the week, just find a way to combine midjourney with UFO rumors, and I'm not even completely joking (though... I am *mostly* joking ;) )
You know what, I don't think it's the personality quiz they were talking about. I mean, maybe it was, but I also think they have another 'demographic quiz' where we can tell the bot about ourselves and then it will provide us a style (something like that, don't quote me lol)
Might be nice if you could just select ANY of the options it gives you rather than picking left or right.
Yah I wonder they did that lol
why even use MidJorney when you have stable diffusion with all its lheckpoints, loras, inpainting, controlnet, latent couple, loop wave, animatediff ... and all thats for free, locally...
You're not wrong. But some people like the Midjourney experience more, and I'm sure a lot of people don't have the GPU to run SD locally
or the time. I use DALL-E for more control - and text of course - and MidJourney for the aesthetic effects. This tuning certainly helps with that
Does it work with weird?
I just tried it and no, "Using custom styles is not compatible with --weird" but you can add chaos and stylize values
@@FutureTechPilot thank you, Nolan!
i love you
❤
This feels like when MidJourney first got good again. This will change how everyone prompts now
The ultimate way to make unique art
Still using Discord and not a dedicated web site / app = next;
the website should have creation any week now
It’s truly a terrible new feature in my opinion. Bad results
hahah I can't get enough!
Any update on text in image? Great Video!
I know they figured out how to do it, but they aren't sure if including it in Version 6 will break some of their other features ... so maybe we see it next month, or maybe we see it in 3 months
@@FutureTechPilot thank you for the update. So weird how the text is so hard for AI image generators when it is the easiest part for a person to get right with precise instructions and seems as it would be amongst the easier things to get stable.
It's cool, but until it's more fine tuned on their end I feel like it's a waste of fast hours. Just my opinion though. Thank you for your videos. Makes it a lot easier to keep up with the changes.
Yeah I agree, it can burn through your hours extra quick! That's why it's really cool that these tests are shareable
Brilliant! Thank you for a quick and detailed explanation. Btw, selecting individual images from one large grid (rather than selecting between 2 sets of grids) can be even more nuanced. This is going to be a major time saver.
I'll mention the individual choices in a future video, but I guess I didn't realize that it would change the resulting style as well. Wow haha
Once again I want to tell you that you are the best at explaining MJ topics in simple to understand video
I really appreciate the comment! Thanks, and I'm glad my style works well for you
Holy Crap Batman! Can you just imagine what Midjourney is going to look like a year from now?
hahah no I really can't, but I'm here for the ride!
Until MDJourney is better trained with anatomy & can actually see what its creating as well as open the API so the ai can actually get to the point it understands what your telling it to do im sticking to Dale3 & Leonardo
that's fair!
The tuner seems VERY beta... Seems it just adding high --chaos and --weird values to the prompt!? The style code also isn't very reliabe... Kind of disappointed
I think there are some quirks but it's definitely way more polished than high chaos and weird
It'll be interesting once there is a shared database of styles that we can easily choose from.
100% - an easy to navigate database would be awesome
I see myself burning through a LOT of fast hours playing with this!! Thanks for the update 😊
😂 running these tests is actually so fun
Jezz! Complex but amazing. Definitely not a click bait, this is huge!!!
A little weirded out on how we are being part of a experiment, educating Midjourney on art styles.
On the other hand, can you imagine what a art historian could do fine tuning art styles like this?
Agree 💯 with everything you just mentioned! This comment ought to be pinned! @FutureTechPilot
We've been being experimented on for years. Wow, that sounded paranoid.
I was reading the other day about how some of the early AI experiments were done. One used a free app that "identified" dog and cat species based on a photo. It was an early stable diffusion test. They got mounds of data in just a couple of months.
Yah I'm a big fan of the precision. And this is only Version 1 so I think it can get a lot cooler
I have earned so many hours ranking image pairs its silly. I WISH I could be ranking image pairs that were building a portfolio of a larger style library that I could catalog for myself and tune specifically with hundreds or thousands of images. hey, I am HAPPY to let you use my time for building a better product in this fashion, I just wish it was useful for ME.
Maybe they'll build something like that in the future!
Deep dive please! 🙏 Innovative idea from them. Wish we didn't have to use up so many fast hours to hone in on styles, because, well, we're training it while we're helping ourselves, but the styles haven't transferred too well onto other subjects for me. But it's v1. Thank you for this! 💯
I've noticed that the 'category' of subject needs to be similar .. like objects work great, but an object -> to a person can be pretty odd
Also, let me know if you ever try the random style option haha
@@FutureTechPilot I did try! Very worth it if you have infinite number of fast hours! 😂
In case you know - can others see the style code?
@@AG_before Unless you're on Pro Plan and paying for Stealth Mode your prompts are visible on the site.
But can we generate consistent characters yet? Dall-e simplifies the process significantly so we don't have to come up with this complicated pre-fab style system.
No not yet, but we at least know it's being worked on
But Dall-e images are kinda buns. 😂
Do a lot of people use fast mode I prefer relaxed mode I'm on this every day so I would use up my fast very quickly
I think it varies from person to person. Relax mode is amazing, but it gets slower the more you use it, so I like to mix it up every couple of days
How do we remove a style? It keeps adding on to my prompts and I don't like the results.
Go into /settings and make sure 'stick style' is turned off
Is it cool yea but I feel for people who use art drawings for work inthis field will dry up
The future will definitely look different!
Excellent tutorial! I share with you the one I´ve done, but I keep looking for a more litographic effect:
retro vintage litography --no camera --style 1nLglodXNunR
I'll try it right now! Thanks!!
wow, so cool! I love them
I don't use Midjourney myself, but I like that prompt user styles might replace artists trying to copyright protect their unique styles. Prompt user styles will probably marginalise human artists styles like 1000000 to 1000 soon.
That's a really good point, I don't like including Artists names and this seems like a good compromise
Reply your best style code!!
let's see the goods people
If you leave the image selection page open you can go back and adjust which images to include in your style. You can't change the prompt, but your style code will change.
Yeah exactly, one test can give you so many styles
I see using this two ways: push or pull. You can tune a very specific prompt, and push the results toward different aesthetics with the tuner for the final output. Or you can tune a generic prompt to design a style that you might pull lots of different subjects through.
That's a great vision you have, makes a lot of sense to me!
Very very helpful - explanations are key !
I'm really glad I could help!
Awesome video. Def going to be coming back to this one
Well thanks for the comment!
lame
maybe! but I think it has a lot of potential
Insannneeee!!! I still didn't figure out what this exactly does, the Images "tuned" turn out to be everything but similar to any style I chose. So again another thing to figure out, why doesn't rhe day has 48 hours or more 😂
hahah I know what you mean. It's weird to think about, but I've been working on the best way to explain the tuning and I should be able to make it more clear in my next couple of videos
when do you get some sleep??? hahaha =D yeah I guess this new feature is a bite more complex to use, but this just makes it even more fascinating =D@@FutureTechPilot
@@rossobosso haha yeah I was up at 5am to film this, but I'm okay doing that once in a while when there is an update!
Keep up the Wonderful Work - Love It.
I appreciate the comment!!
amazing tutorial thank you soooooo muchhhh
Glad I could help!
Question: how can I create a style that is very close to the style of an unknown artist (not recogniazed by Midjourney)? I tried "describe" but not satisfied with the output.
Have you tried uploading an image by the artist to a chatbot that can describe it, putting the answer into midjourney and then picking the best likeness output into describe. Then use a few words that appear repeatedly in the descriptions for your prompt in tuning
Yes I did, but I tuess I need more prctice to be close to the style. @@yoagcur
You could try including multiple reference images in your prompt. Maybe like 3 image links and some text, should get you close to the style
Ok I will try it, and spend some money to try Runway. Thanks.@@FutureTechPilot
Love your work. Thank you
And thank you for taking the time to comment!
This is wild! When I think that Midjourney is getting behind, they always come with something new
I still think they are behind Dalle at this point. Simply on being able to follow a prompt accurately.
hahah yeah seeing A.I progress is the fun part of all this
thanks! that was helpfull
and thanks for the comment!
Hello from Texas!
🤝 sup Texas
Love your videos! A deep dive would be awesome!
Working on it as we speak!
I played with it a little this morning before leaving to work. Can't wait to get back home!
I don't know if i was coincidence but to me at least the first imagine is the original prompt with bo changes.
Yeah it's kinda hard to wrap your head around 😂
best tutorial yet
Glad you enjoyed it!
Thank you bro.
Cheers pal!
This is so huge!
The most potential ever to drop on a random Wednesday night lol
because it does not work? he always gives me an error, saying that it is incompatible with version 5.0 and asks for 5.1, I give him 5.1 and he tells me there is an error! what should I do?
style tuning currently only works with Version 5.2! --v 5.2
Very cool!
Right??
Interesting video...thanks
enjoy!
LOVE ✊🏿
🤝
thanks
🤝
Thanks!
🤝
wow...this such a great update...so helpful...I'll have to try this one...glad you noted about keeping them between 5-10, .I'd be the one that would pick them all lol...but thanks for always keeping us up to date...I'd be lost if you didn't lol.
hahah well I appreciate the comment and I'm glad I could help introduce the feature to you
This new feature is sick, you can blow through so much time just messing around with different images. Midjourney has come such a long way, its crazy.
haha exactly! The evolution is so fun to see
👋
🤝 I'm sure you'll find some use with this. I actually really like adding chaos and changing the prompt a little when using the code
Excellent guide, as ever. How do you blend more than one style, as mentioned towards the end, please? Thank you. Jim
I'll make a video about it but here is the quick answer - Combine multiple codes in one parameter with a hyphen, `--style code1-code2`
I dont think this should cost any minutes. It's just compiling images from millions of similar prompts and giving you a style number.
I don't think it's compiling images (not that I know for sure) but I think it's making a bunch of prompts for you, and then giving you the output, rather than giving you the words
Excellent, i see some pleasant ways to use it ! Thanks ! Great channel !
Cheers!
great new feature and awesome tutorial. New sub, yo
I really appreciate it!
Fuck, this ai hype is getting really boring
it's literally magic, I'm not sure how you could find it boring. But to each their own!
Another great one! Thank you for this again👍🏾👍🏾👍🏾🎉
I appreciate the support pal, cheers!
This is interesting! Thanks for showing this new feature. Can't wait to try...
and please share if you find anything cool!
Thanks for creating this great intro piece. Good work. Looks exciting and will definitely keep me tuned into MJ even more now!
I'm glad I was able to introduce it and it's cool that a lot of you are as excited as I am
It all looks really nice, but there is so little real use for it tbh
consistent style is a big selling point I think
can i get a style code from an image i have generated?
Interesting idea, you can add an image prompt to the /tune - but you need to add some words as well. Give that a try and let me know if it works
That's really good indeed ! Can't wait to use it.
have fun!
Is there an offline version I can run locally
No
nope!
That sound in the background is annoying
sorry!
This is gonna be GREAT for developers and designers. I’ve used MJ to avoid using stock images that cost 💲 but they never fit the same look or style. THIS changes that! Thanks FTP!
Yess, exactly. Enjoy!
Game changer! Great explanation as always FTP ❤
And thanks for the comment pal!
Pick your favorites from a big grid is another option to fine tune a style. might wanna look into that one as well.
Yah which option do you like to use?
Picking one, maybe 2 or three at most for a consistent combined style and keep your prompts real close to the initial prompt you used for training@@FutureTechPilot
Thanks for walking us through the feature 🎉
🤝 glad I could help
Always coming with the 🔥🔥🔥
just trying my best pal
Seems "meh"
I think it's pretty fun to use!
training LoRA is easier and free than this new feature lol
haha I try not to compare one process to another, I'm just happy Midjourney users get to experience this now
@@FutureTechPilot but it's paid and you can't run it locally on your own system.
cool features, i want moreeee!!!
haha I think this will keep you busy
Still without website?
creation hasn't been added yet, probably won't be there for a couple of weeks
Awesome
💯
hope you get to 100k soon! your videos are so so helpful! :D
haha that's the dream! I appreciate your support :)
I hate bragging, but i'm doing it with --no since V4... I figured it out and never stopped doing it... The results are very different and creative
Negative prompting is pretty unreal
Like your vibe man , keep going a good job 👍 thanks
I appreciate the comment pal, thanks!
Great video overview! Possibilities are endless
to infinity, and beyond!
Subbed
🤝
This looks a great update from MJ. I'm hoping that by distilling an entire style into a single code# will mean that the bulk of the prompt can be more focused on delivering on the subject/action and greater prompt accuracy.
That would be soo cool. I honestly think it kinda works like that, but it's still hard to understand the best way to prompt
That's definitely one of the prime features that were on my "reasons to resubscribe to Midjourney" list. Now just waiting for similar tool for saved characters and I'll get my hands on making a comic book of my own (well kinda) 😁
hahah I know exactly what you mean! I can't wait for more control
I still have no Effin clue - Mind blown.
😂
This was such a great video - thanks so much!
Glad I could help!
Generic.
maybe!