*** ZOOM *** Easily the most powerful thing I've seen in Midjourney for a mod. It turns some images I love into something almost profound with some results. Absolutely insane!
Some info I've learned about /shorten, might be helpful to understand what it's providing: the token weight values don't necessarily indicate an overall effectiveness in midjourney's ability to understand or give a repeatable and recognizeable output when using them in any given prompt. So if a token has a 0 weight, that doesn't mean it's useless to ever prompt with that word again. Instead, the weights are relative only to that prompt you gave it, and somehow reflect MJ's tendency to favor tokens in that prompt against one another, and is very dependent on things like the word order, so if you rewrote the exact same prompt reversing the order of your prompt's structure, it would show an uptick in the weight of a token that originally was furthest right, when now it is furthest left. Needs more testing, but this is some feedback I got from a moderator in the discussion channel, and it absolutely lines up with the results I was seeing.
Want to say thanks so much for the updated info.. when i got off work (saw some at work) after eating went to go try it out.. i am truly amazed what it can do.. think of this .. i made a pic prompt of a group of superheroines..so it generated 3 heroines in each of the 4 squares...i then took one i experimented with and had it go zoom out x2... when the picture finally did generated in the four squares, in the bottom right the prompt had added 2 more heroines to the picture from 3 to 5 ...wow !! so now i am sitting back and figuring what else i can do with this ...lol thanks again!!!
Future Tech Pilot, you think Midjourney devs will implement a feature which will allow the program to 'remember' generated characters (facial features, clothes, etc.) and changes their facial expressions and poses?
I know they're working on something along those lines, but I think they're going to put a 'Midjourney spin' on it, and I'm not sure what the final result will actually look like. But I know consistency and story telling is a priority for them and they absolutely have plans for it in the future
I used to be able to make cool black and white coloring pages with V5.1 but now using that same prompt gives me colors mixed in as well. Granted the images look better but my goal is to make black and white line art. Oh well but great video as always.
Might not be popouler, I personaly find 5.1 superior to 5.2. my previous prompts seems to generate a boring and unathntihc images (like MR X By Artist Y). I also noticed 5.1 variation is not working well and messing things up 😕
I still like Version 3 for some things and Version 4 was amazing too 😂 seems like each update will be different enough, where there will be new people who really connect with it
@@FutureTechPilot Correct. I love the new feature. and I just foundout its my stylize value became super sensitive. when punching it down my results came way more than I expected.
agree to disagree about the Shorten feature - even if it's interresting, after a few test with a specific prompt, the "longer" shorten result (if it makes sense) does not give the same stylitic result overall that the initial non shortened prompt (and i'm researching on very specific illustrative type of prompt engeneering). But then, as a comprehension tool it can be good, as a definitive tool not so much. Good video, as always, though. Cheers.
Yeah I'll have to experiment with it more. Because you're definitely right, the shortened prompt is not the same as the long prompt, so something must be getting lost in translation
Good is when you Custom Zoom that you make for example "Printed as a book cover" the zoomed picture be as the book cover, or T-Shirt Design etc.. I love it :D
@@SW-fh7he result quality, sharpness, and coherence all went up significantly… they added the zoom out feature… they added the ability to have more control over the reroll variations thing… and they added the ability to change aspect ratios via the advanced zoom out… it’s the best update in a long time 💪
@@FutureTechPilot it iiiiis! and it´s highly addictive, as MJ itself wouldn´t be addictive enough lol! i think I zoomed out already some 20, 30 times.... i wanna see how far it can go hahaha =D
@@FutureTechPilot Yes there are quirks. It'll get polished by v6 I hope. I was really looking forward to this feature. The thing is, once you open the details box, it nullifies some words which definitely make a difference. For example: 'depth of field' is given a '0.01', and sometimes even a '0'. Also, if you try and use any one of the suggested shortened prompts, it strays reasonably far from the original. Regardless of what genre you might pick, if you include 'photography', it will give it more weight regardless of anything else in the prompt. This second point I'm still confirming. But yeah, the shortened prompts should be used as a guide, and not a very strict one at that. It'll get better. Then they can include --improve . How cool would that be! Overall a wonderful leap forward though.
nice "monkey's paw" reference :) I think it is crazy that shorten removes lenses. lens size definitely makes a difference in depth of field and basically irl lens differences. i don't understand why shorten says it doesn't. it's also strange that shorten doesn't weight the subject higher than random descriptors. I think shorten may be a little wonky and is something they will improve down the road.
Something that MJ can’t YET do and that Stability AI just released, is the ability to zoom out from a user’s own image - one that they upload, that is. Stability’s called it Uncrop. So you could upload a famous painting, for example the Mona Lisa, run it through Uncrop and now you see more on all sides of the portrait.
Right! It would be extremely useful, I wonder if there is some technical limitation that prevents MJ from doing that or if we can expect it sometime in the future ... maybe when they finish their own custom Website
Dang. Ive been on photoshop 20+ years. I lost my job years ago cos i was very slow at creating hi end stuff. Everytime the boss said 'u got 4 hours' it would look like sh*t. So he let me go. Wish this was out then, id still have a job and the boss's import firm would be thriving :(
Excellent High Level Tutorial. You are a Midjourney Professor!! I am wondering about one thing though. {What's the point of using random seeds? There's no need. Seeds are only used to create similar-looking images from older previously generated images. By using seeds on new images it is the same as not using seed at all.} Right?
Thanks for the kind words! And seeds are only applicable to their current algorithm. Using the same seed in Version 4 will result in different pictures in Version 5. However, using the same seed on the same version will allow you to recreate an image, or test how small changes to your prompt will affect the generation. (like happy dog --seed 420423 vs sad dog --seed 420423) And choosing a seed yourself allows you to lock in that variable, so you can test things like a stylize value. But yeah, not choosing a seed lets MJ pick a random one for you. Hope that was clear! Sorry if I didn't explain it well
Not sure why anybody would try to entirely change the prompt when zooming out? If anything it would be just for subtle changes to the original section, or input for the newly added sections from the zoom out
I've experimented a bunch and it seems like erasing the prompt completely yields the best results ... because whatever you keep in the prompt, has the chance to show up again when you zoom out
I do know what you're trying to say: when using Ps Generative Fill you're actually increasing the image size with each addition. With Zoom Out you still end up with a max pixel size per each zoom out you do. A great way to fix this is to toss the image into Gigapixel, which not only upscales the image but also enhances it beautifully! Gigapixel is my FIRST stop on EVERY Midjourney image I save! Flawless upscaling! Midjourney being able to change the crop while keeping the original subject exactly as it was originally makes me think that a future update will have a 'subject masking' feature that will allow us to change the prompt on everything but the 'subject' and create an entirely new scene, or change the colors, or detail, or context or a background, etc while retaining the original subject; something we try to do so unsuccessfully these days with 'seeds'. I KNEW MJ could do better than they are currently doing with seeds, which at best offer a new imagined version of the seed image! How about "female car model leaning on car" in the changed prompt? I'll have to try it and see if now that I have the car photo I want I can add a model leaning on the same car! One area where I have to disagree with the "important tokens" aspect: lens details. When you prompt "shot with 35mm lens" as opposed to "shot with 600mm lens" you get bokeh and background differences based on the prompt (and, more importantly, the f-stop number you use). At least I've found that to be true. Why would the word "cinematic" be so unimportant in the 'shorten' prompt, yet add so much cinematic detail in the 'shortened' prompt? That's a puzzle, because I can only imagine 'cinematic' DID add a lot in the prompt with that word added; nothing else makes sense. ? Where do we find the blurred-out image in the final set?🤣 Honestly... this was a very telling vid! Thanks for sharing. If we hunker down and rake our brains to figure out how Midjourney comes up with its imagery, we'll die with brains of dust, or ashes, or something. It amazes me how MJ works! Oh, to be a fly on the wall at a Midjourney devs meeting!!!
haha I like your optimism and vision for the future of Midjourney! I agree that there are lots of areas for MJ to dive further into, especially when it comes to customizing a 'finished' image. I still need to experiment with the 'shorten' command more. Cause like you said, cinematic obviously has a lot of influence on the prompt ... and from what I've heard, those token weights aren't universal, as in, they change from prompt to prompt. And you're not missing much with the blurry picture haha it just included two things on her chest that resembled nudity in a slightly odd way .. I figured it was best to leave that to your imagination
A little short sighted to only implement a zoom out... and not a zoom in... but I imagine that may be coming in future versions (dare I suggest pan left, pan right?)
I was like 99% sure a 'pan left/right' would be included (I thought I heard them mention it) but maybe I misheard, or maybe we can expect it in the future!
Can you tell me please with the New (Custom zoom) at 3:08 How can I change the aspect ratio of an OLDER image I did some weeks ago at the beginning of this month. I just wanted to change its aspect ratio I did from the Large rectangle shape for the 'Facebook Profile Banner' size, to convert it into the --ar 3.2. Unfortunately because it was made with the previous version 5.1, I don't have that (custom zoom) option available. The only way i could do was to Click the "variations button" to do it again, and now I have the Zoom options. But these are new images. I wanted to alter only that old image, not the new images. Is there a cheat method to do that in Midjourney? 🙂
Sorry, I just tried some things and I don't think it's possible. You could try using your original picture as an image prompt, if you also give it a weight of 2 using '--iw 2' , it MIGHT work. The only thing I found is what you said, hitting variation buttons on an old photo seems to work ... but I think that's it
@@FutureTechPilot Yeah you are right, Only by hitting the variation button to an old image generation seems to work, it is a similar look to a picture it can get. But it is a new picture. It is not the specific picture i wanted. This works only with new images starting from version 5.2. This new feature of changing the dimensions of an image is Amazing!! I wish we had this feature some months ago. I managed to convert a Variation of an old image which had the Wide Aspect ratio of a Facebook banner pixel dimensions of (--ar 2037:754) and I shrink it into to the standard 3:2 aspect ratio. Thanks for your help. By the way did you know that from the version 5 and after you can make ANY picture size dimensions in Midjourney? Using any type of dimensions, such as pixels, millimeters, inches, centimeters, etc, instead of typing the typical Aspect ratio numbers of 3:2 etc... You can type any picture dimension Numbers in the place of aspect ratio. My old image I used had the Pixels dimensions of (2037 pixels x 754 pixels) for the aspect ratio. I learned this trick from this video => th-cam.com/video/hDQiqAD-Xg0/w-d-xo.html
No unfortunately. Not right now at least... You can definitely get close, I'd say like 80% likeness from frame to frame. But anyone who actually wants to make a video like that will probably hate the fact that there is 20% of consistency lost from picture to picture
@@Ecomcommunities th-cam.com/video/ICh9ose7tGs/w-d-xo.html - that video goes over the idea. Hope it sparks some creativity. Let me know if you learn anything new that I didn't try
Here's what I still find endlessly confusing whenever TH-camrs and bloggers talk about new parameters: we never know what BASE settings are being used! It's all well n good discussing the new stylize parameter on 5.2, but how does that interact with the base style settings (0.5, 100,250, 750)??? Does setting a style parameter completely override whatever base setting you have set? And what about RAW? That will also have a direct influence. I really just wish Midjourney would take this entire thing of Discord and simplify the process entirely. It's becoming so convoluted now to the point we spend more time re rolling and thinking "ooh what if I try this, then I might get that" than actually achieving anything at all - and all that re rolling is making MD an absolute fortune. A far FAR better interface would be to create a single image (not 4) then use sliders to simply add or reduce elements on that image. Forever re rolling, mixing, and creating variants to get to where you need to be is so tedious UNLESS you specifically want variation in a series of similar images. Really done with discord its a ball ache.
so far it seems like 5.2 really really prefers lifelike/realistic humans. I was doing a lot of character art for DnD, and 5.2 really needs me to explicitly promp for fantasy art rather than just saying "DnD character art". The same prompts that produced drawn images in 5.1 are producing hyperrealistic people in 5.2.
I have a question let's say I asked the bot to create a character and I liked the character, now I want this character to interact with other characters that the same bot made can I do that? I don't know if you guys understood what I said but I want to create characters that interact with each other without them changing every time I change their background is it possible to do? Thanks.
I'm having a difficult time trying to gain the same style of images with 5.2 vs 5.1. Its almost as if I have something turned on/off that wasnt before.
it's very different! No harm in just using 5.1 if you have some prompts that work really well there. I still think Version 3 can make amazing paintings lol
@FutureTechPilot Yeah, 5.2 is growing on me now that I better understand the differences and how to guide 5.2 differently from 5 and 5.1. Aside from that, do you know if the MJ team will expand the capabilities of the Zoom Out feature so that we can apply Zoom Out to any and all photos which we can upload to the MJ server or website? It's been a while since I used the website to generate images, to be quite honest.
@FutureTechPilot I've been a member of your server for a while but I could never identify you. Maybe it's because the times I'm active in your server you're not there.
I have big troubles with 5.2 since all my mediaeval witches now have dreadlocks and over and over again they appear suddenly like Naomi Campbell!!!!!! Any idea how to turn that craziness off?????
haha 5.2 is definitely different. One thing you can try is use a stylize value lower than 100. 100 is the default, but if you write --s 0-99, you'll make Midjourney follow your prompt more closely. These pics might not look great with a short prompt, so it's a little harder to use. I really don't recommend --s 0 lol maybe 30 or 40 would be the lowest I'd go
MJ 5.2 seems AWFUL unless you want all your “photos” to look like generic stock photos no matter what you write in the prompt. It’s better with art, but simply horrific with photos on the creativity side.
When changing the prompt, it's taking the whole image into account and trying to make something cohesive in the added space. so changing it to something wildly different is going to make weird results. Also, try it with image prompts for some really cool effects.
As always, thank you for another thoroughly enjoyable video. The thing that drives me absolutely crazy about these updates is when I see the results, after working hours, days, weeks and sometimes months on my prompts, now I want to re-roll all of them again in 5.2. It is literally maddening. I love it… Lol!😎✌️
So, I've played around in the 5.2 playground more and indeed the --stylize has a much more dramatic impact on images than previous versions apparently. Normally I use a very high stylize parameter and in previous MJ versions, it's usually very helpful. But in v5.2........ it's definitely not helpful for the vast majority of my prompts so by setting the stylize to very low levels, I'm actually getting more favorable results.
@FutureTechPilot I often prompt in v5.2 --s {0, 100, 250, 750}. Now, if I use --style raw in v5.2, I'll up the numbers to something like --s {200, 500, 1000} since the raw style significantly tamps down on the stylize parameter.
I dunno. I nearly spat my coffee out when it gave cinematic 0.01. You seen in this example how DRASTICALLY that word changed the whole composition. Im up to 15,000 images on MJ and cinematic is my most used. It changes everything and sets the lighting and colors beautifully. ?? Very strange.
@@armondtanzI hear ya. I had the same reaction. I think it's hard to isolate the effects of specific words when combined with others. "Voltage" is a very active word, whereas "Cinematic" is more of a style descriptor. Placing words at the beginning of the prompt also weights them differently. I'll be really interested to experiment with this feature. At least it's an attempt to allow us a glimpse under the hood of how MJ weights image components.
"Shorten" might be worthwhile if you regularly use long, involved prompts, but if you've been doing this for a while, you already know a lot of the useless words in prompts, and remembering what words are "useless" is somewhat taxing. It's helpful that using the word "lens" after "35mm" isn't necessary, but "global" in "global lighting" not being necessary... well, I can learn that from Grammarly. Now, if it says "key" isn't necessary in the prompt "key lighting" I'll have to write this feature off for a while. I don't see this as anything near a revelation.
I think I will try and boycot midjourney. I have hours to use I paid for a month or so ago. I can't use them or generate images because I don't have a paid subscription even tho i paid for extra hours. I have 60+ minutes of gens I could use. But since I dont have a subscription, I can't use it even though i paid extra for the 60+ minutes. This is wrong. I am not demanding anything for free. I just want to use what i paid for.
If you don't mention a seed number, MJ will keep making random pictures for you. But if you choose the seed number and keep the prompt the same, MJ will make the same pictures everytime
I know it's not exactly what you're looking for, but they updated the /remix feature th-cam.com/video/DgAxc8ZPU8U/w-d-xo.html - check out that video. You can KIND OF maintain a style now
Just a stock photography website, but it's really powerful in Midjourney, and allows you to get that specific look without needing a bunch of extra words
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
@@FutureTechPilot Oh ok. Im familiar with the unsplash site but I wasn't aware that it was actually a parameter that would mean something to midjourney. thank you!
@@FutureTechPilot actually I forgot to upscale, that’s why! I made a cinematic visual and then used « custom zoom » twice, deleted my prompt, changed -ar to 16:9, to expose it as an advertising on a huge screen in Times Square! (With a drone view). Mindblowing, so beautiful !! 🤩
@@FutureTechPilot thanks! For your info, once you’re in « custom zoom » you can delete totally your prompt and can write « mockup on a smartphone screen » (+location if you want), or « in a frame hanged on a (color) wall » Your visual is perfectly displayed it’s amazing !! Enjoy !!!
You Made Me SHED A TEAR when i didnt see one your famous Neon Prompts in this video!😆
You know what's so funny? I spent all of yesterday making neon batmans haha I'll try to include them in an upcoming video.
I appreciate you noticing 😂
Absolutely love your enthusiasm and positive energy, thanks again for yet another super informative post 🔥🔥🔥
And I appreciate you taking the time to comment, cheers!
*** ZOOM *** Easily the most powerful thing I've seen in Midjourney for a mod. It turns some images I love into something almost profound with some results. Absolutely insane!
I'm finding it a little tricky to use, let me know if you have any tips!
Some info I've learned about /shorten, might be helpful to understand what it's providing: the token weight values don't necessarily indicate an overall effectiveness in midjourney's ability to understand or give a repeatable and recognizeable output when using them in any given prompt. So if a token has a 0 weight, that doesn't mean it's useless to ever prompt with that word again. Instead, the weights are relative only to that prompt you gave it, and somehow reflect MJ's tendency to favor tokens in that prompt against one another, and is very dependent on things like the word order, so if you rewrote the exact same prompt reversing the order of your prompt's structure, it would show an uptick in the weight of a token that originally was furthest right, when now it is furthest left. Needs more testing, but this is some feedback I got from a moderator in the discussion channel, and it absolutely lines up with the results I was seeing.
Okay cool thanks for the info!! I'll try to experiment more tomorrow
Want to say thanks so much for the updated info.. when i got off work (saw some at work) after eating went to go try it out.. i am truly amazed what it can do.. think of this .. i made a pic prompt of a group of superheroines..so it generated 3 heroines in each of the 4 squares...i then took one i experimented with and had it go zoom out x2... when the picture finally did generated in the four squares, in the bottom right the prompt had added 2 more heroines to the picture from 3 to 5 ...wow !! so now i am sitting back and figuring what else i can do with this ...lol thanks again!!!
haha okay that's a pretty cool story! This feature just adds to the endless rabbit holes lol
Not even 5 minutes in and I’m already buying your prompt pack. Do more of them, I support raw talent!
Well I really appreciate the support! I'm working on more for sure
Future Tech Pilot, you think Midjourney devs will implement a feature which will allow the program to 'remember' generated characters (facial features, clothes, etc.) and changes their facial expressions and poses?
I got told theyre working on external website.
Yes i look at SD and controlnet and wish MJ would implement that.
I know they're working on something along those lines, but I think they're going to put a 'Midjourney spin' on it, and I'm not sure what the final result will actually look like. But I know consistency and story telling is a priority for them and they absolutely have plans for it in the future
I used to be able to make cool black and white coloring pages with V5.1 but now using that same prompt gives me colors mixed in as well. Granted the images look better but my goal is to make black and white line art. Oh well but great video as always.
hmm... that sounds frustrating for sure! I'll have to experiment with that as well
Funny to see shorten putting all supposed 'prompt engineers' on blast 😅
Great feature!
haha now it will just help them be even better!
Midjourney is amazing 👏 👏 👏 👏 👏 👏 👏 👏 👏 👏
💯
Thanks for this well-crafted update! I could not have asked for a better introduction to the new features.
Hey I'm happy to hear that! Cheers buddy
Might not be popouler, I personaly find 5.1 superior to 5.2. my previous prompts seems to generate a boring and unathntihc images (like MR X By Artist Y). I also noticed 5.1 variation is not working well and messing things up 😕
I still like Version 3 for some things and Version 4 was amazing too 😂 seems like each update will be different enough, where there will be new people who really connect with it
@@FutureTechPilot Correct. I love the new feature. and I just foundout its my stylize value became super sensitive. when punching it down my results came way more than I expected.
@@sqorpy Good stuff
The "dog in space" is useful to add a specific entity inside the vehicle. I've had trouble with this before. Thanks for the video 👏
haha that's a good point, maybe it's perfect for adding a subject in context of your picture
I notice that if the original image doesn't have the subject matter in it, it will add it to the outskirts.
Yeah I've heard a few people explain it that way!
Played w 5.2 for a few hours and lovin! 💯🎉👏🏻
It's a big change, eh??
agree to disagree about the Shorten feature - even if it's interresting, after a few test with a specific prompt, the "longer" shorten result (if it makes sense) does not give the same stylitic result overall that the initial non shortened prompt (and i'm researching on very specific illustrative type of prompt engeneering). But then, as a comprehension tool it can be good, as a definitive tool not so much. Good video, as always, though. Cheers.
Yeah I'll have to experiment with it more. Because you're definitely right, the shortened prompt is not the same as the long prompt, so something must be getting lost in translation
Great stuff, thanks for showing!
And I appreciate the comment, cheers!
I appreciate these updates. Thanks.
Cheers!
thanks for the summary, much appreciate the vid. keep up the fab work
I appreciate the support!
Love your videos & your enthusiasm 👍
I appreciate it, thanks for the comment!
Good is when you Custom Zoom that you make for example "Printed as a book cover" the zoomed picture be as the book cover, or T-Shirt Design etc.. I love it :D
woah... now that's a cool idea!
absolutely awesome. thanks for sharing
🤩
And thanks for taking the time to comment!
I think art classes in schools and art colleges will soon be a thing of the past.
Or they'll just adapt and be the ones teaching this tech!
5.2 is so good... they nailed it with this release... feels more like it should be 6 tbh... can't imagine how crazy 6 will actually be 😂
Why exactly?
@@SW-fh7he result quality, sharpness, and coherence all went up significantly… they added the zoom out feature… they added the ability to have more control over the reroll variations thing… and they added the ability to change aspect ratios via the advanced zoom out… it’s the best update in a long time 💪
@@KingZero69 please give me a comparison between two similar prompts between 5.1 and 5.2 😁
@@SW-fh7he you have to ask FTP for that my guy… 😏
@@SW-fh7heMaybe watch the video.
Great video. Commenting for the algorithm
haha I appreciate that! Cheers
Zoom Out is GREAT!!!! it´s almost like generative fill in PS Beta, only now implemented directly in MJ! I´m already testing out and trying out :D
Cool
Yeah it's pretty wild eh??
@@FutureTechPilot it iiiiis! and it´s highly addictive, as MJ itself wouldn´t be addictive enough lol! i think I zoomed out already some 20, 30 times.... i wanna see how far it can go hahaha =D
-- shorten 🤯
Thank you so very much! I was just there! I must run and play!
And please share any quirks you find lol
@@FutureTechPilot Yes there are quirks. It'll get polished by v6 I hope. I was really looking forward to this feature. The thing is, once you open the details box, it nullifies some words which definitely make a difference. For example: 'depth of field' is given a '0.01', and sometimes even a '0'. Also, if you try and use any one of the suggested shortened prompts, it strays reasonably far from the original.
Regardless of what genre you might pick, if you include 'photography', it will give it more weight regardless of anything else in the prompt. This second point I'm still confirming.
But yeah, the shortened prompts should be used as a guide, and not a very strict one at that. It'll get better.
Then they can include --improve . How cool would that be! Overall a wonderful leap forward though.
Soon the basic subscription will not be enough 😂😂
haha the rabbit holes never end...
nice "monkey's paw" reference :)
I think it is crazy that shorten removes lenses. lens size definitely makes a difference in depth of field and basically irl lens differences. i don't understand why shorten says it doesn't. it's also strange that shorten doesn't weight the subject higher than random descriptors. I think shorten may be a little wonky and is something they will improve down the road.
haha that's a good point! I'm looking forward to even more improvements
lets go!!! have been waiting for update..amazing as alwais.
💯!
Really enjoying/impressed by the new version!
I can't believe how different it is!
Something that MJ can’t YET do and that Stability AI just released, is the ability to zoom out from a user’s own image - one that they upload, that is. Stability’s called it Uncrop. So you could upload a famous painting, for example the Mona Lisa, run it through Uncrop and now you see more on all sides of the portrait.
Right! It would be extremely useful, I wonder if there is some technical limitation that prevents MJ from doing that or if we can expect it sometime in the future ... maybe when they finish their own custom Website
Much appreciated, really helping me stay ahead and keep the boss happy!
Dang. Ive been on photoshop 20+ years. I lost my job years ago cos i was very slow at creating hi end stuff. Everytime the boss said 'u got 4 hours' it would look like sh*t. So he let me go. Wish this was out then, id still have a job and the boss's import firm would be thriving :(
haha cheers buddy!
SO helpful!! Thank you. New sub
I'm really glad I could help! Thanks for the support. Cheers!
Thanks, great content! 🙏🏼
Cheers!
Thank you. Can you please show me how to create a new scene with 2 images? For example, I need a specific dog to sit on a specific turtle shell
Not possible in Midjourney but super easy to do in something like Photoshop or Canva
Really great explanations. Love your videos.
I appreciate the kind words and I'm happy my style works for you. Cheers!
Excellent High Level Tutorial. You are a Midjourney Professor!! I am wondering about one thing though. {What's the point of using random seeds? There's no need. Seeds are only used to create similar-looking images from older previously generated images. By using seeds on new images it is the same as not using seed at all.} Right?
Thanks for the kind words!
And seeds are only applicable to their current algorithm. Using the same seed in Version 4 will result in different pictures in Version 5. However, using the same seed on the same version will allow you to recreate an image, or test how small changes to your prompt will affect the generation. (like happy dog --seed 420423 vs sad dog --seed 420423)
And choosing a seed yourself allows you to lock in that variable, so you can test things like a stylize value. But yeah, not choosing a seed lets MJ pick a random one for you.
Hope that was clear! Sorry if I didn't explain it well
@@FutureTechPilot Thanks very much my dear friend. You explained it very well. Thank you! 😘
Absoultly stellar, I hope Adobe can get to this level of quality.
Competition is the best thing for consumers! So I'm definitely optimistic about the future of all these companies
Not sure why anybody would try to entirely change the prompt when zooming out? If anything it would be just for subtle changes to the original section, or input for the newly added sections from the zoom out
I've experimented a bunch and it seems like erasing the prompt completely yields the best results ... because whatever you keep in the prompt, has the chance to show up again when you zoom out
I do know what you're trying to say: when using Ps Generative Fill you're actually increasing the image size with each addition. With Zoom Out you still end up with a max pixel size per each zoom out you do. A great way to fix this is to toss the image into Gigapixel, which not only upscales the image but also enhances it beautifully! Gigapixel is my FIRST stop on EVERY Midjourney image I save! Flawless upscaling!
Midjourney being able to change the crop while keeping the original subject exactly as it was originally makes me think that a future update will have a 'subject masking' feature that will allow us to change the prompt on everything but the 'subject' and create an entirely new scene, or change the colors, or detail, or context or a background, etc while retaining the original subject; something we try to do so unsuccessfully these days with 'seeds'. I KNEW MJ could do better than they are currently doing with seeds, which at best offer a new imagined version of the seed image! How about "female car model leaning on car" in the changed prompt? I'll have to try it and see if now that I have the car photo I want I can add a model leaning on the same car!
One area where I have to disagree with the "important tokens" aspect: lens details. When you prompt "shot with 35mm lens" as opposed to "shot with 600mm lens" you get bokeh and background differences based on the prompt (and, more importantly, the f-stop number you use). At least I've found that to be true.
Why would the word "cinematic" be so unimportant in the 'shorten' prompt, yet add so much cinematic detail in the 'shortened' prompt? That's a puzzle, because I can only imagine 'cinematic' DID add a lot in the prompt with that word added; nothing else makes sense. ?
Where do we find the blurred-out image in the final set?🤣 Honestly... this was a very telling vid! Thanks for sharing. If we hunker down and rake our brains to figure out how Midjourney comes up with its imagery, we'll die with brains of dust, or ashes, or something. It amazes me how MJ works! Oh, to be a fly on the wall at a Midjourney devs meeting!!!
haha I like your optimism and vision for the future of Midjourney! I agree that there are lots of areas for MJ to dive further into, especially when it comes to customizing a 'finished' image.
I still need to experiment with the 'shorten' command more. Cause like you said, cinematic obviously has a lot of influence on the prompt ... and from what I've heard, those token weights aren't universal, as in, they change from prompt to prompt.
And you're not missing much with the blurry picture haha it just included two things on her chest that resembled nudity in a slightly odd way .. I figured it was best to leave that to your imagination
New Midjourney 5.2 zoom out feature + Photoshop beta's generative AI to center pictures = chef's kiss 🤌
It really does feel like creative heaven
A little short sighted to only implement a zoom out... and not a zoom in... but I imagine that may be coming in future versions (dare I suggest pan left, pan right?)
You can zoom in on your own. Just crop the image and then use a free image upsacler. No artificial image image generation is needed.
I was like 99% sure a 'pan left/right' would be included (I thought I heard them mention it) but maybe I misheard, or maybe we can expect it in the future!
Can you tell me please with the New (Custom zoom) at 3:08 How can I change the aspect ratio of an OLDER image I did some weeks ago at the beginning of this month. I just wanted to change its aspect ratio I did from the Large rectangle shape for the 'Facebook Profile Banner' size, to convert it into the --ar 3.2. Unfortunately because it was made with the previous version 5.1, I don't have that (custom zoom) option available. The only way i could do was to Click the "variations button" to do it again, and now I have the Zoom options. But these are new images. I wanted to alter only that old image, not the new images. Is there a cheat method to do that in Midjourney? 🙂
Sorry, I just tried some things and I don't think it's possible. You could try using your original picture as an image prompt, if you also give it a weight of 2 using '--iw 2' , it MIGHT work. The only thing I found is what you said, hitting variation buttons on an old photo seems to work ... but I think that's it
@@FutureTechPilot Yeah you are right, Only by hitting the variation button to an old image generation seems to work, it is a similar look to a picture it can get. But it is a new picture. It is not the specific picture i wanted. This works only with new images starting from version 5.2. This new feature of changing the dimensions of an image is Amazing!! I wish we had this feature some months ago. I managed to convert a Variation of an old image which had the Wide Aspect ratio of a Facebook banner pixel dimensions of (--ar 2037:754) and I shrink it into to the standard 3:2 aspect ratio. Thanks for your help. By the way did you know that from the version 5 and after you can make ANY picture size dimensions in Midjourney? Using any type of dimensions, such as pixels, millimeters, inches, centimeters, etc, instead of typing the typical Aspect ratio numbers of 3:2 etc... You can type any picture dimension Numbers in the place of aspect ratio. My old image I used had the Pixels dimensions of (2037 pixels x 754 pixels) for the aspect ratio. I learned this trick from this video => th-cam.com/video/hDQiqAD-Xg0/w-d-xo.html
@@eurossocial Yeah the possibilites are nearly endless!
Can you create sequential images that maintain characters in such a way as you can build theme based videos from the series of stills
No unfortunately. Not right now at least... You can definitely get close, I'd say like 80% likeness from frame to frame. But anyone who actually wants to make a video like that will probably hate the fact that there is 20% of consistency lost from picture to picture
@@FutureTechPilot is there any training that details how to generate characters whose likeness is maintained to at least that 80 percent.
@@Ecomcommunities th-cam.com/video/ICh9ose7tGs/w-d-xo.html - that video goes over the idea. Hope it sparks some creativity. Let me know if you learn anything new that I didn't try
If you want the photo zoomed back in all u gotta do is go to previous photos lol
hahah yeah it really is that simple eh
Here's what I still find endlessly confusing whenever TH-camrs and bloggers talk about new parameters: we never know what BASE settings are being used! It's all well n good discussing the new stylize parameter on 5.2, but how does that interact with the base style settings (0.5, 100,250, 750)??? Does setting a style parameter completely override whatever base setting you have set? And what about RAW? That will also have a direct influence. I really just wish Midjourney would take this entire thing of Discord and simplify the process entirely. It's becoming so convoluted now to the point we spend more time re rolling and thinking "ooh what if I try this, then I might get that" than actually achieving anything at all - and all that re rolling is making MD an absolute fortune. A far FAR better interface would be to create a single image (not 4) then use sliders to simply add or reduce elements on that image. Forever re rolling, mixing, and creating variants to get to where you need to be is so tedious UNLESS you specifically want variation in a series of similar images. Really done with discord its a ball ache.
I know they're working hard on a standalone website for creation!
You can do the infinite zoom in a video editor by keyframing and scaling the images. I understand it's already quite popular.
I've heard a little bit about it for sure, I'll have to check it out more in depth and maybe make a video on it!
I switched to the V5.2 in settings but, I do not see the zoom function when I click on upscale ??
Did you find a solution to the problem?
No
@@ralphbonilla2996 It might need to be a native 5.2 image
The only reason I don't like this tool is Discord. I hope they'll change this in the future.
They are 100% working on that as a top priority. I have no idea what the timeline is, but it will be coming eventually
@@FutureTechPilot Oh really, that's very good. I had no idea!
so far it seems like 5.2 really really prefers lifelike/realistic humans. I was doing a lot of character art for DnD, and 5.2 really needs me to explicitly promp for fantasy art rather than just saying "DnD character art". The same prompts that produced drawn images in 5.1 are producing hyperrealistic people in 5.2.
Really? 5.2 does seem like a pretty big difference!
I have a question let's say I asked the bot to create a character and I liked the character, now I want this character to interact with other characters that the same bot made can I do that? I don't know if you guys understood what I said but I want to create characters that interact with each other without them changing every time I change their background is it possible to do? Thanks.
No unfortunately, that isn't possible right now
@@FutureTechPilot well, Thats a bummer ill learn about ai for now and wait till it gets better. Thank you for replying.
I'm having a difficult time trying to gain the same style of images with 5.2 vs 5.1. Its almost as if I have something turned on/off that wasnt before.
it's very different! No harm in just using 5.1 if you have some prompts that work really well there. I still think Version 3 can make amazing paintings lol
CRAZY!!
💯
nice, tanks
🤝
bottom left has 2 elbows😆
oh it had some problems with that one for sure lol
5.2 image results are mostly 💩 compared to 5 and 5.1.
Yes, the zoom feature is very good but I'm not liking the aesthetic and framing algorithm in 5.2
It is super different eh? lol I do like 5.1 a lot but I also like exploring and seeing what a new update can make
@FutureTechPilot
Yeah, 5.2 is growing on me now that I better understand the differences and how to guide 5.2 differently from 5 and 5.1.
Aside from that, do you know if the MJ team will expand the capabilities of the Zoom Out feature so that we can apply Zoom Out to any and all photos which we can upload to the MJ server or website?
It's been a while since I used the website to generate images, to be quite honest.
@@LouisGedo haha I haven't heard that suggested! But you could definitely leave that as feedback on one of the Discord channels
@FutureTechPilot
I've been a member of your server for a while but I could never identify you.
Maybe it's because the times I'm active in your server you're not there.
What does Seed mean for midjourney or what's it actually for?
The seed is a number Id for the picture you promt
@@andreawilliams4778 oh ok, so if I want to refer back to that picture at another time I just use that seed number?
It just allows to you re-generate the same pictures. If you don't write a seed number, Midjourney will just pick a random one for you.
AWESOME😊😊😊
💯
"an retro"
😶
I have big troubles with 5.2 since all my mediaeval witches now have dreadlocks and over and over again they appear suddenly like Naomi Campbell!!!!!! Any idea how to turn that craziness off?????
haha 5.2 is definitely different. One thing you can try is use a stylize value lower than 100. 100 is the default, but if you write --s 0-99, you'll make Midjourney follow your prompt more closely. These pics might not look great with a short prompt, so it's a little harder to use. I really don't recommend --s 0 lol maybe 30 or 40 would be the lowest I'd go
But still midjouurney is paid and censored. I still prefer free, unsensored stable diffusion
That's fair!
MJ 5.2 seems AWFUL unless you want all your “photos” to look like generic stock photos no matter what you write in the prompt. It’s better with art, but simply horrific with photos on the creativity side.
Really? I hadn't noticed!
heh, they changed the seed.
hmm?
How come nobody seems to know how to correctly pronounce the word template!? Is this some kind of dorky Canadian nonsense?
😂 how do I say it? How is it supposed to be said?
@@FutureTechPilot you shouldn't say plate like dinner plate, the correct pronunciation = tem-plit (rhymes with fem-slit)
Meh
😶
Very useful, you are the man ! great video !!!! P34c3
I appreciate the support!
His brain can’t figure out how to crop an image 😂 hahahahaha
Dude it messes with my head lol and I don't know why
The "F1/8" doesn't mean anything in real photography. I think it's supposed to be "F/1.8".
😂 yeah, oops. I appreciate the correction
U right
I’m so obsessed with MidJourney!! Can’t wait to play with the new features this weekend! Thank you for such informative videos!
lol me too
Enjoy!
When changing the prompt, it's taking the whole image into account and trying to make something cohesive in the added space. so changing it to something wildly different is going to make weird results.
Also, try it with image prompts for some really cool effects.
😂 okay that's a way better way of explaining it. And thanks for the image prompting tip!
As always, thank you for another thoroughly enjoyable video. The thing that drives me absolutely crazy about these updates is when I see the results, after working hours, days, weeks and sometimes months on my prompts, now I want to re-roll all of them again in 5.2. It is literally maddening. I love it… Lol!😎✌️
hahah I could not relate more ... it's all definitely borderline overwhelming lol
So, I've played around in the 5.2 playground more and indeed the --stylize has a much more dramatic impact on images than previous versions apparently.
Normally I use a very high stylize parameter and in previous MJ versions, it's usually very helpful. But in v5.2........ it's definitely not helpful for the vast majority of my prompts so by setting the stylize to very low levels, I'm actually getting more favorable results.
Lower than 100? I've stuck at 400 and I'm pretty happy with it but I honestly think the default 100 is looking really good too
@FutureTechPilot
I often prompt in v5.2 --s {0, 100, 250, 750}. Now, if I use --style raw in v5.2, I'll up the numbers to something like --s {200, 500, 1000} since the raw style significantly tamps down on the stylize parameter.
WWWGGG
The changing prompt feature is for out painting buddy
I'll keep experimenting!
Will it zoom out a picture you upload?
Hmmm... I don't think so. I think it will have to be a generated upscale from Midjourney. Good question though
You are the man! Keep up the good work with these amazing MJ vids :)
Thanks for the support!
The 'shorten' feature is a revelation. A giant leap in the generative AI multiverse
Why?
Show your work
I dunno. I nearly spat my coffee out when it gave cinematic 0.01. You seen in this example how DRASTICALLY that word changed the whole composition. Im up to 15,000 images on MJ and cinematic is my most used. It changes everything and sets the lighting and colors beautifully. ?? Very strange.
@@armondtanzI hear ya. I had the same reaction. I think it's hard to isolate the effects of specific words when combined with others. "Voltage" is a very active word, whereas "Cinematic" is more of a style descriptor. Placing words at the beginning of the prompt also weights them differently. I'll be really interested to experiment with this feature. At least it's an attempt to allow us a glimpse under the hood of how MJ weights image components.
"Shorten" might be worthwhile if you regularly use long, involved prompts, but if you've been doing this for a while, you already know a lot of the useless words in prompts, and remembering what words are "useless" is somewhat taxing. It's helpful that using the word "lens" after "35mm" isn't necessary, but "global" in "global lighting" not being necessary... well, I can learn that from Grammarly. Now, if it says "key" isn't necessary in the prompt "key lighting" I'll have to write this feature off for a while. I don't see this as anything near a revelation.
What is unsplash?
A stock photography website, gives a really consistent cinematic look without needing a bunch of other words to set the mood
I think I will try and boycot midjourney. I have hours to use I paid for a month or so ago. I can't use them or generate images because I don't have a paid subscription even tho i paid for extra hours. I have 60+ minutes of gens I could use. But since I dont have a subscription, I can't use it even though i paid extra for the 60+ minutes. This is wrong. I am not demanding anything for free. I just want to use what i paid for.
Sounds frustrating! Hope it all works out eventually pal
what is that --seed number about, sorry if I missed something
If you don't mention a seed number, MJ will keep making random pictures for you. But if you choose the seed number and keep the prompt the same, MJ will make the same pictures everytime
Amazing but where can I see that pic of Crono
hahaha I meant to include in the video but I couldn't find a spot for it. Just use 'Chrono Warrior Pixel Art'
mj needs an im2img feature that allows style transfer for people who want to use it with ebsynth and animation/comics.
I know it's not exactly what you're looking for, but they updated the /remix feature th-cam.com/video/DgAxc8ZPU8U/w-d-xo.html - check out that video. You can KIND OF maintain a style now
@@FutureTechPilot Thanks, i like Mirror AI. Im using it to turn my 3d illustrations into comics. Working on one right now.
What's unsplash?
Just a stock photography website, but it's really powerful in Midjourney, and allows you to get that specific look without needing a bunch of extra words
THX
🤝
What does 'unsplash' mean in the prompt?
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
@@FutureTechPilot Oh ok. Im familiar with the unsplash site but I wasn't aware that it was actually a parameter that would mean something to midjourney. thank you!
Amazing! I put 5.2 in my settings but I don’t have the zoom out buttons and the 2 vary buttons. Tried on discord app on iPhone in Paris, France
hmmm ... still having the problem or did you find a solution?
@@FutureTechPilot actually I forgot to upscale, that’s why! I made a cinematic visual and then used « custom zoom » twice, deleted my prompt, changed -ar to 16:9, to expose it as an advertising on a huge screen in Times Square! (With a drone view). Mindblowing, so beautiful !! 🤩
@@calotite Wow, cool idea!
@@FutureTechPilot thanks! For your info, once you’re in « custom zoom » you can delete totally your prompt and can write « mockup on a smartphone screen » (+location if you want), or « in a frame hanged on a (color) wall »
Your visual is perfectly displayed it’s amazing !! Enjoy !!!
@@calotite Awesome suggestion! Thanks
stable difusion , panel controls is 100% better ,
Stable Diffusion is amazing for sure
Did anyone notice that the new Disney series intro for Silent Invasion used ai art?
I heard about it but haven't seen it myself yet. I bet we see it used a lot more in the near future!
These are pretty useful. Zoom and the prompt scoring will come in handy
Yeah eh it adds a lot more depth to the experience
It’s sooo goooood WHOOOAHHH 🤣🤣 thought about you last night
😂 I was up laaaate making this one. Jeeze, lots more rabbit holes to explore now
Still sticking to A1111 & Deliberate v2 since its just better
SD is great!
are you really thinking this is too suggestive for youtube, sad
You didn't see what it was without the blur... it was definitely suggestive lol
Lots of cool new features! Great video!
Cheers!
Daaaamn you´re really quick!!!! =D
😂 stayed up late to edit this one...
Thanks. Great explanation!
Happy to hekp!
Midjourney version 5.2 is pathetic, clinical, and crap.
to each their own!
@@FutureTechPilot No. TO Midjourney, Improve the product for all, and dump this clinical no soul 💩. Version 4 had a soul.
great sharing
Thanks :)
Yay 5.2!
Been a long time coming!
Start with generate a picture, than zoom out and out again and again and again, what’s everything color
What do ya mean?