Midjorney has been heavily criticized for its plagiarism, making only small additions to the original images. And sometimes non at all. But under the guise of "AI" people think it's some super impressive tool that is creating images from scratch. It isn't doing that. For AI to get to that level, we're a long way off.
My first guess seeing your video was that using "portrait" in the prompt might be the main reason for the frontal portrait results you get. Still no excuse for the non-robotic/chrome face but that may be one screw to adjust it. Thanks for the video - very helpful!
I get the impression Midjourney's been trained to have a very distinctive interpretation of things, dark and kinda depressive I would say. It's like a pessimistic AI. So crazy how different it is to Dall-E 2 for example, there's a video around comparing them with the same prompts
You are definitely should try generation in different aspect ratios. Also check the website gallery and search whatever you like to do first, to find out what prompts works the best for your liking
I have had good success using something like this: in the style of Jasper Francis Cropsey|The Valley of Wyoming ------ here I reference the artist and a specific work of that artist separated by a pipe symbol. It's far from perfect, but it gives me the best results I've been able to find vis-a-vis working with an external image. Your mileage may vary.
With all the new updates, I think u should revisit this! I don’t think you were specific enough in your prompts, especially if you’re shooting for an idea in your head.
Our future is like nothing we've ever imagined, but more like what we're collectively imagining with Ai art. As soon as virtual reality/ the metaverse/neuralink is ready, all this will be reality... if it isn't already and we're imagining a time before it, while we're in it.
I think MidJourney is (so far) excellent in horrific imaging! And I mean that in the best way. It probably is so, because it's still in the uncanny valley region of image creation. And I think that's its charm. I hope that when AI gets better you will still be able to go back to this (then) old charm it has now.
I agree that MJ has "personality", creating unexpected things, but they could at least train better the disturbing/creepy distortion of the human figures.
Mixing images with text in the prompt is also fantastic so you can more or less direct your creation. By playing with relative weights you can get there...
@@fromunderthekilt6310 (1) you have to upload your image to some place on the internet from where it can be freely linked; (2) then you use the URL of your image (e.g, ending with .jpg) as the first element of the prompt; (3) you should play with the "--iw" setting to control how much weight your image will have in relation to the other [text] elements of the prompt; (4) you can use several images at once, but then your weight control works for all collectively; (5) you may also control the weight of different text elements of the prompt using ::n. Hope it helps...
Your problem with "Hajime Sorayama" is that instead of grabbing on to his style, it grabbed on to his image, like it literally referenced a picture of the artist, not his art(look him up and compare). I have had this happen using "in the style of H.R. Giger" before--all of a sudden not only does the image come off as a Giger painting, but one of the faces looks a whole lot like Giger himself. Is it a bug, a feature, miscommunication between man and machine? Maybe, it's a bit of each.
It'd be great if you could select and tweak certain parts of an image; maybe in the future command the bot/A.I to do exactly as you tell it. And i hope one day that could be part of an upgrade maybe. Now i know tgis is shittin on Concept Artists, but, imagine, being a director even and waking up in the middle of the night with an idea they cant shake and wanna see it on the fly, just voice in the commands and see a ton of results. That'd be cool, its to random to be art directable for the industry but theoretically speaking, everyone that's using itis an Art Director, id just love to see like tweak that head, body, legs, armour, sky etc individual selections. That'd be a game changer and more results other than jus four to choose from. I've been using it four, four days, and I've done over 1500-2000 pretty badass artwork. I stopped being a concept artist after a car ran over my head, head to hand coordination fucked up. But after gettin this for £30 a month, shit, its amazing, regards, cant wait to see its future. Jonny Bone
It's a common thing to say among artists... you try to draw a bicycle from memory. They are more complicated than you think. To be fair, I have to use a reference (for everything, I have Aphantasia) but Midjourney has a near-infinite number of resources at its disposal.
I tried to emulate styles of artists, and sometimes it gets there, especially the color palette and the overall proportions, but the human figures are horribly melted.
Good video. I ran into same problems. There is an issue with "in the style of..." or "painting by..." Midjourney uses a portrait of the painter plus painters style. These are horrible result and always happen with "David Hockney", "Salvador Dali" even "Rembrandt" and more. So a "woman, painted by David Hockney" will be a woman with Hockney head :( It is very hard/impossible to solve this. For "woman" I highly recommend always use "beautiful woman" - this improves results. Also not all styles have this painter issue. So for example " drawn by Junji Ito" works fine.
Great one, Dimitris! How would you put the link of a source image in MJ? I've tried to upload some images on my GoogleDrive but the sharing link won't work in MJ. Thank you!
Hey there! I had a similar problem. Just make sure that the url ends with the image file. So .jpg .png or whatever else. Any other URL won't work. Hope that helps!
@@marvelousdecay Thank you! All right, so it should end with *imagename.jpg* let's say, but from GDrive I don't see to customize the link in this way, did you use any other website ot upload your images?
I used my own website to host my images. Try maybe some other provider like dropbox or microsoft's one drive. Maybe you can get a better URL out of them.
Great video but if you don't mind a minor critique. In English it's not "how it looks like". It's just "how it looks". No need to add "like" at the end. It's redundant. Cheers!
I believe that is not true. According to the license agreement (midjourney.gitbook.io/docs/terms-of-service): "Subject to the above license, you own all Assets you create with the Services. This does not apply if you fall under the exceptions below." You do grant Midjourney a license to display, reproduce the image etc but as far as I understand you own the image. If you're a paying customer that is.
@@marvelousdecay Its yours, but anyone can use it, like Public Domain some how, CHECK By using the Services, you grant to Midjourney, its successors, and assigns a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute text, and image prompts you input into the Services, or Assets produced by the service at your direction. This license survives termination of this Agreement by any party, for any reason. Please note: Midjourney is an open community which allows others to use and remix your images and prompts whenever they are posted in a public setting. By default, your images are publically viewable and remixable. As described above, you grant Midjourney a license to allow this. If you purchase a private plan, you may bypass some of these public sharing defaults.
As for artistic imagery, I was trying to paste verses of song lyrics. But got stuck when they started to censor/ban absurd words such as FLESH -- wtf? With this provincial mindset they're not going far.
What MJ does is it turns the images into words, it "reads" the picture and then creates something out of its own imagination. This makes things difficult because there are a lot of words that MJ doesn't know.
how did these look without using artist's names? I'm just concerned it was wasted info, like if this AI doesn't know every artist's name, or it if was trained on that artist's work at all. I do see it obviously knows Giger there
I have fumbled with MidJourney, and I think it is great to produce distorted, deformed pictures of living beings. As for inanimate object, unless the commands are very short it get easily lost and never gets the "recipe" correctly. I'm disappointed. Maybe the devs wanted it to have a mind of its own, but who needs a wayward AI? Maybe just creative professionals in search of inspiration?
It's definitely a nice tool for quick brainstorming. It's like a second brain thinking way outside of the box! :D That being said I wouldn't use it for final production but I'm sure that things will change rapidly the coming months. I've already noticed big improvements in certain aspects the past few weeks, so I wouldn't be surprised if in a year's time we will see masterpieces coming out of Midjourney!
I pray to God it will never be directable. It's simply too powerful of a tool, it would completely eradicate artists as we know them and leave them jobless.
imagine how good midjourney will be in 5 to 10yrs, amazing! this is like the iPhone 1 of AI Art. love the upcoming features the mj team is working on
Amazing, thanks to human intelligence that created A.I
Midjorney has been heavily criticized for its plagiarism, making only small additions to the original images. And sometimes non at all. But under the guise of "AI" people think it's some super impressive tool that is creating images from scratch. It isn't doing that. For AI to get to that level, we're a long way off.
My first guess seeing your video was that using "portrait" in the prompt might be the main reason for the frontal portrait results you get. Still no excuse for the non-robotic/chrome face but that may be one screw to adjust it. Thanks for the video - very helpful!
This is EXACTLY what I was thinking all the time when experimenting with midjourney for multiple weeks, I'm so glad I found this
I know, you've taught . It ans so much to so many people, you are helping us pursue our dreams! Love from Sweden
I get the impression Midjourney's been trained to have a very distinctive interpretation of things, dark and kinda depressive I would say. It's like a pessimistic AI. So crazy how different it is to Dall-E 2 for example, there's a video around comparing them with the same prompts
kind of yeah.. noticed it too. maybe it's just more pesimistic seeing the world because of the nature of the artists itself ?
The David Hockney inspired piece is like a mix of Hockney and Francis Bacon, pretty wild. Thanks for your videos!
You are definitely should try generation in different aspect ratios. Also check the website gallery and search whatever you like to do first, to find out what prompts works the best for your liking
I have had good success using something like this: in the style of Jasper Francis Cropsey|The Valley of Wyoming ------ here I reference the artist and a specific work of that artist separated by a pipe symbol. It's far from perfect, but it gives me the best results I've been able to find vis-a-vis working with an external image. Your mileage may vary.
I found if you change the aspect ratio of the image it can vary greatly since the composition depends on it.
With all the new updates, I think u should revisit this! I don’t think you were specific enough in your prompts, especially if you’re shooting for an idea in your head.
Our future is like nothing we've ever imagined, but more like what we're collectively imagining with Ai art.
As soon as virtual reality/ the metaverse/neuralink is ready, all this will be reality... if it isn't already and we're imagining a time before it, while we're in it.
Great tutorial! Super easy to understand!
Great to hear!
I think MidJourney is (so far) excellent in horrific imaging! And I mean that in the best way. It probably is so, because it's still in the uncanny valley region of image creation. And I think that's its charm. I hope that when AI gets better you will still be able to go back to this (then) old charm it has now.
I agree that MJ has "personality", creating unexpected things, but they could at least train better the disturbing/creepy distortion of the human figures.
Mixing images with text in the prompt is also fantastic so you can more or less direct your creation. By playing with relative weights you can get there...
How can you upload your own image? Thanx in advance!
@@fromunderthekilt6310 (1) you have to upload your image to some place on the internet from where it can be freely linked; (2) then you use the URL of your image (e.g, ending with .jpg) as the first element of the prompt; (3) you should play with the "--iw" setting to control how much weight your image will have in relation to the other [text] elements of the prompt; (4) you can use several images at once, but then your weight control works for all collectively; (5) you may also control the weight of different text elements of the prompt using ::n. Hope it helps...
@@adrianokury fantastic sir. Thank you so much, appreciate it!
Your problem with "Hajime Sorayama" is that instead of grabbing on to his style, it grabbed on to his image, like it literally referenced a picture of the artist, not his art(look him up and compare). I have had this happen using "in the style of H.R. Giger" before--all of a sudden not only does the image come off as a Giger painting, but one of the faces looks a whole lot like Giger himself. Is it a bug, a feature, miscommunication between man and machine? Maybe, it's a bit of each.
It'd be great if you could select and tweak certain parts of an image; maybe in the future command the bot/A.I to do exactly as you tell it. And i hope one day that could be part of an upgrade maybe. Now i know tgis is shittin on Concept Artists, but, imagine, being a director even and waking up in the middle of the night with an idea they cant shake and wanna see it on the fly, just voice in the commands and see a ton of results. That'd be cool, its to random to be art directable for the industry but theoretically speaking, everyone that's using itis an Art Director, id just love to see like tweak that head, body, legs, armour, sky etc individual selections. That'd be a game changer and more results other than jus four to choose from. I've been using it four, four days, and I've done over 1500-2000 pretty badass artwork. I stopped being a concept artist after a car ran over my head, head to hand coordination fucked up. But after gettin this for £30 a month, shit, its amazing, regards, cant wait to see its future. Jonny Bone
It's a common thing to say among artists... you try to draw a bicycle from memory. They are more complicated than you think. To be fair, I have to use a reference (for everything, I have Aphantasia) but Midjourney has a near-infinite number of resources at its disposal.
ai artwork generators work better when you dont have a particular result in mind
prompt writers are, therefore, not artists
I tried to emulate styles of artists, and sometimes it gets there, especially the color palette and the overall proportions, but the human figures are horribly melted.
I really don’t understand how you can upload an image in MJ. Having an url and then Copy/paste in the prompt ???
How do you upload an image into the prompts? I had no idea you could do that!
Check the image prompting documentation here:
midjourney.gitbook.io/docs/user-manual#image-prompting-with-url
Good video. I ran into same problems. There is an issue with "in the style of..." or "painting by..." Midjourney uses a portrait of the painter plus painters style. These are horrible result and always happen with "David Hockney", "Salvador Dali" even "Rembrandt" and more. So a "woman, painted by David Hockney" will be a woman with Hockney head :( It is very hard/impossible to solve this. For "woman" I highly recommend always use "beautiful woman" - this improves results. Also not all styles have this painter issue. So for example " drawn by Junji Ito" works fine.
I can't wait to see photographers using DALLE 2 to copy their own style it would be very interesting 🤔
ikr too. i clicked on sotNice tutorialng and now i cant find any tracks and everytNice tutorialng is blank idk what to do
I want to learn more how best to include REFERENCE images, and multiple image to my prompt to steer the render.
Have a look at mid journey's documentation here:
midjourney.gitbook.io/docs/imagine-parameters
What you want is the image weight command.
Great one, Dimitris! How would you put the link of a source image in MJ? I've tried to upload some images on my GoogleDrive but the sharing link won't work in MJ. Thank you!
Hey there! I had a similar problem. Just make sure that the url ends with the image file. So .jpg .png or whatever else. Any other URL won't work. Hope that helps!
@@marvelousdecay Thank you! All right, so it should end with *imagename.jpg* let's say, but from GDrive I don't see to customize the link in this way, did you use any other website ot upload your images?
I used my own website to host my images. Try maybe some other provider like dropbox or microsoft's one drive. Maybe you can get a better URL out of them.
@@marvelousdecay got it, thank you very much for your help, Dimitris!
@@ArgoBeats how did you solve this? Trying with Dropbox but I get an error message
Is there a way to do it on Microsoft Paint??
Great video but if you don't mind a minor critique. In English it's not "how it looks like". It's just "how it looks". No need to add "like" at the end. It's redundant. Cheers!
sway og youngboy bro that’s how everyone starts even Kanye admited that Nice tutorials soft weren’t great at first just keep grinding and you’ll
Great Test !!!! Remember you don't own any images, Midjourney is the owner.
I believe that is not true. According to the license agreement (midjourney.gitbook.io/docs/terms-of-service):
"Subject to the above license, you own all Assets you create with the Services. This does not apply if you fall under the exceptions below."
You do grant Midjourney a license to display, reproduce the image etc but as far as I understand you own the image. If you're a paying customer that is.
@@marvelousdecay Its yours, but anyone can use it, like Public Domain some how, CHECK By using the Services, you grant to Midjourney, its successors, and assigns a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute text, and image prompts you input into the Services, or Assets produced by the service at your direction. This license survives termination of this Agreement by any party, for any reason. Please note: Midjourney is an open community which allows others to use and remix your images and prompts whenever they are posted in a public setting. By default, your images are publically viewable and remixable. As described above, you grant Midjourney a license to allow this. If you purchase a private plan, you may bypass some of these public sharing defaults.
try adding words like shiny metal reflective metal mirrored surface
As for artistic imagery, I was trying to paste verses of song lyrics. But got stuck when they started to censor/ban absurd words such as FLESH -- wtf? With this provincial mindset they're not going far.
How to I get midjourney
Work 100%
What MJ does is it turns the images into words, it "reads" the picture and then creates something out of its own imagination. This makes things difficult because there are a lot of words that MJ doesn't know.
try "in the style of Soriyama"...
robot chrome hajime sorayama woman
gives you chrome results!
even better is:
robot chrome hajime sorayama woman figure dance jump
for you
yes, sort of
OMG! You are a gorgeous man!
how did these look without using artist's names? I'm just concerned it was wasted info, like if this AI doesn't know every artist's name, or it if was trained on that artist's work at all. I do see it obviously knows Giger there
You can see the difference in the Hajime Sorayama prompt. Without the name of the artist the looks was completely off
I wonder when the beta is over. I've applied a few times and zero response. I'm getting very antsy and want to try this out! 😁
Yeah no idea when it's coming out of beta. But look at it from the bright side! You will get to use a more robust AI! That's awesome!
I have a guy that can do it for you and he installs
tNice tutorials, good luck- you'll go far
I have fumbled with MidJourney, and I think it is great to produce distorted, deformed pictures of living beings. As for inanimate object, unless the commands are very short it get easily lost and never gets the "recipe" correctly. I'm disappointed. Maybe the devs wanted it to have a mind of its own, but who needs a wayward AI? Maybe just creative professionals in search of inspiration?
It's definitely a nice tool for quick brainstorming. It's like a second brain thinking way outside of the box! :D
That being said I wouldn't use it for final production but I'm sure that things will change rapidly the coming months.
I've already noticed big improvements in certain aspects the past few weeks, so I wouldn't be surprised if in a year's time we will see masterpieces coming out of Midjourney!
@@marvelousdecay Fantastic perspective!
SO, what the AI created is NOT art, but after YOU change a few things, it is?!?! 🤣🤣🤣🤣
I pray to God it will never be directable. It's simply too powerful of a tool, it would completely eradicate artists as we know them and leave them jobless.
LOL
You talk alot