Here’s a tip to save you guys some time you don’t need to re-upload the image, you can simply right click on the image that was uploaded for the describe feature, and use that one as the reference image file in front of your prompt.
@@cluepac Well considering that I have a Master of Fine Arts degree and over 15 years professional experience. I could easily draw and paint you out of this entire comment section any day.
A little time-saving tip if you’re using a laptop/PC: simply click and drag the uploaded image into the prompt box, then you don’t have to copy and paste anything, it does it for you.
Image weight really helped in getting better results while following these steps. I've been trying to create a character for my other channel, that I can achieve a consistent look in different poses, and kept getting characters with widely varying looks in MJ 5.2 (and even the new v6), same as in Dall-E 3. This has been the closest I've gotten to my starting reference image so far. Thanks. Merry Christmas!
Hey thanks for the feedback! I'm glad it worked well for you. And I've heard that Image Weight can go up to '3' in Version 6! So give that a shot for sure, --iw 3
i love the describe feature ive been rinsing it out for about a month now😆, like seriously i think its one of the best so far ...now u showed me some other tricks to it....wowwww thank u always for sharing!
Image weight is super important, first time I used it and the describe feature. Promote that more on your channel. Can't believe this video is 5 months old.
Hmm I'm trying to do /describe and copied my image address but it keeps giving me errors. Anyone else have the same problem? 'Error message is: You must provide an image or link to an image url to describe.' Which I thought I did.
what do you mean by Midjourney will fix hands in a few months? have they mentioned something about that? or is it because of what they mentioned about inpainting coming soon? i cant wait for inpainting, i guess we will get outpainting first but thats also cool
a little bit of both, but mainly that the tech will just continue to get better. And I think 'better hands' will just be a by-product of the overall progression.
It's also virtually impossible to get "A GREEK LYRE HARP" to depict *"Emperor Nero playing a GREEK LYRE HARP while Rome burns in the background outside a large window".* It keep showing Nero playing a regular wooden guitar like some 1970s hippie.
Midjourney might have a hard time with bows and arrows because they are very thin forms, and the diffusion process may require unusually high resolution sources to properly train for them.
I havent tried it recently but i did a month or two ago but it also has problems with sky full of stars/galaxy, swords and weapons still have some problems with it, snakes and similar shaped stuff and of course hands
Could use some kind of importance sampling, that identifies details like hands, bowstrings, etc, and diffuses those details with higher resolution. Then it enforces a model of how they should be built, i.e. hands can ONLY have 5 fingers, bows must be held by the grip in one hand and the fletched end of the arrow in the other, with the bow string running from the bow ends to the nock of the arrow...etc. I think we're going to need an engine change that builds an unshaded 3d model of certain elements, rigged hands for example, to make sure they obey physical rules, then rerenders the scene with generative details. 3D posing has been useful for generating human animation.
relay hope you can helps me love the idea and video , but i am finding it difficult ,now i can right click but i there is no option for copy image address ,only copy image ,so my problem is trying to get it to use my image
love MJ i just wish they could work on skin tones, race, and body types more, they really have an issue with "light skinned african american" vs "light skinned black man"... for example but the intricate racial skin tone differences are a real thing as well as hair types and lengths within racial groups -- example a black man with a short faded haircut is different than a black man with a ceasar styled haircut. In our world, these are real details the AI needs to be better at across all racial groups -- i say this here because i know you have a connection to them.
Oh I don't have any connection with anyone on the Midjourney team lol but I'm sure they agree with your assessment and I'd bet on it improving eventually
hahaha I tried to do this stuff as well, and it went ridiculous but very funny but we have to say that there are many other things it can barely do, or worse theyre' censored. And i don't mean sex or nude stuff, but even kisses. Have you ever tried to prompt something like that? you'll have hard time to make it or at least, this was on 5.1
Yeah Midjourney is super powerful, so they are trying their best to limit the graphic nature of possible generations. But unfortunately that means innocent and harmless prompts sometimes run up against the A.I Moderator. For a long time, you couldn't make 'Willy Wonkas Factor' because of the word 'willy' lol but I know they're working to improve it! I guess we just need to be patient
@FutureTechPilot then i don't understand. If the reference image gives you a prompt, and there is no reference to the subjects appearance in the prompt, yet it still somehow gives you a close approximation. Then whats the point of a promt at all if its just going to use the refrence image and spit out some variation? Maybe im missing something here
@@jess_n_atx I'm not really sure what you mean. Midjourney is a slot machine, so any generation is 1 in 4 billion. If you thought the skin tone looked similar without the reference image, I would say I just got lucky.
I keep getting a little box on my picture that gives an option to either "Spoiler attacement, Edit attachment or trash attachment. I can't get the window to pop up when I right click :( I have dragged and dropped still same thing what am I doing wrong? My Dell Laptop is upsetting me
oh dang, that sounds frustrating and I'm really not sure what's wrong! I think you should ask for help in one of the 'prompt chat' channels on Discord! Someone will be able to help you troubleshoot - good luck!
if you're making a fresh picture, something like 'holding a hat' would probably work fine, but if you already have the picture made, you'll want to try adding the hat using something like Photoshop
haha I have the exact same problem. I can't get it to look like me either. But there are some good face swap tools out there. I'll make a video about one soon
I've been doing this since they introduced the describe feature in MJ but the results were usually not quite what I was hoping for. But with how crazy fast AI's advancing these days, these features are bound to improve over time🤞
I saw someone else say this .. and I'm really not sure why that's happening lol I just checked and I have the 'upload file' option on a bunch of different channels. Please let me know if you find a solution to the problem
iam a bit tired of your video's, not for the reason you might think. Almost everytime i watch your video's about midjourney etc iam like OMG i like this i want to try this but i want to try 100 other stuff aswell haha. I have to be carefull iam not burn myself with information of your videos and others. AI is so damn interesting and your video's are really informative so thanks sir.
@@FutureTechPilot When I include an image link in a prompt, Midjourney asks for a more detailed description of the request besides the image size and technical data. I have to describe the picture and the result is something a little like the attached reference. Could you try it and make up your mind?
yea midjourney is great but it is terrible with weapons and authentic military uniforms and accoutrements. They need to improve this in their data sets! If you think trying to do a bow and arrow is bad, try doing a civil war musket! :)
You mean steal any pic in on the site that steals art and photos? How to have a car just like your dream car? Steal your dream car then slap your license plate on it and call it yours.
@@FutureTechPilot Yeah but it makes your channel look bad when you are repeating same stuff too much. Why wouldn´t you do a new video about this stuff again. There is always new people. Oh, that´s why. Because you did it already. That´s exactly my point. Still there is good stuff too.
SMH at how MJ requires you to post the url of the pic. What if the page isn't available any longer? Webpages expire all the time so it's not like that's uncommon. Once you upload the pic it shouldn't matter and you should be able to use it as a reference regardless. MJ seriously needs to improve its functionality so it becomes way more intuitive and user-friendly. That's why more and more people are using Leonardo.
Peace this gesture would be the go. Yours is equivalent to 🖕in my neck of the woods. Small potatoes, you do a terrific job and appreciate all your time and effort . I wish you wonder
Here’s a tip to save you guys some time you don’t need to re-upload the image, you can simply right click on the image that was uploaded for the describe feature, and use that one as the reference image file in front of your prompt.
Yeah exactly! Thanks for pointing it out
Came here to say this
better yet, you could learn to draw/paint
@@cluepac Well considering that I have a Master of Fine Arts degree and over 15 years professional experience. I could easily draw and paint you out of this entire comment section any day.
buuuurrrnnnnn 🤣🤣🤣🔥🔥🔥🔥🔥
A little time-saving tip if you’re using a laptop/PC: simply click and drag the uploaded image into the prompt box, then you don’t have to copy and paste anything, it does it for you.
I never knew that! Thanks a lot for telling me lol I'll include that in a future video
@@FutureTechPilot You're welcome!
Image weight really helped in getting better results while following these steps. I've been trying to create a character for my other channel, that I can achieve a consistent look in different poses, and kept getting characters with widely varying looks in MJ 5.2 (and even the new v6), same as in Dall-E 3. This has been the closest I've gotten to my starting reference image so far. Thanks. Merry Christmas!
Hey thanks for the feedback! I'm glad it worked well for you. And I've heard that Image Weight can go up to '3' in Version 6! So give that a shot for sure, --iw 3
@@FutureTechPilot Actually, I had heard that in a comment somewhere, so I tried it and it actually seemed to help! Happy New Year!!
i love the describe feature ive been rinsing it out for about a month now😆, like seriously i think its one of the best so far ...now u showed me some other tricks to it....wowwww thank u always for sharing!
I'm really glad I could help, and thank you so much for taking the time to comment! Cheers :)
Very, very useful trick using MJ. Thank you!
And thanks for the comment!
Image weight is super important, first time I used it and the describe feature. Promote that more on your channel. Can't believe this video is 5 months old.
haha okay thanks for the feedback, maybe I can repackage the info in a new way.
Thank you Brother! You saved my life!
haha cheers pal!
here's a tip: you can use the url of the image from the one you upload on /describe. No need to re upload it
Yeah I didn't know about that one, thanks!
THANK YOU SOOOOOOOOOOOOOOOOOOOOOO MUCH! I've been wondering how some people were getting such highly accurate pics of celebrities.
Yeah if they're using Midjourney, that's definitely how they would do it
or they might be using a Face Swapper tool
This is just what i needed. Thank you. This will save me the trouble I've been having.
Hey I'm glad I could help you out!
This approach works really well
I'm glad to hear!
Exactly what I was looking for thanks.
Happy I could help!
Hmm I'm trying to do /describe and copied my image address but it keeps giving me errors. Anyone else have the same problem? 'Error message is: You must provide an image or link to an image url to describe.' Which I thought I did.
Sounds frustrating. You should ask around in one of the Support channels on Discord. Sorry I'm not more help
No prob! Thanks for the reply. Glad I found your channel!
👏👏👏👏👏🙌🙌🙌 always enjoy coming here, it’s legitimate helpful ❤, thanks again for all 🤗
Hey I'm glad I could help!
Can I add in say "illustrative style" so I don't get a photograph?
That's a great way thing to try but I'm not sure it will work exactly how you're hoping
wow... amazing tips ,, i follow your instruction and it works, i'm subscribing to you right now
I appreciate the comment and I'm happy I could help you out!
Just got the prompt pack, great content. Also you got me my first three sales on Etsy last month. Thank you and keep up the great work.
Hey man that's super awesome to hear. Thanks for taking the time to comment, best of luck pal!
What do you sell on Etsy? How did you get started on it?
JUST EXCACTLY what I was Looking for!!! New Sub / Thumbs UP!!! keep going I want to know more!!
haha cheers buddy I appreciate the support!
what do you mean by Midjourney will fix hands in a few months? have they mentioned something about that? or is it because of what they mentioned about inpainting coming soon? i cant wait for inpainting, i guess we will get outpainting first but thats also cool
a little bit of both, but mainly that the tech will just continue to get better. And I think 'better hands' will just be a by-product of the overall progression.
@@FutureTechPilot oh okay, cool
So where is the difference between image weight and style reference?
This video might help you a little! - th-cam.com/video/XCDiul7Ln2s/w-d-xo.html -
It's also virtually impossible to get "A GREEK LYRE HARP" to depict *"Emperor Nero playing a GREEK LYRE HARP while Rome burns in the background outside a large window".* It keep showing Nero playing a regular wooden guitar like some 1970s hippie.
haha yeah it's tough to generate full scenes like that ... for now
Midjourney might have a hard time with bows and arrows because they are very thin forms, and the diffusion process may require unusually high resolution sources to properly train for them.
I havent tried it recently but i did a month or two ago but it also has problems with sky full of stars/galaxy, swords and weapons still have some problems with it, snakes and similar shaped stuff and of course hands
Good point! I wonder what the solution will be when they finally figure it out
Hair and fingers are also very thin forms but MJ seems to have significantly improved on rendering them. 🤔
Could use some kind of importance sampling, that identifies details like hands, bowstrings, etc, and diffuses those details with higher resolution. Then it enforces a model of how they should be built, i.e. hands can ONLY have 5 fingers, bows must be held by the grip in one hand and the fletched end of the arrow in the other, with the bow string running from the bow ends to the nock of the arrow...etc. I think we're going to need an engine change that builds an unshaded 3d model of certain elements, rigged hands for example, to make sure they obey physical rules, then rerenders the scene with generative details. 3D posing has been useful for generating human animation.
Great video, thx for the tip. The bow and arrow struggle is real my friends...
haha yeah I know what you mean
relay hope you can helps me love the idea and video , but i am finding it difficult ,now i can right click but i there is no option for copy image address ,only copy image ,so my problem is trying to get it to use my image
You need to left-click on the image first, then it will sort of expand and be bigger, then you right-click and you'll see copy image address
love MJ i just wish they could work on skin tones, race, and body types more, they really have an issue with "light skinned african american" vs "light skinned black man"... for example but the intricate racial skin tone differences are a real thing as well as hair types and lengths within racial groups -- example a black man with a short faded haircut is different than a black man with a ceasar styled haircut. In our world, these are real details the AI needs to be better at across all racial groups -- i say this here because i know you have a connection to them.
Oh I don't have any connection with anyone on the Midjourney team lol but I'm sure they agree with your assessment and I'd bet on it improving eventually
I learned so much from this. 😢
I'm glad I could help!
hahaha I tried to do this stuff as well, and it went ridiculous but very funny
but we have to say that there are many other things it can barely do, or worse theyre' censored. And i don't mean sex or nude stuff, but even kisses. Have you ever tried to prompt something like that? you'll have hard time to make it
or at least, this was on 5.1
Yeah Midjourney is super powerful, so they are trying their best to limit the graphic nature of possible generations. But unfortunately that means innocent and harmless prompts sometimes run up against the A.I Moderator. For a long time, you couldn't make 'Willy Wonkas Factor' because of the word 'willy' lol but I know they're working to improve it! I guess we just need to be patient
How does it know the skin tone and features of the man?
I think that comes from the reference image
@FutureTechPilot then i don't understand. If the reference image gives you a prompt, and there is no reference to the subjects appearance in the prompt, yet it still somehow gives you a close approximation. Then whats the point of a promt at all if its just going to use the refrence image and spit out some variation? Maybe im missing something here
@@jess_n_atx I'm not really sure what you mean. Midjourney is a slot machine, so any generation is 1 in 4 billion. If you thought the skin tone looked similar without the reference image, I would say I just got lucky.
its not 'ec-cetera' its ' et-cetera' . etcetera.
I think deep down I knew that lol
@@FutureTechPilot ah I get ya.
1:48 bro so strong he holds the saber by the light and defeats his opponents hitting with the handle only
😂 you don't mess with a guy like that
I keep getting a little box on my picture that gives an option to either "Spoiler attacement, Edit attachment or trash attachment. I can't get the window to pop up when I right click :( I have dragged and dropped still same thing what am I doing wrong? My Dell Laptop is upsetting me
oh dang, that sounds frustrating and I'm really not sure what's wrong! I think you should ask for help in one of the 'prompt chat' channels on Discord! Someone will be able to help you troubleshoot - good luck!
Hi a question, what prompt to use if I want to add a hat into the Hand of the character?
if you're making a fresh picture, something like 'holding a hat' would probably work fine, but if you already have the picture made, you'll want to try adding the hat using something like Photoshop
@@FutureTechPilot I tried with the remix mode to adjust and write it down but it sadly won't work at all.
@@Treasure-bl3cn yeah remix isn't good at small changes unfortunately
Fantastic video!
Thanks for the feedback!
As usual, top tutorial, you are my number 1 tutor on AI - cheers!
Thanks a lot for the feedback! I'm glad my videos work well for you. Cheers pal
Hi. Thanx for the Video. how do i make it so that the face of the person is not changed, only the surroundings
I don't think Midjourney can do that right now!
@@FutureTechPilot Thank you so much! Probably do u know, which prompt do you have to enter if you need a high quality of jpg?
@@Tk-182 Midjourney generations are all at the same pixel count! You can't select a higher quality right now
when i upload my photo It NEVER looks like me even when I upload the image weight.
haha I have the exact same problem. I can't get it to look like me either. But there are some good face swap tools out there. I'll make a video about one soon
"MJ is a slot machine - good luck". Comment of the year.
😂
very very nice!!!! Thank's!!!!!!
🤝
hello thanks for the video , however when i right click nothing happens ...so annoying . do you know why this might be???
You need to left-click on the image first, then it will sort of expand and be bigger, then you right-click and you'll see copy image address
Thank you!
I got you, pal
it’s frustrating that it doesn’t understand specific outfits
💯 frustrating for sure
Great Tips!! 🎉
Thanks pal!
I've been doing this since they introduced the describe feature in MJ but the results were usually not quite what I was hoping for. But with how crazy fast AI's advancing these days, these features are bound to improve over time🤞
💯 Exactly, it's only going to get better
I don’t have the ability to upload in discord. Am I missing something?
I saw someone else say this .. and I'm really not sure why that's happening lol I just checked and I have the 'upload file' option on a bunch of different channels. Please let me know if you find a solution to the problem
absolutely amazing and correct! So tired of those weird images
haha cheers!
Gracias!!
Cheers!
In my case the problem with “bow” was related to the fact that it has multiple meanings depending on context.
That's really interesting! Never thought of that
iam a bit tired of your video's, not for the reason you might think. Almost everytime i watch your video's about midjourney etc iam like OMG i like this i want to try this but i want to try 100 other stuff aswell haha. I have to be carefull iam not burn myself with information of your videos and others.
AI is so damn interesting and your video's are really informative so thanks sir.
hahah I feel like I burn myself out with all the information as well ... this tech is a lot to wrap your head around
one problem with MJ is when the reference image is blurry, almost all of the resulting images will also be blurry
Yah I guess that's true!
Thank U
🤝
What about paintings?
What do ya mean?
Can I remaster an uploaded painting, not a portrait? It seems to me I can't..
@@catress_online I don't see why that wouldn't work! What kind of problems are you having
@@FutureTechPilot When I include an image link in a prompt, Midjourney asks for a more detailed description of the request besides the image size and technical data. I have to describe the picture and the result is something a little like the attached reference. Could you try it and make up your mind?
@@catress_online That is why you should use the /describe feature first!
I need to paid for this option on midjorney or free sir ?
Midjourney needs a paid subscription!
love this video!
Cheers!
yea midjourney is great but it is terrible with weapons and authentic military uniforms and accoutrements. They need to improve this in their data sets! If you think trying to do a bow and arrow is bad, try doing a civil war musket! :)
haha I can only imagine!
Archery? - Does not compute....🤖😟
Bows and arrows are too old school for A.I
Awesome insight! Thanks for all your valuable vids. I will keep watching more. 💯😎#relaxzd
Cheers!!
When was John Boyega ever in a Star Wars movie? WTF?
...
I can't wait until we can actually converse with MJ like CGPT
The future is going to be absolutely wild
3:46
#3 🤣😂
Deadliest shot in the West
I thought everyone know about this lol
you'd be surprised!
You mean steal any pic in on the site that steals art and photos? How to have a car just like your dream car? Steal your dream car then slap your license plate on it and call it yours.
Hope you have a good day!
👋
🤝
Midjourney be racist calling Finn a Star Wars Villain.
haha maybe it was just the facial expression that signaled 'anger'
Nothing new.
To some people it will be
@@FutureTechPilot Yeah but it makes your channel look bad when you are repeating same stuff too much. Why wouldn´t you do a new video about this stuff again. There is always new people. Oh, that´s why. Because you did it already. That´s exactly my point. Still there is good stuff too.
@@tomi71 you should have a go with all your new ideas..........................waiting
@@mattwelcome2154 I have a new idea. Shut up.
SMH at how MJ requires you to post the url of the pic. What if the page isn't available any longer? Webpages expire all the time so it's not like that's uncommon. Once you upload the pic it shouldn't matter and you should be able to use it as a reference regardless. MJ seriously needs to improve its functionality so it becomes way more intuitive and user-friendly. That's why more and more people are using Leonardo.
I know what you mean, but as long as you can see it in the Discord chat, you'll never lose the pic (I think)
MJ is terrible with regard to weapons and objects that humans hold.
😂 sometimes it's randomly perfect but yeah overall, not the best
Peace this gesture would be the go. Yours is equivalent to 🖕in my neck of the woods.
Small potatoes, you do a terrific job and appreciate all your time and effort .
I wish you wonder
haha sorry! I appreciate you taking the time to comment. Cheers buddy