V6 Prompt Design - Midjourney Beginner Tutorial
Vložit
- čas přidán 15. 06. 2024
- Midjourney Beginner Course ➡️ ftpai.thinkific.com/courses/m...
-
-
I thought I would break down what a successful prompt is actually made of. This might be a basic English lesson disguised as a Midjourney tutorial, but I hope you find it helpful to learn what Version 6 can handle!
NEW Free V6 PDF ➡️ ftp.live/V6Glossary
2-4-2 V6 PDF ➡️ ftp.live/242
Discord Thread ➡️ / discord
Exclusive Prompts ➡️ ftp.live/patreon
Podcast Link ➡️ / @futuretechpodcast
► FREE PDFS
21 Prompts ➡️ ftp.live/free-prompt-pack-pdf
Prompt Ad-Libs + 20 Category Key Words ➡️ ftp.live/ad-libs
Camera Angles + Lenses ➡️ ftp.live/camera-lenses
200 Aesthetics ➡️ ftp.live/200Styles
Midjourney Cheat Sheet ➡️ ftp.live/CheatSheet
► PROMPT PACKS
Midjourney V5.1 Pack ➡️ ftp.live/V51-Pack
Photorealism Prompt Pack ➡️ ftp.live/photoreal-pack
Midjourney Variety Pack #1 ➡️ ftp.live/variety-pack
A.I Art Ad-Libs ➡️ ftp.live/ad-libs-premium
► SOCIAL
Discord - / discord
Instagram: / futuretechpilot
Thanks to www.merriam-webster.com/dicti... for some extra info
Timestamps:
0:00 Intro
0:26 Nouns
2:12 Prepositions
4:04 Adjectives
5:44 Adverbs
7:58 References
9:21 Prompt Example
If someone tell me ten years ago that I will try to learn English just to communicate better with AI, I would say that person is crazy. Yet here I am. Learning about adjectives, prepositions and nouns... btw- sometimes v6 prompts give incredibly beautiful pictures in v5.2. Check it out sometimes by adding {,--v 5.2} at the end of prompt.
Right? You're not the only one. I'm a native English speaker who never paid attention in school. Learning out of necessity is a trend I've noticed in my adulthood.😅
haha life is funny, eh? And that's a cool tip about using prompts in the old versions! I think you should even try --v 3 😂
I feel the same way!
Don’t know if it’s worth an additional video covering it, but I find that planning the main subject will also significantly affect the composition in a way that’s hard to do otherwise. e.g. "a person walking in a city park" will compose the image so the person is at the center and fairly close to the viewer, whereas "a city park with a person walking" will widen the angle to view more of the park, and the person can be place anywhere within it, usually further from the viewer.
That's some really interesting insight, thanks a lot for sharing and I'll try to explore that on my own a bit. Maybe I can slip it into a future video!
Awesome tip! Thanks @sidewaysdesign!
i do this when i want a wide shot, describe the environment first, then say theres a person there and describe the person--describing any person in the prompt gives them massive priority, so sometimes i have to say 'a person in the far far distance' (it still makes them be more than 50% of the frame lmao, but its a trick that works
This is all cool, but can you try to do a deep dive into getting images that are in an editorial advertising style. Meaning, no space aliens, glowing plants, planets etc, but a dad with his two children on the backseat of a car blowing soap bubbles for example. Real natural, to put on a billboard. - most midjourney tutorials go hard into cinematic, dark moody scenes, but what are those for other than look cool? So It would be great if you could dive into something more practical. Im sure people are looking for that!
Yah sure!
ive been taking a break from midjourney for the last couple weeks and never really gave it a try till now. my god the creativity of v6, especially raw mode is mind-blowing. i did some side by side comparisons, with the same prompt in vanilla stable diffusion and the difference is crazy. i may bump up my sub to the 30 dollar level once they add zoom/pan/inpainting. i feel like this is something you can genuinely make art with. the *any skin showing at all* censorship is annoying but for how good it is i can let it slide
haha I know what you mean. There are many rabbit holes to explore and each prompt offers so much potential
Thanks to you and the MJ crew for putting these guides together to help us all level up.
Cheers!
One of your best videos. Thanks for sharing the knowledge.
I'm glad you liked it!
I'm finding image references in combination with adding weights:: to strengthen certain aspects of the prompt is the best way to achieve specific/complex results in V5.2 & V6. Especially if you're trying to achieve characters in actions/environments for storytelling. It removes a lot of the randomness, improves coherence and delivers way more accurate results than a more poetic, verbose style.
now just imagine if we had controlnet capabilities! =)
Care to share any tips you might have? What have you found that doesn't work? What do you think works best?
@@FutureTechPilot
OK. I needed to create the scene of an old man being interviewed by a young policewoman in a warehouse. The style is very cartoony and the characters are caricatures. After many hours and failed attempts using plain language descriptions I couldn't get the result I was after. So...
I started with a simple prompt to get as close as possible to the desired outcome. I chose 6 images that varied in style and included their url's after the first heavily weighted subject...
a policewoman interviews an old man::16
[ADD THE 6 IMAGE REFERENCES]
a female police officer listening to an old man::8 an old man and a young police woman in a warehouse::8 caricature in the style of B&W comic illustration::8 grotesque caricature illustration::8 clean black lines, white background, heavy line-work, Comicbook, highly detailed, rule-of-thirds --chaos 5 --ar 16:9 --stylize 250
As I get images I like, I replace the original image references then paste the new prompt into another image (that has stylistic qualities I'm chasing), via "vary strong" This helps build and maintain coherence.
The benefit of this method is that you can easily increase or decrease weights to control various aspects of your scene.
ATM, in V5.2, (using this prompt as an example) I can also add weights to the image references, which makes a huge difference.
Unfortunately, in v6, MJ limits you to 7 total. Hence I'm currently getting more accurate results from v5.2 with this method.
@@pedxing i dont think they'll ever have true honest-to-god controlnet sadly. their --iw parameter is so weak its a joke. i get they dont want image to image because probably nudity concerns, but why not give us edge/depth/pose to image ? would be powerful
@@spacekitt.n fair point and I would LEAP at the chance to use depth maps in MJ. move this to the top of the dev list! =)
Thanks a lot for this great video !!! from Paris, France 🇫🇷
Merci beaucoup!
Fantastic highlights Nolan bro! 🎉
I appreciate the support!
Funny experiment I did a while back. I wondered what would happen if I just typed in the hex code of a specific color instead of describing it and to my disappointment, MJ didn't recognize it at all and associated the hex value to something absolutely random and the colors weren't even anywhere near to what I wanted. Just thought I'd share but I still definitely hope MJ will have this feature in the future. Would be cool to have.
haha hex codes are the dream! Seems like it would be pretty easy to implement but what do I know 😂
Fantastic video. It actually is _really_ impressive how much it picks up on the use of certain words. I even get a difference between 'a dreary morning', 'a cloudy morning' and 'a sombre morning'!
An excellent and very helpful video! 💯 Add an unintentional positive side-effect to using MJ - improved diction. Or is it 'for using MJ? lol Or 'of using MJ'...🤔 AWESOME! And be careful about spelling because I tried 'man turning into many bats', and 'man turning into man bats' - _very_ different images! edit: ALSO, THANK YOU FOR THE PROMPT IN THE NEWSLETTER! It's _awesome_ !
haha wow, thanks for reminding me about spelling. I definitely should have emphasized it more, because v5 didn't really care about your spelling at all lol now it's a big deal for sure - also, I'm glad you liked the prompt - it was too good not to share
THANKS,, THANKS, THANKS,, you are very generous with us
I'm glad I could help!
Great resource and guides Thankyou! 😊
I'm glad I could help!
Omggg you explaining all this really well!!! Thank you!!! 🤩🤗
And thank you for the feedback!
🎯 Key Takeaways for quick navigation:
Foundational prompt elements: Nouns, prepositions, adjectives, adverbs, and references.
Nouns establish subjects and settings, forming the backbone of prompts.
Prepositions define spatial relationships, adding depth to scenes.
Adjectives enhance description, adding visual and emotional depth.
Adverbs modify actions, emphasizing and refining descriptions.
References provide cultural and historical context, enriching scenes.
Strategic word placement enhances prompt effectiveness.
Understanding linguistic elements empowers prompt creation.
Made with HARPA AI
Wonderful information. Thank you!
Cheers pal!
I'm learning prompting and English grammar at same time ;-)
Yah that's what I was hoping for!
This has been a good grammar lessobn for me :)
I'm happy to hear that!
Thank you for the video:)
And thank you for the comment!
LOL, I love how learning the most advanced concepts of modern imaging are incredibly primitive, it's like a 2000 year old circle with hieroglyphics (which I needed the auto spell checker to spell btw). Is there actually some kind of natural language API? there should be that has name recognition for archetypal development IMHO. Like a baby names book for attributes. p.s. to OP, saying how each language item helps direction would be great for a PDF, we know what nouns are but how do the interact in this world. Great tut, thanks!
"saying how each language item helps direction would be great for a PDF" - could you elaborate on this a little please? I'm not sure I understand what you meant. Thanks!
Sometimes I go to chat gpt and ask for the nouns, adjectives in a description. I know lazy. But it works!
No that's a brilliant way to use GPT
Great video. Your info is always apprecieated. Cheers.
And thanks for letting me know!
You are my mj go to!
I'm so glad I could help!
Thank you for everything you do! You are a real Man
Cheers pal! I appreciate you taking the time to comment and support the channel
Don’t forget punctuation. V6 ~totally~ understands punctuation, especially commas.
Yah great point!
Learning Mid-journey V6 prompting is quite analogous to learning English grammar.
It's definitely turned into that!
Thank you for this video!
But I still hope that Midjourney has features like other AI Image platforms. Where we can put photos of our original products, then create an environment for them by midjourney.
I know right? I think that would be so cool
Sheesh! love how he makes his videos. Even if you are already utilizing these techniques. The way he articulates what's happening when you apply these prompting strategies and breaks it down in a simple digestible presentation is very beneficial. To me anyway. Thank you!
I really appreciate the kind words and feedback. I'm super glad my style of teaching works well for you, thanks!
My name is Nolan and I'm here to teach you the basics of grammar. lol.
😂 time is a flat circle
I found MJ prefers the word below rather than beneath
The funny thing about Midjourney is all the small things that change - like maybe in Version 7, it will prefer the word beneath😂 just good to remind yourself of words with similar meanings
loving your videos Nolan@@FutureTechPilot
can you do a video where we can sell maybe the pics?
Yah I've been taking notes on the business side of Midjourney for a while. Not sure when I'll get around to making the video but thanks for the suggestion!
I came for v6 prompts, i stayed for U3 crabcakes
😂 all hail the Crabcake family
@@FutureTechPilot I need me that u3 bro, already spent my standard plan fast rerolls trying to get that Mr crabcake pic 😂
@@CanisoGaming I'll send it to you if you really want it. email me - ftp @ futuretechpilot . com - and tell me which one it is
thx for this really good video. but ist it possible to access the .pdf without entering a malware infested site? would be great, thanks
malware? It's hosted from my own site. I've checked with Norton and it's safe
@@FutureTechPilot sorry but my AntiV block all my attempt to reach this site, even if I let it in the "allowed" list, I could skip two block but then a third one appears and I cannot pass it, "Malware detected, bla bla...." impossible to enter the site. ;(
@@martinweber8299 dang okay. thanks for letting me know. You can email me - ftp @ futuretechpilot . com - and I'll send you the PDFs
So a style (Andy Warhol …) is basically an adjective? Your opinion please.
I considered that to be a reference. But since Andy Warhol is like, extra famous, you could definitely argue it was an adjective haha
Brethern turned this to a English class
😂 I know I feel bad
ok wait, what is the -c and -s doing? I cant find it on their prompt page lol
--c stands for chaos and it adds variety within the grid ... --s stands for stylize and determines how much creative freedom the bot has. I'll work on another cheat sheet to explain it
🎉🎉🎉
Did anyone else notice images #1 in each of "... Green smoke calmly..." and "... violently...", are identical except for the contrast/brightness?
I think that's one of the advantages of seeds - it has a chance to maintain the shape but change what is inside - like a dog -> to a cat maybe
❤❤❤
🤝
👋
🤝
1:58 for real thought you were going to say "open a dictionary." ;)
😂 that would have been a good one
They all left out the most important thing. How to Zoom out in V6 ?
It hasn't been added yet! but they're working on it
I've subscribed to the PDFs but have not received the emails.
Please check your 'spam' folder - maybe because it gets sent to a lot of people, it will end up there - if not, please let me know and I'll look into it!!
@@FutureTechPilot I checked the spam folder, it's empty.
@@FutureTechPilot I've used a different email address and it worked straight away. Thanks
@@LeoL30 Awesome! Sorry for that frustration
It has gotten more annoying and almost never get what I want. A few months ago I used to get almost always what I was looking for with a few word sentence. Now I have to write a paragraph and it doesn’t get close to what I’m looking for.
Dang. Sorry you have that experience. I guess I can suggest that you still use those older versions you had success with
In your prompt example I recognize at least the art of one 3d artist.
oh yah? who does it remind you of
Andy Warhol should be capitalized.
capitalization doesn't matter!
@@FutureTechPilot Try it both ways and find out for sure. This video is essentially a grammar lesson, so why not follow all the rules of grammar?
@@willdwyer6782 cause the bot doesn't see a difference
it seems like the AI is really understanding how humans talk
Yah good point, it's on its way to being perfect
But it can only create white or black women
You can prompt for all kinds of people
Midjourney after so many years now.. still a hugely flimsy tool.. so disappointed that they don’t seem to be making as much progress as we’ve seen previously.. we still need much more precision, we still need much more “stable diffusion”, we need also style tools to be separate from the prompt so you can be more precise about the style and that not to be mistaken with the prompt subject..
I know they're working on things similar to what you're asking for, but I don't think this company is interested in turning it into something complex like Photoshop