AI Prompting Mastery with the Ultimate AI Image Prompt Recipe
HTML-код
- Опубликовано: 17 май 2024
- *New!* get the ultimate prompting eBook! bit.ly/AI-prompt
Go to my sponsor aura.com/makingthephoto to get a 14-day free trial and see if your personal information has been leaked online. Today, we're doing a deep dive into prompt structure. and I'll give you the Ultimate AI Image Prompt Recipe.
You'll find out what to include in an AI image prompt and if the word order matters and how can you get the AI to pay attention to what you really want!
**Let us know your favorite prompts tricks in the comments below.**
Subscribe for more tips like these bit.ly/MtP-Subscribe
If this video was helpful, remember to like the video and subscribe to the channel. And a Thanks would be much appreciated!
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
VIDEO TIMELINE
0:00 Start
0:36 Parts of a Prompt: Subject
3:00 How to describe people
3:55 Describe context and atmosphere
4:21 Use sensory language
5:14 Describe style
6:00 Does word order matter?
6:26 Does length of prompt matter?
7:32 How to weight AI prompts
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
GET MORE!
►Secrets to Creating Stunning AI Images: Expert Prompts
• Secrets to Creating St...
►20 Art Terms you should be using in Midjourney
• 20 Art Terms you Shoul...
►Making Rainbows: Mastering Color in AI Art
• Mastering Color in AI Art
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
MY FAVORITE TEXT-TO-IMAGE GENERATORS
►MIDJOURNEY
www.midjourney.com
►Leonardo AI
leonardo.ai
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
GETTING STARTED VIDEOS
►►VIDEO: Starting Your Midjourney
• Starting Your Midjourney
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
FOLLOW
►Follow on Facebook / witsendphoto
►Follow on Instagram / jennatwitsend
►Follow on Flickr www.flickr.com/photos/jae_at_...
►Follow on Midjourney (subscribers) www.midjourney.com/app/users/...
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
MAKING THE VIDEO
Images by Jenn Mishra via Midjourney, Lexica, & Leonardo AI
Music by Jenn Mishra avia Garageband
Video created with Canva & Audacity
►CANVA
www.canva.com
►AUDACITY
www.audacityteam.org
►GARAGEBAND
www.apple.com/mac/garageband/
STUDIO EQUIPMENT
►MacBook Pro amzn.to/3KR5awo
►Audio Technica 2020 usb microphone amzn.to/3Zhy1hV
►Stedman PS101 Microphone Pop Filter amzn.to/41BhE10
►BenQ SW271C 27 Inch 4K Photo and Video Editing Computer Monitor
amzn.to/3Ys6Z6g
►Sony A7RV amzn.to/3y8Z175
►Sony FE PZ 16-35mm F4 G amzn.to/3SLmncu
►Neewer Ring Light amzn.to/3mjAxW9
Some of these links are affiliate links. I earn a commission when you purchase products. 
Хобби
I've noticed a more granular description is sometimes needed.
For example "Red Hair", produces fire-engine red in many instances, but "a redhead" works slightly better at producing a more natural color. Then "Iridescent light orange-brown hair" produces an even more natural looking hair, as "redhead" gets the color right, but doesn't describe how it interacts with light very well. "Iridescent" seems to be the secret sauce there, giving it the characteristic edge glow when backlit, which a flat color won't reproduce.
Also "With a subtle lens flare" enhances some realism. Even though you will scarcely see any lens flare in the output, it seems to pull from more actual photographs, and less from digital art that simply looks like a photograph.
I also find that playing with word associations can be beneficial. Instead of describing things in adjectives, associate other objects with the scene, and then place those other objects "behind the camera". This can be tricky to describe sometimes. Imagine you want a closeup photo of a wolf. There's millions of wolf photos, but you want to make it look more like a candid photo taken in the wild, and less like a tamed wolf. Wolves in the wild usually run in packs. So a prompt like "A photo-realistic closeup image of a wolf's face, set against a black background, the other members of it's pack are behind the camera". You have now reduced the set of images it pulls from to be closer to what you want, even if you are not explicitly showing those objects in the generated output. It's a little counter-intuitive, but the results are impressive when done correctly.
Terrific tips! You're right, working with the exact phrasing/term is a must. Sometimes we have to think laterally about the look we want.
Really great information. Also, your visuals are stunning. Thanks so much
My pleasure! Thanks!
Thank you! Looking forward to implementing...
Thanks a bunch! Can't wait to see how you put the tips into action. Keep me updated on your progress!
Thanks Jan! Always so helpful and easy to follow!!! You are a gem!!!!☺☺
😃Thank you!
You deserve a 🥇 for this ai prompt tutorial 👍🏼
You've created an amazing channel! Thank you so much!
Thank you! That means a lot!
Excellent explanation!!!! You have earned my subscription!!! Congrats....
Terrific! I hope you're having fun creating!
Thank you very helpful information!!
Any time!
Beautifully Done!
Thank you very much!
As always, important tips.👍🌟👏
Thanks!
Thanks for this tips! 👍
No problem!
Thank You very informative ♥️👏
Hope this helped! Enjy!
Great again, with lots of helpful information, thanks so much! I like the Chekov quote, cause it really hits the issue of "observing what a thing really looks like" instead of presuming "Ai will know what I mean". It often does not. As a non native English speaker the best words don't jump to my mind but often the shades between a "rough" and a "polished" surface make the difference. Which non-native speaker knows "matt surface" or "dull surface". So I keep my thesaurus open or print out my lists put together with, for example, your help. Your polished and furry help. See, my attribute collection grows candidly with every episode!
😀 Thanks! I'm used to visualizing scenes from photography, but I'm not really used to describing them with words. I've had to practice a lot when prompting AI images! I'm always amazed at what the AI understands. What's holding me back is my own thinking.
@@MakingPhoto Thanks for answering even wordy individuals like myself! ;-)
@@ChristianFuchsBlues 😁
Exactly, i believe we non-native speakers, learn english in a more practical way for the sake of communication not precise description, fancy stuff or flourish with words get cut out from classes, i can use basic words to describe like; a sharp face or a chubby one, thin or thic limbs, tall or short people, but when it comes to the (in betweens) is where i get lost...
Btw, my personal main weakness is describing body positions, man, if you need to make a complex scene where two bodies interact with each other and you have a rusty english like me... Give me nightmares just to think about it
Great tips! Thank you!
Have fun creating, Meredith!
Thanks for your useful advice. It will help to beginner to get start their AI journey. Recently I used Blue Willow and I am really amazed to use it
I like BlueWillow, too! Enjoy creating!
AI is making the world a better writer..we are imagining more words and creating instant reality this is amazing
Yes! I'm learning all sorts of new terms and ways to talk about images.
Well its not exactly us that create any single thing. Its AI and only AI. Human talent is getting smaller in the process.
@@danjohannessondkI disagree. I think it’s making people having to actually think. In order to use AI effectively, you have to THINK
Simply spectacular. Fantastic video. Absolutely marvelous. Unbelievably awesome. Just simply wonderful. Watching from Bangalore, Karnataka State, India. Just discovered your channel today 5th July 2023. Super impressed by your video and needless to say, superbly helpful to everyone who watches your videos. Am certainly going to watch all your 155 videos. One by One. All of them. Keep producing more, keep posting more such awesome videos.
Thank you so much! Enjoy creating!
Very useful video . I tested it in my gothic video for my channel and the results were very good.
Glad it helped. You can really guide your results with prompting.
I personally feel AI tools make my life more easier. Recently I am researching so many AI tools and I am amazed to use Blue Willow that is text to image generator. I can generate unlimited images by using Blue Willow and most importantly it's completely free
Yes. I use BlueWillow, too.
Thank you so much for all of your videos and demonstrations of how you got the images you show. I am new to Ai and feel like I am trying to catch up or if I ever will. May I ask how long have you been working in Ai art forms. I am in Midjourney as a paid member as I wanted my own space as I knew what I was going to produce wouldn't be that great to start :) and I would be trying many different ways to get a picture I liked. I decided that learning to give great "Prompts" would be the first place to put some time into. Ultimately I would like to create images that would be beautiful and fascinating to look at and would work as a piece of print on canvas art or be able to print on all types of items depending on the image. Thank you again. You are very appreciated by this newbie.
Welcome, Carla! AI is all about exploring. Finding what you like. You can make anything you want, but there are tricks as you're finding out. I'm a professional photographer, so I like to use AI to explore digital art that I can't really make with my camera. I lucked into Midjourney just as they opened it for public beta testing back in August 2022. I wrote a review for the website "Expert Photography" and I was hooked!
Great Ideas
Thanks! Hope this helps.
changing the resolution up or down can greatly effect the quality as well almost as if you are focusing with a microscope. resolution adjustment will also get rid of conjoined or extra unwanted body parts.
Good idea - how do you prompt for this?
Amazing mastery of AI prompting! Considering your expertise, you might find ZeroBot AI, the first verbal chatbot, intriguing. It could add an exciting dimension to your content. Looking forward to possibly seeing it in a future video. Keep the AI knowledge coming!
OMG I LOVE ZEROBOT. Its like we're in a movie these days
Interesting! I'll check it out. Thanks!
Great
what prompt texts would I use to restrict the AI image to present for instance a single planet in the sky without the AI giving me multiple planets ? I've tried the obvious use of prompt words like "single" , "solo" etc but no success.
Sometimes when the AI doesn't get what you're asking for, add an image prompt. Find an image online that sort of looks like what you want - even if you have to erase or cut out some of the planets. Then try your prompts again with the image.
Is there a way to keep the original image person's face? It always keeps the representative image but changes the face!
Not on traditional AI image generators. They always generate - create - change. There is a Face Swap tool on Discord, but I haven't played with it yet. In MJ, you can add multiple images to one prompt. This helps. Also, if the face is similar to a celebrities, add this person's name to the prompt. That helps.
I am trying to make a logo for my farm with a horse, a cow, and a goat and for some reason no matter what i type i only get one animal or if i get more than one they are like some kind of hybrid creatures. lol I can't figure out how to fix it
This is a known problem. The AI always wants to combine two or more things. You may have to make the animals separately and combine them in a photo editing program like photopea.
Great content. Can you please do this similar video dedicated to transforming an "existing image" (maybe it only needs to be 5min) E.G. I want to create a photorealistic portrait of an image I have of myself in Leonardo using an iphone image and just change the clothing. But when I tried to do it, it is significantly changing the face of my image. I need this image for an AI video Avatar I want to create for my Linkedin profile.
You need inpainting for this sort of thing. A generative AI will always GENERATE - that means changing. So, you can get a similar face sometimes, but not the same face. A better tool would be Photoshop's new Generative AI. I have a video on this "You Need Photoshop's AI Generative Fill: Let me Show you Why!" ruclips.net/video/2eYeUSiT2-o/видео.html
@@MakingPhoto thank you 😊
@@tonettecarrion5152I’m pretty new to AI art but isn’t this what you would use negative prompts for? Telling the AI what you do not want in the picture?
Where do you get your background music from and is it free?
I use Epidemic sound sometimes and occasionally RUclips audio. But for this video, I constructed the music in GarageBand using loops
I have to say, your videos are always well crafted and informative. If I may give one little bit of critique, would be that the music here is really annoying and too loud to properly focus on the VO, Hope this helps...
Thanks for the feedback!
This was a rather helpful video, but it didn't answer my main question... In general, how do Ai image generators see commas? If I try to describe something in a sentence, but it needs a comma, then will the Ai not understand it? I want to know the specifics about how the commas play into the image generation. If you have a video to link me about this, I'd greatly appreciate it!!!
My understanding is that the AI doesn't care what punctuation you put in. It's more for you than the AI. Some people use commas, others use | or . Though on Midjourney, there are some punctuation that means something specific - like :: indicate weights and { } are permutations. In Stable Diffusion, ( ) indicate weights (( )) is more weight.
@@MakingPhoto Interesting, I'll have to experiment around more I suppose, thanks.
In the future you say vacuum the floor, go to work and when you get home you realise your robot cleaner have moved all the furniture in the house as you didn't specify to only vacuum where there is no furniture in the way.
I'm ok with this. As long as it vacuums the floor
im on MJ5 and do have my seed, but its not reading the command --seed 12345, I even tried --seed:12345, but all in all, its giving an error for the word "seed", thoughts?
Hmm. Your use of the seed parameter in this comment is fine. --seed 1234 Double check that there is a space before the -- Also, is it at the end? Maybe if I see the entire prompt, I'll see something.
there shouldn't be a comma behind the seed number, just a space for the next --
I am really looking forward to this. I overexplain to humans, so it is counterintuitive to me to know how to not waste my AI image credits trying to get what I want. I keep telling the AI that there are two women, one in an employee apron and one in a black jacket (or whatever) and that the employee looks confused. More often than not I get two women, both confused-looking, both in aprons. hopefully after this I can avoid that. :)
A prompt is worth it thousand pictures. 😂
Yes it is!
I'm having trouble creating vampire fangs.
It's funny at 4:10 you used a clip of a woman sniffing Hemlock, which is extremely poisonous and can kill you if ingested 😂
Right now I have a prompt with over 550 letters describing both a person and the place she's in (and some objects), this isn't too much is it?
¯\_(ツ)_/¯ You can put in as much as you want! BUT the AI may not listen to you. It's like talking to a teenager. It will probably pick and choose what it wants to do!
@@MakingPhoto Forgot to mention I'm using fotor.
Is the voice ai generated too?
No. Blah. My voice is just my voice.
Thanks you have confused me more 🥴
So sorry! What are you confused about?
Am I the only one who's prompt doesn't seem to go any further than the first :: ? I can't seem to give weight to certain sections...
Great video!
Thanks! Let me see your prompts. Maybe I can see something.
@@MakingPhoto cinematic still::1 Cleopatra in the style of retrofuturism, Iris van Herpen, jewellery, mythical, surreal::2 in a giant bathtub full of thick white fluid standing in the middle of the room, donkey stands in the room::3 In the background we see an ancient Egyptian cityscape::2 Complimenting pink colors, hyper photorealistic still lifes, extreme details on her face, Colors like Hasselblad, rich and vibrant::1 dramatic Studio lighting, fill light. Shot with Leica M6, 16mm lens, F16.0 --ar 3:2
@@WStroo I think you're saying that you get a cinematic still but not the subject you want. try moving the main subject to the front of the prompt. Move cinematic still towards the end. You don't need a weight (the default weight is 1).
Also try consolidating some of the elements. You may lose some art direction but it may be closer to what you want. Cleopatra in a room with giant bathtub of milk, a donkey, with ancient Egypt in the blackouts.
@@MakingPhoto Putting weights on things gives different results than not putting weights. But not specifying the weight gives a default 1 like this ::
5:42 💀
❤❤😂
is it your actual voice or Ai, just asking? 👀
It's my voice.
❤❤❤❤😅😂😊❤😊❤
Nordic people are not the same, the finns looks like Russians and Norway, Sweden and Denmark are Scandinavians.
The AI are getting better, but will have difficulty with fine tuning looks.
but if u just put words, at some point does not make sense to the AI
its not about the words, its how you make AI understand what is front, back, what details you need in the characters or what exactly should connect with what.. etc
how you can do that... how you build the text in a way that allow to create and shape the things i need specifically?
when to use the symbols ( , . ! / ) etc .. to give the exact reading and separations?
This video has a recipe for you. Start with what you want to see (a photograph, digital art, cartoon), then describe the subject of your image, add details of background.... Most AI don't need the symbols, but it's common to separate sections with commas. Some symbols mean something specific to the AI. In Stable Diffusion, wrapping a section in parentheses ( ) gives it more weight.
❤❤❤❤😅😂😊❤😊❤
❤❤❤❤😅😂😊❤😊❤
❤❤❤❤😅😂😊❤😊❤