MidJourney Prompts For Ultra-Realistic Images
Vložit
- čas přidán 27. 02. 2023
- In this video, we explore how MidJourney can create ultra-realistic images. Inspired by a Reddit thread, I go down the rabbit-hole of trying to figure out how to make the most realistic images possible.
Resources Mentioned:
MidJourney: www.futuretools.io/tools/midj...
Reddit Thread: / i_entered_every_fancy_...
Maximize Video: • Create Ai Photos That ...
Linus Tweet: / 1630092895666552833
🛠️ Explore hundreds of AI Tools: futuretools.io/
📰 Weekly Newsletter: www.futuretools.io/newsletter
😊 Discord Community: futuretools.io/discord
🐤 Follow me on Twitter: / mreflow
🐺 My personal blog: mattwolfe.com/
🌯 Buy me a burrito: ko-fi.com/mattwolfe
🍭 My Backgrounds: www.futuretools.io/desktop-ba...
Outro music generated by Mubert mubert.com/render
#MidJourney #AIArt #generativeart - Věda a technologie
*Here is the prompt from the video if anyone wants to try it or a variation of it:*
portrait of an Indian village woman in forest in Himachal Pradesh, clear facial features, cinematic, 35mm lens, f/1.8, accent lighting, global illumination, (add uplight and v4 or v5 yourself because here it turns into an emdash -- )
thanks
thanks a lot
Please do more people realism stuff like this. I was never a fan on the RGB fantasy land graphics but this really piques my interest!
Hi Matt! Glad i could give a bit of inspiration with adding the polaroid! Big fan! Next time try out some 'fujifilm fujicolor' instead of polaroid to really get some old school film images!
You guys are awesome, thank you for all of this updates and tips!
I personally love your channel I always learn something new, I use the upscale you taught me everyday.
Following both of you from the beginning 🫂
Pewds?!
@@jayc6159 thanks Jay! Happy to hear :)
Matt- I really respect your crediting the sources.
Great video brother! I'm trying to upgrade my Midjourney game and this was really helpful ❤
Just amazing... your method to get hyper realistic images works great... always enthusiastic about the quality of your advice and feedback... many thanks for that... Bye (from France)
Thank you for your examples! I loved the explanations of those prompts, they are great!
Hi Matt, I just subscribed to your channel and website after watching your last video. Please keep up the excellent work.
You’re still in the top 5 AI and AI image generation gurus on Planet CZcams. Keep ‘em coming, Matty.👏🌟👍
Sometimes something is so incredible you can only shake your head and/or curse in response. Like how can a tool like this that can generate these types of results actually exist? This was one of those times. Thanks for sharing.
does feel like were straying into "indistinguishable from Magic" territory with these emerging machine learning systems...
Thank you for the great videos. I've watched allot of videos from people covering AI stuff but you are the only one I subscribe to. Great job!
Amazing tutorial. Matt really adds value to the AI world. This is the end of the crazy MidJourney prompting industry as we just need to be smart in using photography terminology (aperture, shutter speed, depth of field, exposure, ISO, white balance, focal length, bokeh, filters, histogram, RAW, composition, exposure compensation, focus, megapixels, bracketing, flash, lens, metering, image stabilization). So, human intelligence comes from understanding these terms. AI will do the rest. This list was generated with ChatGPT and reviewed by an experienced photographer.
I love this
As an experienced photographer myself (25+ years as a pro, and 40+ years total) I'm really skeptical about the utility of most of the items in that list. I'm not shocked that the long involved prompt in the original Reddit experiment could be pruned to
I have noted giving the film type has yielded some amazing results. I love the Kodak Tri-X, and Agfa films. I’m going to start delving into the technicolor and other cinematic films.
Matt, it would be interesting to see how these prompts work with stable-diffusion-webui with v1-14 and v1-5. Hoping to see a comparison soon!
These prompts have taken my "creations" to the next level, outstanding, and thank you Matt!
It’s not a creation and it’s not yours
@@KoralTea Fair use, and my comment was directed at Matt, not you.
@@KoralTea lolololol
Ofcourse Its a Creation in this AI world..We built our home using bricks, cements which were originally manufactured by someone.
Great video! A few things to add to this. If you add 'film grain:': it can get rid of the plastic look that makes it look like it's AI many times. Another cool one to play with is movie films. You can probably remove some of the ending prompts, but it's got some good stuff in it. Also try some previous versions with it for different looks. Prompt: A wide angle movie still from Alien (1979) by Ridley Scott showing [your choice here] | sci-fi and supernatural horror | uneasy and uncanny atmosphere | shot on celluloid with Panavision Cameras | Panavision Lenses | 35mm Film Negative Width | Anamorphic Projection Format | Edited by David Crowther, Terry Rawlings, Peter Weatherley | Cinematography by Derek Vanlint | Critically acclaimed | Oscar winning practical effects --ar 16:9 --v 4
You made my day
Mine too
mindblowing results, thx a lot
I used your prompt and got an absolutely amazing result from 1 of the 4 results. So much fun. Thanks, Jonathan. Great movie, by the way.
@@dirk.landers if you can share it somewhere let me know. I have a lot of fun with this one.
This is amazing! Thanks for sharing the tips. It's greatly appreciated.
Far out, this is insane. Thanks for sharing!
Hi 👋
Thanks for mentioning my reddit post. Glad to help the community 😊
Hai archit can I get your instagram id
Good find Matt. I only used MJ for fantasy collage, landscapes and portraits. This is a good find.
This video is exactly like i was looking for! I'm glad I didn't search few days ago :D
Thanks for interesting tips. I have been also playing with midjourney lately. You definitely can find some edges where it still struggles to provide that mindblowing result. Like "from above" where it surely doesn't have enough resources because people usually don't shoot from above right :) ... Also when recreating portrait from uploaded picture is still not there yet. But i am excited to see what progress the future brings.
Thank you. I'm so glad I found your channel
Insanely cool!! Thanks Matt
Thank you so much for sharing!❤
I've learned so much about prompts from this guy.
Thank you so much for this great video .❤
So now you specify: f1.4 , Summilux lens, Leica color science look, National Geographic look...such fun but now I see you anticipated my suggestion! I shouldn't make comments before watching to the end.
Thanks for the tips will be using them.
Thats amazing mate!
Thank You ! for sharing your knowledge, you are a very good presenter !
great video, many thanks
another great vid man, i have been a subscriber for a few weeks now. in 40 minutes i will be premering my song called "TIMEKEEPER" with a music video using 200+ ai images generated with midjourney! i had so much fun using this technology and hope that people can get lost in this immersive story
Hey Matt, you"re a Master! Thanks a lot!
Your video is useful as usual. I wonder if you could create a video on how to make a realistic portrait of an existing person in midjourney. We want to make a gift to our colleague at my work and give him his portrait created in the style of his favorite painter. Is this even possible?
Great video, Matt. I would love to know more about the prompts for making a realistic avatar of myself as well as a cartoon-style character - would you have some tips for this ? Cheers Inge
Great post!!
Great video! I also tried this in Blue Willow, and it also picks up the Polaroid switch. I think a good video would be to compare MJ, BW, and the other leading AI generators to see which is the clear winner.
For you to do that in Blue Willow, was it free? Or how much did you have pay monthly to be able to use these prompts?
@@ramonep7973 For the time being BW is completely free. There are no limits to image generations, and the account is also free.
@@louisd95714 Thanks for that information bro!!
Hi @mreflow, thank you very much for this amazing video, very interesting that we will be able to work with photo realistic with Midjourney, as I do not have more credit in it, I will try the prompts in Leonardo AI, Bluewillow and in Playground Ai and see what happens, maybe I will be back here and point some URLs to people see, as you know that has a lot of persons that can not afford with the cost of Midjourney.
Editing looking good
Thanks man! My guy is getting the hang of it. :)
You can boot up Blender or any 3d/scene software drag in some assets to set up scene and play with camera settings until you set it just as you want it then take those settings into AI as prompts and u gonna get stunning results. Also if you wanna enhance it even more set up lightening and pull in prompts for type of light, angles and intensity (UE lumen setting work pretty amazing with this) and results are pretty much indistinguishable that they are being done by AI. Give it a go!
Thanks for the tip but i doubt midjourney would recognize SolidWorks Visualize settings, don't know about other 3D programmes.
text prompt are so much faster. Also assests aren't free. This is like giving everybody power of a fiverr artist to everybody. Soon 3d and movies will be there as well. Can't wait for music and research lol.
@@feilox wtf u on about? It's about prompts not about stuff in 3D space. Turn on ur brain !
Got some great photos thanks!
thank you for this awsome prompt. it is a real Gamechanger!!
this defintely blows my mind, its just crazy
Thank you for the helpful information! I've been using Midjourney for about a week or two now and I'm really impressed. I've realized that there are many ways to use it effectively and there's a lot for me to learn. I have a question: Is it possible to give a prompt or command to refine my sketches to a certain level of color or style? Basically, if I provide a rough sketch, can I receive a prompt that guides me to complete the drawing with a certain color or style?"
It's really good at making photos of actual real things and scenarios (like your examples), but I like making images of unrealistic events, futuristic fashion, etc., and it is very hit or miss.
Hi Matt, really enjoyed your video. Recently I was thinking about creating some images of some Indian saints, the kind you normally see in Kumbha festivals in India. But I am curious to know If I can give some actual photos of people to be used as reference while creating the image and achieve some resemblence with them or upload landscape photos to be used as reference while creating a background. Its LIke creating portrait of a person whose features and dess you can describe, and also upload one or more reference photos of the person so that the created photo resembles him. Thank you.
MidJourney MidJourney MidJourney , I feel like the majority of people are being left out. Not everyone can afford to pay for MidJourney. How about making a video on helping us to make images as realistic as possible with other (free) AI generators. Which free AI generators are the best at making the most realistic faces, animals, spaceships, objects, landscapes. And how to prompt them. Thank you Matt for all your cool content ! 💯👍
Have you tried BlueWillow??
LEONARDO AI
@@rittzbyitch3682 Thank you rittz. Aware of that site and the multitude of other free sites. Enough videos of MidJourney realism. What about tutorials on realism for free AI generator sites. Leonardo included.
@@DianeLanglois7744 Yes. Thank you all for suggestions. Just would like to see more tuts for free AI. The majority of tuts are for MidJourney.
Awesome video
Cinematic, 35mm tens, f/1.8, accent lighting, global illumination -uplight --v 4 --q 2 --uplight
Love the video Matt, only thing, your ring bell effects are waaaay to loud.. 🔊
Thank you, thank you, THANK YOU!!!
Thanks Mike, i'm a newbee your tutorials helped me generate my first images xx'
Wow! I grew new brain cells watching your video.
Hermano me podes creer, me salio altiro el prompt, liked and suscribed , tal vez no duerma esta noche me quedo genial un abrazo
What a time to be alive ah?
I have just seen that a photographer named Boris Eldagsen won the first place in the creative category at the 2023's Sony World Photo award and turned down the prize. His argument was to show that normal photography shouldn't compete with A.I. photo and i agree with him. In case you are curious, his creation was named The Electrician from his series Pseudomnesia...A beautiful image nonetheless. 👌
Hi Mr Wolfe,
love the videos and you have got me hooked
In your video of installing the midjourney model into your local stable diffusion install, why do i get such bad bad images? What settings are optimal to get these images i see online?
thank you for your time
*Awesome!*
Love your vids, but the bells are REALLY loud in Airpods :) Thanks, Matt!
Hey Matt Please describe how we can generate studio-style photos the next time.
You can use several camera types + filmstocks to get very realistic stuff. The trick is not to go off the rails and prompting anything MJ could process into a style which makes the picture artifical. It's challenging when you want your subject be part of a subculture (goth, anime, e-girl/boy etc.). Thats what the 2 creators did: simple prompts, no fancy stuff.
Could you use this same technique to do let's say a cinematic movie poster of yourself and your friends in MJ? And how much monthly does this cost in MJ to use these prompts?
Great thnx!
Midjourney works best when prompted with atmospheric attributes, more than subject attributes.
very usefull video thankssssssss.
Can you get these same results with substance as well?
Imagine being able to export these as true RAW files. Insane.
I think pics could be even better with some keywords in negative prompt ("--no" parameter or "::-1" at the end of your weight prompt) and no "--uplight" parameter as it just makes smoother/less detailed pics using Midjourney v4 :)
I mostly use the following negative prompt for portraits : --no inconsistency, bad proportions, undeveloped concept, unrealistic, 3d, bad anatomy, abnormal hands, abnormal fingers, simple, generic, tumour, low quality
In case you wanted to copy and paste this prompt, I wrote it all out for you.
Portrait of (ethnicity) (biological sex) clear facial features, Cinematic, 35mm lens, f1/8, accent lighting, global illumination, --uplight --v 4 -- q 2 -- uplight
Though when I tried it out on my paid standard version of Midjourney, it did not recognize v4 or q2
Amazing stuff. Still a little uncanny but it's all happening so fast. It's wild
Who knows what will be like in the near future, this summer is gonna be fun for sure!!
@@klaustrusselyou were right
Incredible info Matt, thanks a lot. Where can i find Archi? Y entered in reddit but i can't see his work
My husband has been using Midjourney like crazy lately. :) I am slowly learning how to get different AIs help me with my CZcams channel. In my case, AI is a force of good.
great video
Midjourney Prompts collection available for stunning AI Artwork ❤
Hey Matt, Love your channel and your website.
You will help me build prompts for AI generative art systems. There are some rules to follow that I will explain to you:
1. I will use the command "/imagine" followed by a subject I want to render.
2. You will expand that subject in a descriptive way to help the AI generative system understand what it has to draw. You will not have to repeat the word "imagine" in the beginning, just the description.
3. You will append to the description the name o a famous Artist depending on the style you want to give to the picture. Example: "by Stanley Artgerm". The name will be followed by a full stop "."
4. You will add the name of a professional photo camera model.
5. You will add some comma-separated painting parameters useful for the AI generative system to set up the painting for the scene, depending on the kind of subject. You will decide the parameters based on the style of the artist. An example is: "oil painting, old photograph, retro, sketch drawing,".
6. You will define a lighting style for the shot based on the artist and the style we want to achieve. Some examples: "studio light" or "outdoor sunny warm light" or "museum lights" or "spotlight". Feel free to use other lighting setups you know apart from these.
7. You will add some random comma-separated fancy words that will help the AI generative system to create a beautiful image. Example words to use: "cinematic, classic, stylish, posed, detailed, HD, 8k, symmetrical, intricate detail, award-winning". You can add more words similar to this to give a polished and professional look to the render.
8. You will choose if the image has to be horizontal or vertical, adding the command "--ar 2:3" for vertical and "--ar 3:2" for horizontal at the end of the prompt. You will just provide the command without any other indication.
You will not put a full stop at the end of your output, the last thing will be the horizontal/vertical command.
Are you ready to start?
Thank you, man!
bro can you have sa sample prompt
well i was testing the same thing with other text to image generator they also create good result but still lot of catch up to do with midjourney
Excellent
It still has that slightly airbrushed quality. Using the analog style model for stable diffusion produces the most insanely realistic photos.
Do you have a link or a video source?
dude put the prompt in the descriptions. thousands are checking for it
Repeat: f/1.8 critical. f/2.8 also excellent. Amazing difference in quality, with AI and DSLRs, too.
I don't know for you people but as realistic as these are, midjourney faces are so easy to spot for me, there is a very distinctive smoothness on the skin likely due to the upscaling process and I feel like I can distinguish midjourney portraits from other aiart models
What are you viewing the images on? Tablet, desktop...smartphone? Obviously the smaller the screen the more difficult it'd be to distinguish telltale characteristics of an AI-gen image.
@@SilentKnight43 my computer is hooked to a very big 4k screen.
With V5 some images are indistinguishable at first glance. V6 should be out in around 2 months.
Once things are indistinguishable, midjourney will eventually graduate to video. It's going to get silly willy.
Hey man, thanks for the video! Was wondering, how's it going with stock photography on your end? Maybe a follow-up video on that as well? Still uploading only on Adobe Stock or?
I have a followup video coming. It's in the works. :)
@@mreflow awesome, thanks for the feedback!
Midjourney pads the f out of prompts, as much as I dig using it… it’s all about Stable Diffusion. More of a peek behind the curtain. Thanks for sharing!
Clear facial features, cinematic, 35mm lensX f/1.8, accent lighting
Thank you number 1.
wow! Cant believe it took me so long.
This type of technology added to video games would be crazy
I love love love your videos! You've taught me so much about MJ...so fun! I'd like to put in a request that you stop using keyboard typing sounds 😛 I listen to videos on headphones almost exclusively and they are balanced with your voice. It really hurts my ears so instead I listen to the videos quietly and miss some of your words. Just a request if you feel like it, but I'll keep watching regardless! 💜 Thanks for being awesome!!
It’s so real that it looks unreal.
Bokeh is the term for the blurred background with the wide open lens aperture.
Great video. But to make this a business, how do you actually sell those and what would be the business model?
Thanks for the video. Are all these images created by Ai free to use for commercial purposes?
'photographed by Herb Ritts' or some other famous artist works also great. No need for to many fancy words.
Camera settings do a lot of heavy lifting it seems . Also , BBC documentary is a good one to add to a prompt
Using weights would change the game. Commas are at a ratio of 1:1 so the importance of each description fall heavily on itself and the other description. I’d give that a try.
When you removed Agfa Vista, that was the film type. I wonder what results you could’ve gotten.
Hi Matt, any recommendations on using mid journey weights to reference a portrait to get results as close to that reference image. Basically being able to reference a portrait and then mix an environment. I have struggled with the weight parameter and can't get it to work to well. Any advice would be gladly appreciated
You can't use weight for reference images as of now, but it's about 1:1 weight with the text prompt
@@CptDangernoodle Is there a way to get results as closest to the image? Any examples?
@@MFFranchise Using Stable Diffusion doing img2img with low prompt strength. You can't control the strength of the image on Midjourney
@@CptDangernoodle Thanks for the reply. I Will give Stable Diffusion a try
I noticed on the detailed portrait prompt, it had within it --uplight --v 4 --q 2 -- uplight. you didn't mention that second uplight. I wondered if that had anything to do with the results. I think the first one was connected to the -- and the second one had a space after the --. Secondly, I noticed that you did not copy and paste the uplight prompt, are you leaving that ON in your settings or off and on depending on what you are developing? But I'm curious if you leave it on MOST of the time, kind of question?
I dont know if you read these comments, but my preferred platform is my ipad pro. IOS has no right click feature in which to get the url for ones imported pic. Is there anyway around this? Like maybe a photo holding website where the url is not hidden by right-click access?
It looks like NatGeo materials! Cool
WOW!!