Steal AI Art? AI Creates Text Prompts from Any Image!

Discover reverse Artificial Intelligence models that convert art or any image to text prompts. We'll test how similar artwork these prompts can generate compared to the reference image with platforms like Midjourney, Stable Diffusion, and Dall-E. In this video, we'll explore the platforms: "img2prompt" and "clip interrogator" to convert your favorite artwork or photograph to text. I hope this video helps. Thank you so much for watching :)
► LINKS TO "IMAGE TO PROMPT" GENERATORS:
1. img2prompt: replicate.com/methexis-inc/im...
2. clip interrogator: colab.research.google.com/git...
► TIMESTAMPS:
00:00 Steal Prompts FAST!
00:38 What's It All About?
01:14 AI Platform #1 (Fast)
01:52 AI Platform #2 (Descriptive)
03:25 Artist Reference Check
03:52 Testing the Prompts
04:06 Testing: Midjourney (img2prompt)
04:39 Testing: Midjourney (with Reference)
05:30 Testing: Midjourney (clip interrogator)
06:02 Testing: Stable Diffusion
06:52 Testing: Stable Diffusion (with Reference)
07:44 Dall-E Demo
08:19 Is It Stealing?
08:29 Recommendations & Thank You!
► CHECK OUT THE OFFICIAL PIXIMPERFECT COMPOSITING PLUGIN: pix.live/compositing
► SUPPORT THE CHANNEL & GAIN PREMIUM ACCESS: / piximperfect
► RECOMMENDED SOFTWARE & CREATIVE RESOURCES:
✅ Official PiXimperfect Compositing Plugin: pix.live/pixcomposite
✅ The Best A.I. Powered High-End Portrait Retouching Plugins: pix.live/retouch4me
✅ Unlimited Photoshop Actions, Plugins, Brushes, Assets, etc: pix.live/envato
✅ The Best Professional Plugin for Color Grading: pix.live/infinitecolor
✅ The Best Professional Plugin for Skin Tones: pix.live/unify
✅ Best Plugin for Landscape & Astrophotography: pix.live/propanel
✅ Try Photoshop for Free: pix.live/photoshop
✅ The App I Use for KZbin Growth: pix.live/ytapp
✅ Music Used in PiXimperfect Videos: pix.live/music
► THE GEAR I USE:
✅ My Graphic Tablet: pix.live/wacompro
✅ Recommended Budget Tablet: pix.live/wacomlite
✅ Primary Microphone: pix.live/mic1
✅ Budget Microphone: pix.live/mic2
✅ Audio Interface: pix.live/audiointerface
✅ Camera: pix.live/camera
✅ Lens: pix.live/lens
► PIXIMPERFECT MERCH STORE:
teespring.com/stores/piximper...
► PIXIMPERFECT ACTIONS:
gumroad.com/piximperfect
► SHARE: goo.gl/IUhnUl
► LET'S CONNECT:
Instagram: / piximperfect
Facebook: / piximperfect
Twitter: / piximperfect

[@smortonmedia]

Hi Unmesh, there is something that is missed in this video that I think is worth mentioning.... but for MidJourney, using either "--test" or the new "--v 4" models give SIGNIFICANTLY better results than the standard --v 3 model. I got amazingly high detailed and dynamic results that are much much closer to the original image.

[@luckyenigmas]

This was insightful being that I play around with other AI image generators... and they are definitely more miss than hit and I keep wondering if it's my prompts haha This was good. Keep rocking!

[@TuPhongMusic]

Wow, this is really interesting! I can't believe how well the AI was able to generate text prompts based on the images. It's amazing to see how far AI has come and the creative possibilities it opens up. I can't wait to see what other innovative uses of AI will be developed in the future!

[@joeboulden7042]

Always enjoy your videos. To the point and informative, but joyful at the same time.

[@mohsen1]

Hi Unmesh, thx for the great video, keep up the good work brother!

[@TOEC]

Adding a reference image is a bit tricky because if you give the reference image full strength, you essentially get the exact same results as the origin image, where with no weight, the origin image is pretty much ignored. Stable Diffusions image to image section is, in my opinion, it's most powerful tool. With a proper prompt and an origin image, you can have results which in their overall composition are radically different from the origin image, yet have a theme from that image. For example, an origin image of blackberries with a proper prompt can result in an image of fine dining china that has blackberries as a theme for their decoration. The Origin image and the Generated image will be 99% different except for that one detail.

[@HMaxTube11]

Excellent exploration. Well done.🌟👏👍

[@gforce1102]

Do Any of these Apps also give you the Camara lens used in a particular Image that is uploaded ??

[@foreropa]

I tried with my own generated images, and it´s not even close of what I said, but then, using the prompts it gives me, it has amazing and beautiful results, so I don´t care if it´s close, even more, I´m happy it gives me new ideas. As an artist, this AI world is really exciting, it gives me SO many ideas for my own creations.

[@Selene-AI]

I totally agree! Sometimes it gives other results than expected but they're even better than what we wanted. I'm embracing that too. Using AI to get inspired for your art is so nice too!

[@mrcead]

Image 2 prompt works better for standalone Stable Diffusion for me at least, clip interrogator gives me results much closer to what I want in Stable Diffusion

[@mifyitaad]

I've been looking for something similar, like forever. Thank you!

[@timothygraue-yg7mi]

This is old but I didn't know about the clip interrogator. Thanks a subscribe well deserved!

[@Queenbeez786]

it keeps getting better

[@sien_visuals6394]

Really. This cycle would never end it seems. Soon we'd see "AI generates new AI system from AI images"

[@LeoGarciaPhotos]

already has been done. DA's new generator adds the generations back into the model.

[@sien_visuals6394]

@@LeoGarciaPhotos okayyyy... Well thanks for the update. I haven't personally explored the world of AI generated "art" so I'm a bit dumb in that field.

[@staceykimbell9324]

Derivative art is my favorite art

[@liz_lore]

Yep, just an endless cycle of soulless crap.

[@StevenLamPhoto]

Thank you for showing us this.

[@aludrenknight1687]

1:28 writing a prompt makes you an artist? That job used to be called an editor.

[@Sichel22]

that whole thing is just a way of making the prompter feel good :D its like someone using google is a liberian or an author or something xD

[@sabbiano1257]

Yes

[@aludrenknight1687]

@@sabbiano1257 "/imagine a guy screaming on a bridge, colors swirl, hands to face, scream, scream, expressionism." - boom, I'm an artist. If anyone uses that prompt they owe me royalties. 🙄

[@sabbiano1257]

@@aludrenknight1687 Do it, no one is denying you that opportunity to make money.

[@aludrenknight1687]

@@sabbiano1257 I did do it, it's right there! If the prompt makes you the artist then you don't need the ai result. Any ai bot will give a result.

[@evefavorse5939]

Really informative video! TFS

[@designholistically7447]

Thanks so much for this video. I have struggled with how to describe my art for a long time. I have even hired people from time to time to write descriptions for me. I will give these tools a go and see how they do.

[@graphosxp]

back in the 90's when i started out with Photoshop 4.0 people said i was "cheating" at art!🤣

[@Plankton_menance]

You are missing the point. Traditional art and digital both involve the same manual process. I still need to sketch, compose, choose my color swatch have skills to do what I do. Even photography involves composition. "Prompt engineers" can just be any rando that copies some text from chat gpt and pastes it in mj sd dalle

[@lxic-bz8hf]

amazing, thank you 🙏🏻

[@cantobioenergetico]

I loved your video. Thanks a lot

[@AlexandreMS71]

They just stole the new created jobs like prompt specialists.

[@Charles_Bro-son]

LOL at this point, the dynamics between emergence and downfall of hole job markets has become impossible to follow.

[@jichaelmorgan3796]

Guess I'll take a sit and wait approach, optimistically assuming the curve will flatten instead of the singularity.

[@sevdev9844]

@@jichaelmorgan3796 Collecting money approach is better.

[@sevdev9844]

Jobs can't get stolen, since they aren't things and are not a property.

[@jamesholdername]

They made new jobs... then kill it. This is life now.

[@jm7720]

Is there any AI platform out there that you can FEED it an image, and it will tell you the best FONTS to use with that image and the best colors for those fonts? So basically an AI typography?

[@user-rb2]

Great channel, thanks!

[@ofelipebrasileiro]

I love yours tutsss 👏🏼👏🏼👏🏼

[@albert93911]

Quick note, if the image is a png, chances are the prompt and config are embedded. Might be worth checking that out first. Thanks for another great video!

[@Caad3ll]

Is there a better file type to use?

[@HaroldsMind]

I'm curious as to PiXimperfect's opinion on using living artist's name to generate AI Artwork? Specially given the gray line of legality, ethics, and morals for how some of these companies (ie: Stable Diffusion, Midjourney, etc.) acquired the image data sets. As well as using living artist's copyrighted artwork to train their AI Art models, and in the end create a for-profit product.

[@pixeldojo]

It is quite simple: since AI steals pieces of real artists' work and combines them it violates copyright law (more precisely - it creates a derivative work).

[@SKD3SIGNS]

@@pixeldojo which gets bypassed with literally any touch up on photoshop since it now had human interaction. I might of paraphrased it but I'm driving in the crazy streets of Miami so sorry lol

[@HaroldsMind]

@@BalkanWizard that's a false equivalence. In a real world you can't attribute protections, laws, or many other things when comparing a person to a system of machine learning. They are not the same, and the law sees them differently. I'd encourage you to look into how Stability AI and others gathered the image sets (which contains copyrighted artwork, and in many cases personal information) though data laundering, by means of shielding themselves from accountability by using nonprofit research entities.

[@gondoravalon7540]

@@pixeldojo > *It is quite simple: since AI steals pieces of real artists' work and combines them it violates copyright law (more precisely - it creates a derivative work).* Well, except no, it doesn't - that's not how the technology works. Images are fed to create data - that is used to train the AI to recognize shapes, combinations of objects, styles, etc - and then discards the data. If it kept the image data, the training dataset wouldn't be 2-10GB, but many terabytes to petabytes in size. It'd literally be impossible to distribute it like it is with some of these training models. Interestingly, creating a derivative work isn't **always** a copyright infringement.

[@HaroldsMind]

@@gondoravalon7540 except when AI models are prone to memorization and over-fitting using copyrighted artwork. The merits of it being a derivative work becomes murky at that point. The "Bloodborne" (Playstation game art) case is a good example of that.

[@Zayne285]

AI bros thought typing words into a prompt was too difficult so they needed an AI to do that as well

[@WinnersForToday]

As you can tell, thanks!

[@kimerapolar]

And now, this is the official announcement that you already have ended your artist career and you seems happy about it.

[@Project_2501]

Give it a few years and it will create whole movies for you just by typing some general frame work. Spooky.

[@soweitsobitter9288]

and that is just the tech that has been released to public.

[@staceykimbell9324]

They already did it with rings of power

[@mjt1517]

@@staceykimbell9324 lawl

[@hombacom]

No no give it a more years and it can do everything you can think of!

[@infographie]

Excellent

[@soulxoart]

Thanks 🤗

[@VideoNash]

Are these images and image variations being pulled somewhere from the web or they completely unique and original?

[@pastuh]

Such job was all my life

[@williamkuroki7683]

Anna Dittmann probably think it's stolen.. but you don't care. That's the magic of the AI, my friend.

[@fathiabushanab3534]

Very interesting :) in reversing the process. l believe if you used the V3 or V4 (--v 4) in midjourney the result may looks way better also specifying the quality while writing the prompt will defiantly affecting the result.

[@dongraparesh]

love you always bro,,,,,thank for make my day

[@rajeshguddaniit01]

informatic tutorial

[@jayvdequito8828]

You can also use Chatgpt to get prompt.

[@kamu747]

Whoa. Makes sense. That thought never occurred to me. Thanks.

[@coincrazy3563]

You are such a boss

[@emonhassan0]

Love you bro ❤️‍🔥❤️‍🔥

[@bikashthapa8622]

Thank you

[@EnriqueRevollo]

WOW great 10x a lot!!

[@mnoir7257]

It's like a snake eating its own tail

[@dgr.visuals]

The more I see what AI can do, the more I'm happy I didn't choose art as a career but as a hobby.

[@dayleywhaley2420]

Lol I left my restaurant job to see if I could do art full time, but the way Ai is advancing it’s hard to see a future in that 😂 I’m making sure I know how to use all the Ai though

[@pastuh]

True. Yesterday I was shocked how fast I got amazing oil painting results :)) Its ok for printing.. but replicate in real is much harder..

[@friendlyvimana]

@@AnthonyGoodley how are you verified with just a k subs?

[@cmerritt6799]

@@pastuh 3D printing is also improving. Is it that hard to imagine 3D printing oil paint onto a canvas? Or, eventually, printing the canvas too...

[@therealOXOC]

@@friendlyvimana check your eyes

[@peerlessportraits]

Do you have any "Keep Prompting" merch?

[@it_designer]

Thanks

[@joshparks3264]

Is there a way to merge a real life image and a illustration by dropping two images together?

[@nic-ori]

Thanks.

[@iamYork_]

Even Piximperfect is doing AI tutorials... what a world...

[@Am4ndeepSingh]

we all can see worse future for our Fav Artists.. then Ai in people's normal jobs too

[@LoydaYoung]

AI will take over everything like Skynet

[@charlesscott4929]

Why not? Be grateful and open your mind to the future.

[@iamYork_]

@@charlesscott4929 im all about it... My YT channel has been dedicated to teaching AI for creative purposes for months now... Everyone is evolving... I work in animation and film professionally and AI has been a major topic for the whole community... I love Unmesh and have been a long time fan... My point was these massive channels are even getting into discussing AI since they also know it 100% is the future and is turning the creative world upside down... It is quite fascinating...

[@iamYork_]

@@BalkanWizard for most... I embrace it and I teach it in my spare time so I am ready for it...

[@monkesh6043]

Science is becoming a curse..countdown begins..Congrats piximperfect

[@leumnanthavong]

Wow! Nice contents channel. I am a new subscriber. Thanks for sharing video. I have learn from yo channel

[@1979fabrice]

Please can you also review Canva's text to image AI drawing tool?🙏

[@Soulartist13]

A.I. another medium to explore. Have fun.

[@starboy001]

Hi could u please tell usabout Unreal Engine. In some video

[@SKD3SIGNS]

Great now I gotta go off on more people copying my prompts lol bad enough I had to quit graphic design 🙃

[@SKD3SIGNS]

@@MKultra81 like human photo manipulation ? Lol

[@mjt1517]

Why would you have to quit graphic design?

[@recnepsyarg]

How do you use image weight with midjourney 4.0? I get an error stating invalid parameter when trying :/ Any advice is greatly appreciated!

[@edmartincombemorel]

image weight is not available at the moment with V4, the default value is 0.5 right now i think, u can always try in older version and then remaster it with V4, Good luck

[@CobraDove1111]

People who make money by selling art are not artists.. people who create art are artists.

[@bboystrech]

God Bless you sir :D

[@nurpiet]

oh shit this is what ive been missing

[@user-jg8ff7oj3g]

Thanks for the video, How can you create AI-based images without distorted faces?

[@sirmiluch6856]

The key is to use negative prompt in Stable Diffusion. Midjourney model is doing it behind the scenes probably, because it's way easier to use and get great results.

[@safrota4]

WoOw ...Excellent

[@MaazKhan-bz3br]

Thanks For Informative Video, But Is There Any Trick To Convert Our Basic Text Commant Into Perfect Prompt ? Plz Make Video On It...Great Job 👌

[@thebangalman8478]

Just wow

[@salmanmehmood470]

Great work 💖

[@blastoise77]

Dude. You're gonna be rich soon with that voice and teaching style. Seriously you just blew me away with that.

[@VideoNash]

thanks

[@mike5636]

Mai Ligne - Back in the Days

[@PaladinCiel]

for the short of images I've been trying to backwards engineer, these were a phenomenal waste of time. Didn't get anywhere close to the results I needed.

[@felixsmittick9128]

This is fun and inspiring. I love playing with these AI tools. Thank you

[@photobyTaps]

Very nice thank you! When using the resulting image, are there copyright concerns?

[@mjt1517]

How would there be? The resulting image is original. You’re not republishing the prompt. You’re generating an image based on the prompt. If you’re nervous, just change a word or an attribute.

[@photobyTaps]

Someday, AI could demand equal rights :-)

[@bunnywar]

Ai scraped copyrighted works, medical files and private photos , so even if you don't get a copy of an existing work there are still a lot of legal issues to worry about

[@AlexeyShort]

I need to do this in batch. Can i?

[@istuff4137]

I don't seem to be able to think of a use function for these AI things? I suppose we can sell the results without investing time?

[@SKD3SIGNS]

There are two types of ai artists the ones who make their prompts from scratch and those who just copy prompts. You can tell which one is hated by the ai art community lol

[@abandonedaccount747]

This depresses me.

[@georhodiumgeo9827]

Can you make an AI to copy the photo and put in into the prompt generator? Is so hard, need more AI.

[@MsRuell]

If ai is hard, try drawing. Maybe that will be easier for you x)

[@NotThatOlivia]

You are sooo late here - try the V4 of the MidJourney, and You'll be amazed! (What You are using here is at best V3, not even a beta!)

[@ilmariangefou6190]

WONDERFUL, thanks for the tips!!!

[@ruofanmo2125]

Hi! Thank you so much for your video, and it is so interesting. I tried to use the code but batch processing from image to text doesn't seem to work. Could you please check it out and maybe talk about how to use it? Your help will be very much appreciated.

[@TheSteveBoyd]

So here we sit, gleefully watching in wide-eyed wonderment as technology makes more and more humans obsolete. We can't help ourselves, because we LOVE science fiction, and it's all becoming a reality right before our very eyes. Your children will tell stories to their disbelieving grandchildren about how they used to write music, paint, and build things before the Silicon Oligarchy made ordinary humans obsolete.

[@hombacom]

Imagine how impossible mobile touch screens and internet was and now are everyday tools. You forget that humans learn and improve too with all tools available for everyone. People are more creative than ever with tech.

[@AM-kb9cz]

@@hombacom AI isn't a tool. AI is a replacement.

[@hombacom]

@@AM-kb9cz Maybe you feel replaceable but don’t spread fear. It’s just software anyone can try to learn and understand it’s limits.

[@AM-kb9cz]

@@hombacom It's not fear, it's a reality. A tool is something a human uses to do work with their skills and according to their will. A screwdriver is a tool. You can use it to screw things. If you ask your employee to screw something, are you doing the job through him? Is he a tool? No, he's doing the work, he's a worker. Now let's say you fire your employee and buy a robot, and you ask the robot to screw something. If you just tell it what to do and you see it doing it completely external to you, are you screwing things using it? Or is it screwing everything while you watch? Yeah the robot in that case isn't a tool, it's a replacement for a human worker. It's the same with AI art, you see. Someone with no artistic training but lots of money needs one or several images done, so he hires an artist or group of artists to do the job. But why do that when they could just ask an AI to do it? He's got no artistic training, he won't be able to tell if something is really good, or just flashy. He may think he knows, but he doesn't. You may say "well then he hires an artist to supervise everything!" Yeah that's called art supervisor and it's one job of many. Still kicking a lot of people out. And I hope no one suggests all artists become art supervisors because that would show a lack of knowledge about hierarchies at the industry Yeah why isn't everyone at mcdonalds a manager? Why isn't everyone at the office the boss? That's laughable. "Then they should just give up art and learn other skills" yeah still a bunch of people are now looking for jobs and the number of jobs offered isn't growing. Wait until new different AIs take other people's jobs and see unemployment rates rise. Super fun for the economy eh? Who's going to pay to see the new 99% AI generated pixar movie? AI is a replacement, not a tool. Saying otherwise is deceitful. Cheers.

[@hombacom]

@@AM-kb9cz AI is a lot of different things so I dont see the point of simplify like that. For someone Photoshop is a tool. If you just “ask the AI” you get often odd or boring average of everything it learned, but you can spend a lot of time to tweak it if you want. But it feels like random one hit wonders and hard to make images that belong to each other. Is it ethically right to train on public images, maybe not. I don’t like AI hype or to make fun of artists. Some jobs are in danger but I don’t see that humans become obsolete.

[@jamiecropley]

AI's working with other AI's working with other AI's 🤝

[@paulspvk6049]

For Midjourney, use Version 4

[@DeZzBeZooked]

how do u use v4?

[@whiteraimentevangelism]

That picture looks like a CorelDraw image. Over the years they have had many pictures of a woman with flowers to advertise their product.

[@parsley8188]

Lol i remember seeing a video by u a while ago. Cool to see ur into ai too

[@simoneaston8107]

Damn you bro

[@zerosalary]

Dall E on another level...liked it

[@cryptoharrypotter4190]

please recommend me that I should buy AI NFT as a investment please help me to select and where to buy and which one? advise me ASAP

[@pluckypluckster]

Sad thing is today deviant art is FULL of AI whether the users admit it or not, its quite obvious. Earlier this year b4 i found out about it(sad i know) i was amazed at all this work people were doing. then i noticed this lil description ai generated and was like -- hmm wait u mean this is all ai created- what is that exactly? and i dove deep. At the surface level its just fun at how easy it is, 2nd level i realize all those phone apps are basically this( i don't own a smartphone- so im kinda behind) :D, 3rd level is actually trying to make your images unique with specific phrases/ words and getting rid of obvious hints that its ai and not a photograph (photography is what i like), i'm still diving deeper trying to find the bottom

[@jamesand8078]

💙🤍 keep work 👍

[@2nd-place]

WE'VE COME FULL CIRCLE GUYS

[@JessiDeerSims]

When I use the reference link, midjourney basically ignores it. Even with the IW prompt... Is there something I'm doing wrong

[@LeoGarciaPhotos]

Make sure there is a space after the .png

[@JessiDeerSims]

@@LeoGarciaPhotos thank you I'll try that!

[@Abhi-tt8pr]

Bhai I'm from India one video on Hindi on how to start AI Art brother love from India ❤️💯

[@walterbobrowski4761]

To quote from my favorite 1994 movie, "The Little Rascals" with regards to creating appropriate txt2art prompts, one needs a "rather extensive vocabulary, a phenomenal grasp of grammar and superlative command of syntax". Or just let 'clip interrogator' do it for you! Thanks for sharing this approach!

[@neoanimates]

♥️

[@KONDRAA]

Amazing ❤❤❤❤❤

[@user-vn3dg9lp2g]

From the best to best where you want to go ?

[@mr.psyduck4093]

Pro sir :D

[@cranberriarts]

Doesn’t adding artist descriptors like that encourage other people to use AI to generate off other artist’s works without their consent?

[@konradjablonski7727]

People are crying over AI art but they don't realize that "Artistic Licensing" has been a thing for centuries lmfao

[@abandonedaccount747]

yep and this is exactly why I will not support this channel anymore.

[@staceykimbell9324]

@@konradjablonski7727 yeah...licensing has generally been a thing that people PAY for

[@cranberriarts]

@@konradjablonski7727 I do know art licensing exists, I do this kind of thing professionally. It’s fine if the artist has licensed their art for Ai programs like this. But the unfortunate fact is a lot of these Ai art programs learn by being given images from artists that haven’t done so, which is the main problem with Ai art. If the artists have sold their art or otherwise given permission via contract for Ai gen purposes, then there’s no problem with using Ai

[@purplevincent4454]

here we go again with the crybaby artists.

[@Miniocalypse]

still waiting for Text to 3d model to STL file.

[@El_The_Best]

How little empathy there is for real artists who create original pieces and make a living from their work.

[@Lioness_of_Gaia]

You're not wrong...

[@El_The_Best]

@@Lioness_of_Gaia The decision to use or not an AI application is up to each person but here there are several factors that cannot be denied, a legislation on the unauthorized use of pieces of art that obviously affect the artists who created them, and the background of the creation of the different AI engines as a race of which one gets the best software regardless of the consequences that affect the artists. Morality is left aside when there is a market, it is the truth, the real artist is in the second place.

[@Lioness_of_Gaia]

@@El_The_Best Yeah, it's a mess. It's hard enough as an artist.

[@Lioness_of_Gaia]

@@El_The_Best You've got some awesome Playlists!

[@bluefernlove]

But artists are already using this to speed their own workflow, without stealing from anyone.