DALL-E 2, developed by OpenAI, is an advanced AI system capable of turning simple text descriptions into photorealistic images that are entirely new and imaginative. This remarkable technology takes the concept of image generation from text input to the next level with higher resolution, improved comprehension, and an exciting new capability called “in-painting.” In-painting allows DALL-E 2 to realistically edit and retouch photos, seamlessly blending AI-generated imagery with the original based on natural language descriptions.
The technology behind DALL-E 2 involves training a neural network on images and their corresponding text descriptions, enabling it not only to recognize individual objects but also to understand relationships between objects and actions. This innovative development has several key outcomes: it empowers people to express themselves visually in unique ways, helps assess the system’s comprehension, and aids in understanding how advanced AI perceives and interprets our world. However, DALL-E 2 is not without limitations, as it can be influenced by incorrect labels and gaps in its training data.
Despite its limitations, DALL-E 2 showcases the potential of combining human creativity with intelligent AI systems, allowing us to explore new realms of imagination and creativity while expanding our understanding of artificial intelligence’s capabilities in the visual realm.
Subscribe to our Newsletter!
Amazing model, can’t wait for public availability for it!
Well, hope you have a confortable chair xD
@@miguelalvesmiguel7688 Lol. Been in my chair for years sometimes waiting for opeanAI to release models.
@@miguelalvesmiguel7688 It took 2 years for them to release GPT-3 for the public. They should not do the same mistake they did last time, people got impatient to the point of buying rented accounts to use GPT-3.
Make future releases public for god’s sake!
@@miguelalvesmiguel7688 In your avocado chair.
@@jon7707 Well, sure but they need to take as much time as they need to make this publicly available in a way that can’t be abused (they talk about that on their site), and since there’s no company regulating AI we should just be grateful they’re not just throwing it out there and taking their time instead 🙂
Sometimes when i read a novel, lack of images that i have in mind limits my imagination greatly and it could make me quit reading.
Dall-E 2 can provide a lot of images for the story and help people imagine far much better… reading a book full of good imaginations provide more fun than a movie. I hope more people can use it in daily life.
This could help so many people in terms of accessibility to information!
Imagine being able to put in any arbitrary book, and have AI generated voice-over in the language you’re fluent in as well as illustrations, or in the future video representing the story.
That sounds cool to me, though I don’t think book readers want images
Man, I’m with you. English is not my 1st language, so sometimes I really had a tough time to imagine something whenever I read books written in English, especially sci Fi books. If this technology will ever be fully functional and accessible, which I really hope that it will, the 1st novel that I would like to try is Anathem by Neal Stephenson.
yeah
How about dear human beings? Haha…
I find your reflection very interesting, in fact, before reading your comment I have been reflecting on the SAME…
But let me make a CRITICAL statement, that is, QUESTIONING:
Why do YOU BELIEVE that you have a GREAT INCAPACITY, as you mention, to IMAGINE certain things within your PSYCHIC apparatus?
DON’T YOU BELIEVE that NOT being able to do something that EVERY HUMAN BEING DOES is a FULL FAILURE within your psychic apparatus?
That is to say, you are thinking, as a human being who is NOT looking for the ROOT of his inconvenience…
I want you to understand me, THE HUMAN BEING, self-corrupts himself and by NOT knowing how to repair himself, he then looks for TOOLS that make up for HIS INCOMPETENCE…
Do you think it COHERENT that we do that?
Do you think that we SHOULD use technology to SUPPLY our SELF-INDUCED psychological flaws?
You can tell me; I have not decided NOT to be able to imagine. Ok, I accept it, but… You DON’T THINK you have a BIG PROBLEM, because you CAN’T imagine? Let’s say that you have not self-generated said inconvenience, as I have stated before, but even so, you STILL HAVE said inconvenience, right? So…
What do we do about it? Do we look for the ETIOLOGY of said inconvenience, or on the contrary do we use technology to NOT SELF-KNOW OURSELVES and then CONTINUE AVOIDING THE ROOT OF OUR CONFLICTS?
Tell me HONESTLY, what do you think is the most appropriate thing to do in your case? Use technology MEDIOCALLY to AVOID our ROOT problems or on the contrary, use it AFTER we are already HEALTHY?…
Do you get me?…
Kind regards, teacher.
A big greeting from Argentina.
This is such a great video explaining clearly what dalle is, love it when companies spend time and resources creating clear explanations like those! Thank you OpenAI! Glad to share your work on my channel!
10 Best YouTube Bots for Boosting Views, Likes & Subscribers!
This is going to revolutionize all kinds of art and media 🙌🏽 I am hoping in the future Dalle 3 4 or 5 would have the ability to create 3D files & possible animations or even entire films 😮
Wow that sounds super cool
for an entire film you’d have to have access to an absolute monster of a pc lol
@@cursorguy The questin is which tech is gonna advance faster, AI or PC Hardware?
Aka putting artists out of work & devaluing the idea of creativity.
@@TheFatblob25 art is gonna become a hobby instead of a job
Damn I just learnt about the motivation and theory behind GAN today and I thought it was super impressive already. Now this is a whole new level.
I’m afraid that the future generations would no longer be able to tell the difference between reality and ‘just another image created by AI’
the real cookies taste better
There’s already a lot of DALL-E images on the subreddit that i couldn’t tell if it was a real artist or an AI
And there will be movies created by AI with no real actors/directors.
Don’t worry. It will fool you in your lifetime
What is reality?
An outstanding achievement. The best part is that OpenAI is cautious of the hype and points out the pitfalls.
10 Best YouTube Bots for Boosting Views, Likes & Subscribers!
@@mikek545 What do you mean?
Could we create videos with this in the future? Or even interactive game worlds instantly rendered from what you type in in real time? You could basically recreate a lucid dream with that if used together with VR
i can’t decide whether i love this or hate it. it’s honestly a bit creepy, but it’s also so fascinating
?
Same here I feel really conflicted about this stuff! Though I must admit the stuff it can come up with is pretty damn impressive… Which is kinda why I hate it! It almost renders human artistry redundant.
This seriously deserves so much more attention. Just incredible.
I used to be a professional visual artist, but now I mainly write and take photos.
What excites me about this tech is the opportunity to cram more of my personal expression into the amount of life that I have been given on this planet.
Words are my current fav for self expression, but visuals can add a lot to not just the experience of those words, but in discovery of that text. If I publish a blog for instance, I always include a photo, as I find that many folks use that visual to decide whether they will spend time reading the text. Visuals embedded within the text can also provide extra layers of info and expression.
Prob is the time it takes to create both compelling words and images.Computer based design software has sped up the production of visual expression, word processors have sped up the production of words, DAW have…. etc.
I’m looking forward to seeing how much time can be saved using AI tools such as this 🙂
Get ready for not being paid then I’m afraid, dude. Professional Visual Artists are no longer required. These guys are working on text generating AI as well, just to make sure you have zero revenue streams going forward.
So literally you are doomed.
Educating people on what AI actually is and what it’s abilities are is as important as creating the AI itself. Thanks for making videos like this one, which focus on single issues and explain them in layman’s terms.
A artist that I deeply admire described Dalle2 perfectly.Its impressive just like any other simular product. It’s interesting. But most importantly it is to art what reality TV is to the human experience. It’s a soulless commercial tool, made by people that want to milk every last bit of authenticity they can get their hands on for its last dime. Given that OpenAI is basically the NFT-tech bro of AI companies I would rather support any other simular project than this.
They are probably perfected and published by a less greedy company before Dalle2 is available
I’m an artist who wants to work with eather ilustration, comics, and animation, and the only thing I’m good at it art, and visual story telling. This makes me sceard for my future. I literally pray to every God, Godess, Demon, and Angel from every religion, and mythology in the hopes, that at least one of them is reall, and listens my prays, for the survivel of indi projects, and indi animaters, ilustration artists, comic artists, and artists.
@@Alexandraadftxr7052 now that OpenAI was trashed by more ethical services, you can use this tool to your advantage
You guys should do some live streams! That way, everyone can participate dynamically with the usage still being moderated.
These text-to-image generators need to allow for users(individual or team) to add their own subset of image data to the training dataset. By using that new added image set along with their own unique caption data the user could extend the possible image styles. This would allow for even more specific and nuanced image outputs. LOTS of individuals and cgi teams are sometimes going to want as much precision as possible.
I’ve been playing around with the Beta for a day or two now, I’m very impressed with it’s fidelity, but also unfortunately underwhelmed by it’s lack of flexibility/creativity.
A simple prompt like “A bathroom in the style of H.R Giger”, when input in crAIyon delivers exactly what you’d imagine, a bathroom with biomechanical xenomorph looking surfaces and pipes, while Dall-e delivered bathroom’s that are mostly flat metal with odd sinks.
I tried “Spock and Kirk in Among Us” with crAIyon and got a blue crewmate who who’s visor/helmet was shaped like Spock’s bowl cut hair. When I tried the same prompt in Dall-e, I got what looked like 1990s video game box art render of two people, neither looking like Spock or Kirk or anything remotely Among Us related in the background. I tried being more verbose with Dall-e, submitting: “A screenshot from the video game ‘Among Us’, with the crew-members customized to look like Star Trek’s Starfleet officers”. It backfired spectacularly and gave me something that looked like another low budget CW young adults series.
Pardon if I’m sounding overly critical, but I’ve been generating content for months using other (less-refined) neural-nets, and had very high expectations for Dall-E. I will be putting Google’s Imagen through the same type of tests when/if ever it becomes available.
Interesting point you made.
Maybe its network is not connected to pop-culture references, but it just uses nouns / verbs / adjectives and the human logic we give them. It’s very difficult to teach a machine “in the style of” because even in the human world that is very interpretable and hard to define except for very famous cases that launched an entire genre of art.
Maybe if you tell it “in the style of Van Gogh” it will know because he’s such a well known artist?
For example I tried using terms such as surrealism and it worked just fine.
I loved this review! Can you do more videos with a Blue Willow comparison to other image AI generators?
It’s both brilliant and at the same time kinda scary. Certainly it’s technical capability is incredible and it opens the door to all kinds of weird and wonderful new imagery. Though I take some issues with the notion of it “amplifying our creative potential”…. Is that statement really valid when it reduces the need for human artistry and skill to just typing a few words? The same could be said for AI music generators. It’s certainly a clever system but I’m really conflicted over whether it actually enhaces humanity or deminishes us buy making our own creative faculties increasingly redundant. 🤔
Developing artificial intelligence technologies like GPT is an extremely difficult task, but the developers who made it possible have achieved something truly remarkable. I deeply appreciate their work and contribution to the field.
I am an indie game developer and this is really useful especially for games that have images, I even manage to create a game with the use of this ai art.