OpenAI has unveiled its latest image generating AI, DALL-E 2, which uses a diffusion-based model to gradually transform noise into detailed images based on text inputs. This technology has the potential to revolutionize digital art and allow anyone to illustrate any novel. DALL-E 2 also features variant generation, image inpainting, and image outpainting capabilities, which are demonstrated through five outstanding examples. The AI is able to create high-quality inpaintings and outpaintings by filling in the holes or adding context to the image beyond what is visible in the initial frame. OpenAI plans to deploy this innovative technology to 1 million people in the near future, democratizing art creation for all.
Source : Two Minute Papers | Date : 2022-08-06 16:22:48 | Durée : 00:07:50
➡️ Accéder à CHAT GPT en cliquant dessus
Abonnez vous a Chat GPT
Comments
"Scholars holding on to their papers" 🥰🥰🥰
Well, I've finally located Ren….now where is Stimpy?
that's big for indie game !
That someone generated an image for "colorless green ideas sleep furiously" is cheeky, amusing, and legitimately interesting
I finally remembered who the video speaker sounds like: Peter Lorre, actor known for films like Maltese Falcon (1941)
Why aren't you covering MIDJOURNEY bro???
This democratizes creative endeavours like game development even further.
I'm sorry to ask you this, but why do you have so many tremolos in your voice when you speak English? It sounds weird!
RIP Concept artist, illustrators and design professionals. Artists are becoming art curators.
"Artists losing their job to A.I. and becoming suicidal"
Doing this prompt would be very ironic lol
Me in 2033 – "Computer, generate Mona Lisa with ahegao face"
wow
Hey, horrible accent. But great content. Greetings from Hungary 🙂
Hey…..I like the information…. You give…… in your videos………. And …… I’d like to…. Give you a suggestion… to get move viewers…. And…..watch time….On your videos…. You…..should…… try toooooooo….. use and AI voice… with correct inflection……timing….and….well…pronunciation.. 😂
Dalle 3 might do x amount of fps for y amount of seconds and there might also be 2 prompts one for considerations the other for the scene description
How long before we can get what Neal Stephenson imagined with the Young Lady's Primer in the Diamond Age?
Fantastic video
VFX artist jobs are scary 😪
I guess next level is video. after all it could easily generate 30fps videos with small varation from a source, then v4 will add music and synth voice !
Ya that out-painting does demonstrate AI is creative.
The image in-painting of the videos is unreal. It feels like magic. Wow.
And the out-painting was so cool.
I think that the ladybug just turned to wasp
I have made a living as an artist for over fifty years. I have seen people input my artwork to make variations. This raises many legal and ethical issues. Who owns the art inputted into the system? I wonder if there will be lawsuirs in the future…
5:15 this is too good the layers the complexity the symmetry and the hands of God at the end it’s shocking to know that this was made by a machine
3:40 It wouldn't take that long to notice the trickery… That lady bug has 8 legs!
i think you are ai too😂😂🤣
Absolutely amazing as scary at the same time. There is so much ways that people WILL use it in wrong way. Also, most of artists gonna loose their jobs, when SW could generate many hours or days of their work in seconds.
Pixar will be the first to go…
#4 literally took my breath away
wow
I'm really sorry for concept artists
but holy cow this is useful
Your constant pausing between words and vocal upswing/downswing pattern of speaking is so unnatural I couldn't stand to watch half of the video.
https://youtu.be/TW2w-z0UtQU?t=30 Dall-e does not use the Diffusion model.
After reflecting a bit, I came to the conclusion that AI art in the current state is no longer something I can support, and I will be straight hostile to people using and promoting it in a positive light, I came to learn that much of the training data is stolen from real artists that gave no such authorization for their work to be used like that, artworks are being sampled in whole chunks to the point even the signatures are appearing on the final output, artists are being used and are getting nothing for their work, the ethics side of AI training data is looking like a nightmare considering no artist has ever given permission to have their work used like this, all AI training data should be licensed and used with permission from the actual artists, ideally all artwork generated should be tracked and be able to compensate the artist of the original training data with royalties and proper credits, software engineers committed the mistake of believing that "if its in the internet it's free to use", which seems to simply be art theft with extra steps, if professional artists need to have licenses and credit sources when using photobashing, ai art should be of no exception, ethically the whole thing seems much more condemnable now than a few months ago.
IDK for DALL-E 3, but in a year Google would release combined Imagen/Parti model that would not just put all of the previous generations to shame, but close the "can the AI be creative" question for good.
The elephant in the room is that AI has surpassed all human intelligence and nobody is discussing this.
Whilst this will most definitely "democratise art creation" for the masses, it is also going to destroy thousands of freelance artists' careers. Be careful what you wish for.
I just got access and I can confirm Midjourney is far better. Dall-E 2 was super overhyped.
Could you please also post some of the worst images Dall-e 2 makes. I haven't tested it before but did play a bit with a different image generation AI and some images are just bad.