OpenAI’s DALL-E 2 - podcast

OpenAIs breakthrough - this time visual

Hello everyone, so in this episode we're going to talk about some new release from Openai. It's actually pretty insane what. What AI is capable of? Uh. So this project is called Dal E2 or Dal E2. So you can describe an image to create or edit it instantly. O basically probably followed open AI before and with that you were able to. You know to write some text and then. Like AI generates more text, so now it generates generates images. So that's basically. Yeah, how it works. There was like a first version, which was OK, but not that good. But this version is like actually really amazing. And maybe it will open a new frontier for so many developers and designers. But As for now, the API is closed and you can only try if you can apply and you can stay on the waiting list. I'm just reading here for Twitter of Maddie Siemens. So images are such an incredible medium for AI collaboration. Usually information rich but slow to create with daily creation and iteration can happen virtually as the speed of sod. Though it's early dice, I think this has the potential to be one of the most natural and powerful of human human computer interactions. And then you can just write something like. Father of monkey. Astronaut Monster ice cream cone digital art. And yes, so then you can just generate all these images. Yeah, it's just incredible what this technology can do. On their website you can go to Openai com that lead to is a new AI system that can create realistic images and art from a description in natural language. Can create original realistic images and art from the text description it can combine concepts, attributes and styles. So for example, you can say something like an astronaut. Riding a horse in a photorealistic style and then you have all the images and then you have different styles of that image. Also, really amazing that Dali too can make realistic adds to existing images from a natural language caption. It can add remove elements so there is some other picture where you need to add a Flamingo and then this Flamingo is just. Really added to. Yeah, to that specific. Image. Uh, yeah. So like in a different shape so it's. Yes, so it's not like. You know, like if you just passed in Flamingo to the some parts of the photo, but it's just depending on where it's located, it just changes shape and then it also can take an image and create different variations of it. Inspired by the original so you can have classic picture like Mona Lisa or Clint or I don't know. Some other famous pictures. And then it just creates variations based on it. So that's actually pretty cool. So yeah, it's amazing. Version one delay was introduced in January 21, but now it's. Additional version so there you have it. Definitely you can try to sign up on the waiting list. But yeah, let's but now it's the API is not open yet, so it's just interesting how. It can change the design industry like you imagine. You can just generate limitless. Photos. And images like maybe you don't need like stock photos anymore. And stuff like that. So just imagine doing that. So yeah, kind of really, really amazing.