**Greg Isenberg** (0:00)
It's a big day, because ChatGPT Images 2 just came out, and we're going to take it for a spin. We're going to actually see which use cases you can use to make money from it, and actually help you build your business. I'm going to go through a bunch of use cases. By the end of this episode, you're going to be just feeling comfortable and knowing how to work with this thing. But that's not all. We're not just going to go through ChatGPT Images 2 in this episode. I'm also going to give you a startup idea that I think someone, please, please, steal. I'm also going to give you an AI tool that I think is really, really interesting that I wish more people were talking about and I think you're going to really like it. And we're going to talk about a framework that people can use to come up with vertical AI agent startup ideas to make money. It's one of the spaces I think are just ripe for opportunity. So I'm going to give away a framework for that. So we've got all that covered in this episode. So it's going to get your creative juices flowing. But first, let's start with ChatGPT Images 2.0. So.
What are the three things that the old one couldn't do? Like, what's actually changed here? Well, the first thing is, it's going to be able to go to 2K resolution. You can get to 3-1 aspect ratios, and it's spitting out eight images per prompt. So that's a big deal, it feels more powerful. The second thing, and I know a lot of people are gonna be happy to hear this, it's rendering dense and tiny multi-language text correctly. So one of the biggest problems with not just ChatGPT Images as a product, but really a lot of the creative LLMs is getting text wrong. Now, Images 2 in my testing, I will say, hasn't been right 100% of the time.
I won't say it's perfect, but I will say it's noticeably better than the last version. Also, the ability to have multi-languages there, Japanese, Korean, Chinese, Hindi, I'm sure a lot of people are going to be happy about that.
It's also the type of creative that you can use 2 for, it's just a lot more vast. Now, you can use it for things like UI, whereas before, I really wouldn't use it for UI, I wouldn't use it even in infographics. It wasn't that good. Packaging, it's actually pretty good at packaging, posters, stuff like that. And the third thing, and for me this is probably the most interesting, is it has what's called thinking mode. So basically, when you prompt Images 2.0, it's not just going to look at your prompt and then just spit something out. It's going to look at your prompt, but also search the web, do some fact checking, and generate up to eight related images that stay consistent with one another. So, kind of obvious in hindsight that it should actually crawl the web, but happy it's finally in there. Thank you, Sam Altman. So, let's actually go through some prompts that I did that I think people should steal. I can include some of them in the description. Let me know. So, the first is my buddy Sahil Bloom has a skincare line called Wild Roman. So, I basically was like, create a visual style reference for Wild Roman, a modern skincare brand for men 25 to 40 So, I basically give it the aesthetic. I give it the film camera. So, I said it's a context T2 film camera. I give it the lighting vibes, right? Shot between 3 to 5 p.m. golden hour. I give it the color palette, warm creams, some bleached terracotta. I give it the mood, you know, Mediterranean lifestyle. And then, of course, I give it the subjects, like who's actually going to be in the photos. Real humans, natural posture, slight imperfection. Important to include slight imperfections, because otherwise you get stock photos. So, that's something I learned over time. So, I asked it to generate 8 example images, a product shot, a lifestyle shot, a hand shot, a detailed texture shot, an environmental shot, a lifestyle portrait, a packaging flat lay, and an ingredient story shot. And Images 2 just cooked. Like, does that look, does that look AI to you? Does that look AI to you? These look like real photos. That product looks like something that people would buy. And, you know, I just think that this, it's really good at this. But you do need to have the aesthetic dial. Then you have to be very specific with 2.0. Like, really, really specific. It, you know, if it does, if you're not specific, what I'm finding is you get, like, really stock looking images. You'll see some of that in some of the other examples I'll give. And then you're just going to quit. So, highly recommend you copy this. You can also use ChatGPT to help you just, you know, maybe you don't know the difference between a Context 2T film camera, 35 millimeter and 40 millimeter. I didn't. So, I used ChatGPT to help me figure this out. So, yeah, this is a good opportunity to create, like, brand books, packaging. Even, you could, you know, there's a startup idea right there, like selling this as a service to companies. I think a lot of people would pay for it. So, the, another prompt I did, this was to try to basically get photos that I can basically use to then plug into Cdance 2 to create ads and commercials. So, one of the things that you're going to need if you're going to create, you know, if you've seen these AI videos, these commercials that are incredible, and I've done tutorials around how to create them with my friend PJs, one of the things you need is a visual direction. So, I said I'm developing a Super Bowl-style ad, 30-second ad for Shopify, and I have this whole story line where an employee hates his boss, and he creates like a football merch line, and, you know, he basically sells it on Shopify, and then he ends up quitting his job full time. And I basically said generate eight different visual directions for the same story, Wes Anderson, Apple shot on iPhone, Nike just do it. And you can see that some of these are good, and you'll see some of them are not good. So like the Wes Anderson, like this is Wes Anderson vibe, like as a reference images, as a set of references, images that you'd actually create and put into a C-dance too. This is really good. Although what I don't like about this is this looks like Jason Schwartzman. So if you listen or sorry, if you watch Wes Anderson films, you know, this just looks like an image of Wes and of Jason Schwartzman, who plays in a lot of his films. So it's just a little too close, right? But you know, the vibe is definitely correct. This looks a lot like Bill Murray again, a little too much. But I think they did a really good vibe check on that. The Apple shot on iPhone, like looks like stock photos. The Nike just do it. I think it looks really good.
19 more minutes of transcript below
Try it now — copy, paste, done:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000763114970
Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000763114970